WO2000025268A2

WO2000025268A2 - Computerized tomography for non-destructive testing

Info

Publication number: WO2000025268A2
Application number: PCT/IL1999/000320
Authority: WO
Inventors: Avraham Robinson; Alex Roytvarf
Original assignee: Romidot, Ltd.
Priority date: 1998-10-26
Filing date: 1999-06-14
Publication date: 2000-05-04
Also published as: EP1145194A2; WO2000025268A3; KR20010081097A; IL126761A0; JP2002528828A; AU4287599A

Abstract

A method of CT image reconstruction in which different values for at least one reconstruction parameter are used for at least two different portions of the image. Preferably the at least one reconstruction parameter comprises pixel size.

Description

COMPUTERIZED TOMOGRAPHY FOR NON-DESTRUCTIVE TESTING ~

FIELD OF THE INVENTION

The present invention relates to CT imaging techniques, especially for high resolution non-destructive testing. BACKGROUND OF THE INVENTION

Many CT (computerized (axial) tomography) reconstruction techniques are known in the art, and are used when it is desirable to obtain an image of a slice through a body, in situations where it is inconvenient to physically open the body. A most important use of CT is in medical imaging, where, for obvious reasons, it is undesirable to cut open a patient. In a typical CT procedure, a plurality of projections of the imaged slice are acquired and the slice is reconstructed from data acquired for the projections. The projection data may be obtained by trans-illumination of the slice with X-rays. In other medical imaging techniques, the projection data is obtained by emission of radiation from radio-nucleotides introduced into the body.

CT imaging has also been used to some extent for industrial uses, such as non- destructive testing. However, these systems are usually characterized by a relatively low resolution. Medical systems are also usually characterized by relatively low resolution, however, they are generally optimized to minimize radiation exposure and to provide suitable contrasts for imaging low-contrast tissue.

US Patents Nos. 5,450,462, 5,379,333 and 5,400,378, the disclosures of which are incorporated herein by reference, describe a two step CT imaging technique in which an initial scan with no reconstruction is used to determine appropriate radiation levels for an actual imaging scan. It is noted that the thrust of these patents is to minimize the radiation level while maintaining an acceptable noise level. Japanese patent publication 05305077, PCT publication WO 98/33361 and U.S. Patents 5,696,807, 5,228,070 and 5,485,494, the disclosures of which are incorporated herein by reference, describe other methods of determining a desired variation in X-Ray radiation intensity. In particular, a reconstructed slice of an object, at a same or a nearby axial location, may be used to determine the desired radiation for a particular location.

High resolution CT imaging is achieved for small objects using micro-focus X-Ray systems. However, the number of pixels in the image is approximately the same as for other CT imaging techniques. Generally, the complexity of reconstruction increases as the square of the number of pixels, making the reconstruction of large high-resolution images impractical.

Two-dimensional X-Ray photographic imaging for non-destructive testing is also known. Manufactured objects which include cavities are currently inspected by cutting open the object and performing measurements on the cut open object. This inspection is important since the amount of raw material required to manufacture the object is often dictated by manufacturing tolerances of the thickness of a portion surrounding the cavity. Clearly, such inspection cannot be performed on every manufactured object. If CT imaging were to be applied to inspect objects with cavities, two main issues are to be considered. First, a greater than conventional dynamic range is required; and, second, the reconstruction time for a large, high resolution image would be prohibitive.

SUMMARY OF THE INVENTION One aspect of some preferred embodiments of the invention is providing a method of

CT reconstruction which allows fast reconstruction even for large image arrays. Alternatively or additionally, the method allows a reduced memory requirement for such reconstruction. In a preferred embodiment of the invention, the reconstruction method is used for industrial inspection of manufactured objects. An aspect of some preferred embodiments of the invention is providing a method of

CT reconstruction which utilizes a-priori knowledge about an object to be reconstructed. In a preferred embodiment of the invention, the a-priori knowledge is provided from design specifications of the object. Alternatively or additionally, the knowledge is provided from a photograph of the object and/or a slice thereof. Alternatively or additionally, the knowledge is provided from expected manufacturing tolerances and/or expected problem areas in an object. In a preferred embodiment of the invention, the knowledge is used to modify any part of the reconstruction of an image of the object, including, and not limited to, reconstruction technique, reconstruction constraints, reconstruction parameters, reconstruction resolution, data acquisition parameters and/or post processing of an image. Alternatively or additionally, to utilizing a-priori knowledge, an estimated (rough) image may be generated by a fast reconstruction of at least some projection data, for example, by back-projection.

An aspect of some preferred embodiments of the invention is providing a reliable method of reconstructing an image of an object even in the presence of inconsistencies in acquired data. Such inconsistency may be caused by noise. Alternatively or additionally, the inconsistency is caused by near total absorption of radiation by the object in portions of at least one projection. Alternatively or additionally, the inconsistency is caused by partial or negligible absorption of radiation at corners of an object and/or other thin parts. In a preferred embodiment of the invention, such inconsistencies are resolved by enhanced reconstruction of the problematic area and/or by rejecting bad data. An aspect of some preferred embodiments of the invention relates to providing individualized data acquisition and/or reconstruction for different portions of the image slice. Preferably, individual treatment of each portion of the image includes locally setting one or more of pixel size, reconstruction type, reconstruction parameters, initial value, expected deviation and/or certainty level. In a preferred embodiment of the invention, the individual attention is provided by imaging only a difference between an estimation of an image slice and an actual imaged slice of an object. Where little or no difference is expected between the imaged slice and the estimated image, the reconstruction is preferably performed using fewer resources (e.g., memory and CPU cycles). As used herein "reconstruction parameters" refer to parameters of the reconstruction method itself, which when varied, modify the mathematical and/or procedural behavior of reconstruction. For example, in ART, the number of iterations, relaxation coefficients and stopping conditions are examples of reconstruction parameters. In back-projection, examples of parameters are weights to apply to the traces (projections of rays) and a function which selects which traces to apply and which not. It should be noted that the filtering portion of "filtered" backprojection is not considered to be part of the reconstruction, rather, it is a preprocessing step which does not affect the reconstruction but the data used for the reconstruction.

In a preferred embodiment of the invention, portions of the estimated image slice are set to certain a-priori image values, and only other portions of the slice are reconstructed. Alternatively or additionally, different portions of the image slice may be associated with levels of certainty with respect to a-priori image density values of those areas. Alternatively or additionally, different data acquisition methods may be applied for projections or parts of projections containing data from these different image portions. Alternatively or additionally, different reconstruction methods may be used for different portions of the image slice.

An aspect of some preferred embodiments of the invention relates to reconstructing significantly fewer than all the pixels in an image slice. Preferably, pixels whose image value can be estimated with a high certainty level are not reconstructed. Alternatively or additionally, a lower quality reconstruction is used for pixels which have a lesser effect on the final usage of the reconstructed image. An example of such a usage is measurement of a corner, where the sharpness of the corner has no effect on a desired measurement of a wall thickness. Alternatively or additionally, a lower quality reconstruction is used responsive to a lower expect error level in the object manufacture and/or in the image reconstruction. An aspect of some preferred embodiments of the invention relates to a method of selecting those pixels on which reconstruction is to be performed. In a preferred embodiment of the invention, an estimated image (or other representation) of the object is used to determine portions at the object at which a deviation from the estimation may be expected or for which it is important. The amount of expected deviation may be different for each portion of the object, depending for example on feature thickness, location in the object, material and/or manufacturing method. Alternatively or additionally, the expected deviation may be determined responsive to a design-allowed deviation at the location. In a preferred embodiment of the invention, the boundaries of the object are determined. Pixels which are far enough from a boundary, either by virtue of being inside the object (full pixels) or outside the object (empty pixels) are preferably attributed a fixed value. The pixels which are not attributed a fixed value generally define a band around some or all of the object boundaries. However, in some preferred embodiments of the invention, the pixels to be reconstructed may define geometrical shapes other than bands, for example, islands. Preferably, only pixels inside the band areas are reconstructed.

An aspect of some preferred embodiment of the invention is a reduction in reconstruction artifacts. In a preferred embodiment of the invention, artifacts are restricted only to parts of the image which are actually reconstructed and do not affect parts of the image which are assigned a fixed a-priori value. Alternatively or additionally, the reconstructed image is relatively immune to noise, since the noise is also limited to the reconstructed areas only. Alternatively or additionally, the fact that a significant part of the image is known before reconstruction may be used to determine the presence of artifacts and/or noise in the image, for example by comparing a reconstructed image with an estimated image. In a preferred embodiment of the invention, spatial noise, distortions and/or artifact problems are detected on portions of the image whose structure is known. Once the artifacts are recognized, they are preferably subtracted from the entire image. In one example, Compton scatter statistics may be obtained from analyzing portions of an image in which no deviation is expected. Preferably, different statistics are gathered for different parts of the image, so that each reconstructed pixel may be corrected using Compton statistics of a nearby unreconstructed pixel. An aspect of some embodiments of the invention relates to a significant reduction in a computational complexity of reconstruction. As can be appreciated, the number of pixels in a band image is significantly smaller than the number of total pixels in the same image. Possibly, the number of pixels to be reconstructed is reduced from n^ to kn, with a bounded k, which k is preferably determined by a complexity of the object. The achieved reconstruction complexity is preferably nlogn. Additional techniques for still further reducing the number of reconstructed pixels are also described herein. In a preferred embodiment of the invention, the reconstructed image is a binary image. Alternatively or additionally, the reconstructed image is a discrete image, having only discrete image values. Preferably the number of discrete density values is the same as the number of materials in the imaged object, plus one for air. Alternatively or additionally, a greater number of discrete values may be used, for example to correspond to boundary areas, but possibly even a continuous range. However, it is expected that, by limiting the number of density values, a faster reconstruction may be achieved in some preferred embodiments of the invention. An aspect of some preferred embodiments of the invention relates to a multi-step image reconstruction method. In a preferred embodiment of the invention, first an estimated image is provided and then the estimated image ("first image") is used to reconstruct a second image which is more exact, at least in certain portions thereof. In a preferred embodiment of the invention, the estimated image is provided based on design specifications of the imaged object. Alternatively or additionally, the estimated image is provided based on a fast reconstruction of the image, possibly at the expense of quality, preferably using less and/or lower quality data and/or a lower quality reconstruction method, for example, using filtered backprojection, a small number of projections, a lower spatial resolution of the projection data, a larger pixel size (for reconstruction) and/or a shorter data acquisition times. In some preferred embodiments of the invention data is acquired at a large number of projection angles, for generating the estimated image. Alternatively or additionally, the angles used for the estimated image reconstruction are determined by an analysis of a-priori knowledge, such as a CAD design of the object. In some preferred embodiments of the invention, a concrete first image is not reconstructed or otherwise provided. Rather, the second image is reconstructed utilizing a-priori information, optionally utilizing a sinogram of the first image and/or a partially reconstructed first image.

The more exact reconstruction is preferably based on the estimated image (or the a- priori information). Preferably, pixels having values that can be guessed to a high level of certainty in the estimated image, are not reconstructed in the second image. In a preferred embodiment of the invention, after the more exact reconstruction, the projection data used for the reconstruction is cleaned up and a third, even, more exact image is reconstructed. In a preferred embodiment of the invention, further additional clean-up and/or reconstruction steps may be performed, possibly using different clean-up techniques and/or parameters and/or different reconstruction techniques and/or parameters for at least one of the repetitions. In a preferred embodiment of the invention, after a final, more exact image is reconstructed, edges and/or other features of the image are extracted in order to perform measurements. In a preferred embodiment of the invention, the edges are determined to a sub- pixel resolution using moments. Alternatively or additionally to edge determination, the reconstructed image may be used for other measurements and/or uses.

In a preferred embodiment of the invention, the estimated reconstruction uses a filtered backprojection method. Alternatively or additionally, the exact reconstruction and/or the more exact reconstruction use an ART (Algebraic reconstruction technique) reconstruction method and/or its variants. Preferably, the ART technique is iteratively applied. Alternatively or additionally, a different iterative technique, for example Baysian or maximum likelihood, may be applied. Additionally or alternatively, a non-iterative technique, such as backprojection may be applied. In some preferred embodiments of the invention the second, third and/or subsequent reconstructions do not all use a same reconstruction technique. Preferably, the results of a previous reconstruction are utilized as a starting point for the next reconstruction. In a preferred embodiment of the invention, the radiation levels used for each projection to acquire data for the second (and possibly subsequent images, if additional data is acquired for them) are determined based on previous reconstructions. Thus for example, the estimated image may be used to set up desired radiation levels, so that a desired signal to noise level, dynamic range and/or sensitivity of the CT imager are achieved. Preferably, the radiation level is selected to optimize portions of projections which include rays which pass through pixels which are to be reconstructed, possibly at the expense of portions of projections which do not include pixels which are being reconstructed. In a preferred embodiment of the invention, a particular portion of the object may be imaged at several radiation levels, possibly from different angles, in order to better image fine details. In a preferred embodiment of the invention, a sub-section of the imaged object may be singled out for special reconstruction, for example at different radiation levels and/or at a different resolution of data acquisition and/or reconstruction. Preferably, such singling out is a result of analyzing a reconstructed image.

An aspect of some preferred embodiments of the invention relates to setting initial guessed density values for pixels to be reconstructed. In a preferred embodiment of the invention, the pixels to be reconstructed are characterized by being in a band around object boundaries. Preferably, each pixel is associated with a density value responsive to a distance from the nearest boundary(s) and/or other features of an object. Alternatively or additionally, the density value is a function of the width of the band and/or other parameters of the object or portions thereof, for example its moment. These initial density values are preferably used as initial values in an ART type reconstruction method. In a preferred embodiment of the invention, the initial values are chosen in accordance with a moment of the projection data.

An aspect of some preferred embodiments of the invention relates to defining an outer reconstruction boundary "hull". In a preferred embodiment of the invention, the hull is used to clamp density values estimated in a reconstruction step. In a preferred embodiment of the invention, the hull is defined to cover a subset of the image that is bounded by the pixels from which non-zero projection values were obtained. Thus, the hull may be approximated by a polygon, defined by the outermost pairs of radiation rays that yield a non-zero projection value. All the pixels outside the hull only yield zero projection values when a radiation ray passes through them. An initial and/or final density value for these pixels may be set to zero. Alternatively or additionally, during an iteration of reconstruction, an estimated density value may be taken to be a minimum of the value of a previous iteration and a hull value. In one example, the hull value is initially taken to be "one", if the pixel is inside the hull or "zero", if the pixel is outside the hull. Alternatively or additionally, the hull may be defined to include more than two discrete values for calculating the estimated density values, for example based on estimated errors and/or based on an identification of a plurality of materials. Alternatively or additionally, to using the hull to clamp values, the hull may be used to perform other mathematical operations on pixels values, for example multiplication. In some preferred embodiments of the invention, the pixels outside the hull are clamped to a non-zero value, for example, if the imaged object is immersed in an x-ray attenuating fluid. Preferably, the density of the fluid is greater than the (x-ray) density of the object. Alternatively, the density of the fluid is smaller than the density of the object.

An aspect of some preferred embodiments of the invention relates to determining relaxation coefficients for constraint based iterative reconstruction techniques. In a preferred embodiment of the invention, the reconstruction technique is an ART-like reconstruction method, for example, ART, MART, SART, ART-3 and/or other versions of algebraic reconstruction techniques. In a preferred embodiment of the invention, the reconstruction method is applied iteratively and different relaxation coefficients are determined for each iteration, preferably based on results of a previous iteration. In a preferred embodiment of the invention, relaxation coefficients in an ART-like reconstruction method are established as a function of deviation of the inconsistent values. Preferably, the deviation is calculated relative to an allowed deviation gate, for example 3σ. In one example, the constraints are established as a square of the deviation. Alternatively or additionally, the constraints are limited to a maximum value. Alternatively or additionally, the constraints are scaled to the maximum value.

An aspect of some preferred embodiments of the invention relates to discarding a portion of acquired raw data, to increase reconstruction quality. In a preferred embodiment of the invention, projections and/or portions of projections which represent rays which pass through a large amount of material (so their value is very low and/or noisy), are ignored.

An aspect of some preferred embodiments of the invention relates to removing constrains in an ART-like reconstruction process. In a preferred embodiment of the invention, constraints are suspected as being unsuitable if they cause a deviation in a reconstructed image relative to projection data, of a greater magnitude than a predetermined value. Alternatively or additionally, constraints may be "suspect" based on their magnitude relative to a mean and/or other statistical considerations. Alternatively or additionally, the deviation of the constraint over several iterations may be taken into account. In a preferred embodiment of the invention, suspected constraints are rejected if they form only a small percentage of the constraints for a particular direction. Alternatively or additionally, the rejection may be based on the relative deviations from constraints generated from projection data from a same or near angles.

An aspect of some preferred embodiments of the invention relates to generating constraints for an ART-like reconstruction method based on moments of the projection data. Preferably, lower order moments, such as zero and first order are used. Alternatively or additionally, higher order moments may be used.

An aspect of some preferred embodiments of the invention relates to utilizing a non- uniform pixel size in reconstruction. In a preferred embodiment of the invention, a higher pixel resolution is used for pixels of greater importance (e.g., for reconstruction or measurement) and or pixels where an expected error and/or variations in the local pixel values is greater. Alternatively or additionally, pixel size may be determined responsive to an effect of lower pixel resolution on other reconstructed pixels, for example those sharing a same radiation trace. Alternatively or additionally, the projection data may be acquired and/or stored at varying resolution levels. In one example, portions of a projection which only pass through pixels that are of lesser importance in the usage of the image, may be acquired and/or stored at a lower spatial and/or density value resolution.

In a preferred embodiment of the invention, the reconstruction data and/or the projection data are stored using a hierarchical data representation. Preferably, the hierarchical representation maintains a spatial organization of the data. Preferably, the data structure is a Quad tree. Thus, significant data storage is required only for "important pixels". Preferably, higher resolution is provided at band pixels. Preferably a higher resolution is provided for pixels nearer an estimated boundary of the imaged object, possibly related to an expected error level and/or density level. In a preferred embodiment of the invention, the local grid resolution may be changed during reconstruction. Alternatively or additionally, a fixed (but possibly spatially varying) resolution is used. In a preferred embodiment of the invention, the higher pixel resolution is approximately the same as a detector resolution. Alternatively, a higher or lower resolution may be used.

An aspect of some preferred embodiments of the invention relates to minimizing a number of projections required for reconstruction. In a preferred embodiment of the invention, fewer than 360 projections are used to reconstruct the data. Preferably, fewer than 64 or 32 projections are used. In a preferred embodiment of the invention, the projections are selected to be useful projections. In one example, projections angles are rejected if a significant portion of the acquired values are at or near noise levels, as a result of passing through a large amount of attenuating material. The restriction of projection angles may be applied to minimize the number of acquired projections. Alternatively or additionally, the restriction is applied to minimize the number of projections taking part in the reconstruction. In a preferred embodiment of the invention, the object to be imaged and/or its expected errors are analyzed (preferably offline) to determine a small number of projection angles, suitable for reconstruction. An aspect of some preferred embodiments of the invention relates to automatic registration of a to o graphically imaged object to a coordinate system. In a preferred embodiment of the invention, an object is placed on an imaging platform in an arbitrary orientation and/or position. In a preferred embodiment of the invention, a rough scan and reconstruction are used to estimate the orientation. Alternatively or additionally, the position may be estimated by comparing a size of the imaged object with an expected size and a known fan beam geometry. Alternatively or additionally, an identification of the object, for example to better select an estimated image from a library, may be performed based on the results of the rough scan. As can be appreciated, in many cases the raw data of the scan (e.g., a sinogram) may be sufficient for these tasks, without requiring an image to be reconstructed. An aspect of some preferred embodiments of the invention relates to determining a threshold for binarizing an image (or thresholds for setting discrete image values to more than two levels). In a preferred embodiment of the invention, the threshold selected is one in which a zero order moment of the thresholded image matches a zero order moment of some or all of the original projections. Preferably the moment is found by binary searching a range of thresholds between 0 and 1. Additionally or alternatively, a higher order moment may be used for binarizing or otherwise thresholding an image.

There is thus provided in accordance with a preferred embodiment of the invention, a method of image reconstruction for tomographic imaging, comprising: providing an indication of an internal structure of an object to be imaged, which object is of non-animal origin; selecting projection angles for reconstruction responsive to the indication; and reconstructing an image from data acquired at the selected projection angles. Preferably, the method comprises: acquiring data at a plurality of projection angles; and reconstructing a low quality image from the acquired data, where the indication comprises the low quality image. Preferably, the data acquired at selected projection angles comprises a sub-set of data acquired at the plurality of projection angles. Alternatively, the selected projection angles and the plurality of projection angles each include at least one projection angles not found in the other.

In a preferred embodiment of the invention, the data acquired at the plurality of projection angles comprises a sub-set of all data acquired at the selected plurality of projection angles.

In a preferred embodiment of the invention, selecting projection angles comprises selecting an estimated minimum number of projection angles which would yield a suitable reconstruction. Alternatively or additionally, selecting projection angles comprises selecting projection angles responsive to an object complexity. Alternatively or additionally, selecting projection angles comprises selecting projection angles responsive to indications of heavy absorption in the indication of internal structure. Alternatively or additionally, selecting projection angles comprises selecting projection angles responsive to indications of low absorption in the indication of internal structure. Alternatively or additionally, selecting projection angles comprises selecting projection angles responsive to a particular feature in the reconstructed image. Preferably, the particular feature comprises at least one sharp comer.

Alternatively, the particular feature comprises a boundary area of the object. Preferably, the boundary area comprises a boundary between two materials which comprise the object.

In a preferred embodiment of the invention, selecting projection angles comprises selecting projection angles responsive to an incidence angle of traces of the projection angles with features of the indicated internal structure. There is also provided in accordance with a preferred embodiment of the invention, a method of image reconstruction for tomographic imaging, comprising: providing an indication of an internal stmcture of an object to be imaged, which object is of non-animal origin; selecting at least one resolution for data acquisition responsive to the indication; and reconstructing an image from data acquired using the at least one resolution. Preferably, the at least one resolution comprises at least two different resolutions for two different projection angles. Alternatively or additionally, the at least one resolution comprises at least two different for data acquired at a single projection angles. Alternatively or additionally, the resolution comprises a spatial resolution. Alternatively or additionally, the resolution comprises a gray-scale resolution. Alternatively or additionally, selecting at least one resolution comprises selecting an estimated minimum resolution which would yield a suitable reconstruction. Alternatively or additionally, selecting at least one resolution comprises at least one resolution responsive to an object complexity. Preferably, the object complexity comprises a local object complexity.

In a preferred embodiment of the invention, selecting at least one resolution comprises selecting at least one resolution responsive to indications of heavy absorption in the indication of internal stmcture. Alternatively or additionally, selecting at least one resolution comprises selecting at least one resolution responsive to indications of low absorption in the indication of internal stmcture. Alternatively or additionally, selecting at least one resolution comprises selecting at least one resolution responsive to a particular feature of the reconstructed image. Preferably, the particular feature comprises a reference feature used for measurement on the image. Alternatively, the particular feature comprises a boundary area. Alternatively, the particular feature comprises a feature which is itself measured on the image. In a preferred embodiment of the invention, the indication comprises an estimation of internal stmcture of the object. Preferably, the estimation comprises a design specification of the object. Alternatively, the estimation comprises an at least two-dimensional representation of the object. Alternatively, the estimation comprises a previously reconstructed image of the object. Alternatively, the estimation comprises an image of a previously imaged object of similar manufacture.

In a preferred embodiment of the invention, the indication comprises a possible deviation from a desired internal structure. Alternatively or additionally, reconstructing an image comprises reconstructing only a portion of the object, responsive to the indication. There is also provided in accordance with a preferred embodiment of the invention, a method of image reconstruction, comprising: providing projection data of an object, which object is of non-animal origin; and reconstructing an image from the projection data, where a different reconstruction treatment is applied to at least one portion of the image, where the different reconstmction treatment comprises a different reconstruction method for the at least one portion.

There is also provided in accordance with a preferred embodiment of the invention, a method of image reconstmction, comprising: providing projection data of an object, which object is of non-animal origin; and reconstructing an image from the projection data, where a different reconstmction treatment is applied to at least one portion of the image, where the different reconstruction treatment comprises using a different value for at least one reconstruction parameter of a reconstruction method used for the at least one portion. There is also provided in accordance with a preferred embodiment of the invention, a method of image reconstruction, comprising: providing projection data of an object, which object is of non-animal origin; and reconstmcting an image from the projection data, where a different reconstmction treatment is applied to at least one portion of the image, where the different reconstruction treatment comprises reconstructing the at least one portion at a different spatial resolution.

In a preferred embodiment of the invention, the at least one portion comprises at least two portions, each receiving different treatment from each other and from at least a third portion of the image. There is also provided in accordance with a preferred embodiment of the invention, a method of image reconstmction, comprising: providing projection data of an object, which object is of non-animal origin; and reconstmcting an image from the projection data, where a different reconstruction treatment is applied to at least one portion and at least a second portion of the image, such that at least three different treatments are applied to the image, one for each of the portions and for at least another portion of the image.

In a preferred embodiment of the invention, applying a different reconstruction treatment comprises reconstructing at a different resolution. In a preferred embodiment of the invention, the method comprises: providing an indication of an internal structure of the object, where the different reconstruction treatment is applied responsive to the indication.

There is also provided in accordance with a preferred embodiment of the invention, a method of image reconstruction, comprising: providing projection data of an object, which object is of non-animal origin; providing an indication of an intemal stmcture of the object; and reconstmcting an image from the projection data, where a different reconstmction treatment is applied to at least one portion of the image, where the different reconstruction treatment is applied responsive to the indication.

Preferably, providing the indication comprises reconstructing a low quality image of the object. Preferably, the low-quality image is reconstmcted using a backprojection method. Alternatively, the low-quality image is reconstructed using a Baysian method. Alternatively, the low-quality image is reconstmcted using a Fourier transform method. Alternatively, the low-quality image is reconstructed using a Maximum Likelihood method. Alternatively, the low-quality image is reconstmcted using a Maximum Entropy method.

In a preferred embodiment of the invention, the low-quality image is reconstructed using a different resolution grid than used for the reconstmcting an image. Alternatively or additionally, reconstructing an image comprises reconstmcting using an algebraic reconstmction method. Alternatively, reconstructing an image comprises reconstructing using a Baysian method. Alternatively, reconstmcting an image comprises reconstructing using a Fourier transform method. Alternatively, reconstmcting an image comprises reconstmcting using a Maximum Likelihood method. Alternatively, reconstructing an image comprises reconstmcting using a Maximum Entropy method. Alternatively, reconstructing an image comprises reconstructing using a backprojection method.

In a preferred embodiment of the invention, reconstmcting an image comprises reconstmcting using a finer resolution grid than used for the indication. Alternatively or additionally, the special treatment comprises setting up an initial estimate of the image, responsive to the indication. Alternatively or additionally, providing the indication comprises retrieving an indication generated responsive to an imaging of a similar object. Alternatively or additionally, providing the indication comprises providing a design specification of the object. Alternatively or additionally, providing the indication comprises providing a manufacturing specification of the object. Alternatively or additionally, the special treatment is responsive to an estimated distance, of the at least one portion, from at least one edge of the object, where the estimation is based on the indication. Alternatively or additionally, the special treatment is responsive to an expected reconstmction error, of the at least one portion, where the expected error is based on the indication. Alternatively or additionally, the special treatment is responsive to an expected manufacturing error, of the at least one portion, where the expected manufacturing error is based on the indication. Alternatively or additionally, the special treatment is responsive to a confidence in an internal stmcture of the object, where the confidence is based on the indication. Alternatively or additionally, the at least one portion is at least one band surrounding the object. Preferably, the band comprises a region that overlaps at least an outside edge of the object. Alternatively, the band comprises a region that surrounds an aperture of the obj ect.

In a preferred embodiment of the invention, the at least one portion encompasses substantially only a feature of the object. Preferably, the feature comprises an area having a particular density.

In a preferred embodiment of the invention, the special treatment comprises utilizing an estimate for selected pixels inside the at least one portion, instead of reconstructing the selected pixels. Alternatively or additionally, the special treatment comprises utilizing an estimate for selected pixels outside the at least one portion, instead of reconstructing the selected pixels.

In a preferred embodiment of the invention, the special treatment comprises providing a lower quality reconstmction outside the at least one portion. Preferably, the lower quality reconstmction comprises a reconstmction with a greater error. Alternatively, the lower quality reconstmction comprises a reconstmction with a lower spatial resolution.

In a preferred embodiment of the invention, a same pre-processing is applied to pixels inside and outside of the at least one portion. Alternatively, a different pre-processing is applied to pixels inside and outside of the at least one portion.

In a preferred embodiment of the invention, the pre-processing comprises filtering. In a preferred embodiment of the invention, the special treatment is responsive to a level of detail required in the at least one portion. Alternatively or additionally, the special treatment is responsive to measurements to be performed on the at least one portion. Alternatively or additionally, the special treatment is responsive to a material composition of the object. Preferably, the object comprises a composite material and the special treatment is responsive to at least one characteristic of the composite material. Preferably, the at least one characteristic comprises a fiber direction of the material. Alternatively, the at least one characteristic comprises a cell size of the material. There is also provided in accordance with a preferred embodiment of the invention, a method of image reconstruction, comprising: providing projection data of an object, which object is of non-animal origin; providing an indication of an internal structure of the object; and reconstructing an image from the projection data, where the reconstructing comprises only reconstructing pixels from the data in at least one certain region of the image, which region has a shape determined responsive to the indication. Preferably, the at least one certain region comprises at least two non-contiguous regions. Alternatively or additionally, the shape comprises a band shape enclosing at least one area of non-reconstructed pixels. Alternatively or additionally, the shape is determined from boundaries of the object which boundaries are indicated by the indication.

There is also provided in accordance with a preferred embodiment of the invention, a method of image reconstruction, comprising: providing projection data of an object, which object is of non- animal origin; providing an indication of an intemal structure of the object; and reconstructing an image from the projection data, responsive to at least one potential- problem area in the indication. Preferably, the at least one potential-problem area comprises an edge. Alternatively or additionally, the at least one potential-problem area comprises a suspected crack area. Alternatively or additionally, the at least one potential-problem area comprises a suspected void area. Alternatively or additionally, reconstmcting comprises varying a spatial resolution of the reconstruction, responsive to a location of the at least one potential problem area. Alternatively or additionally, reconstructing comprises varying a gray- level resolution of the reconstmction, responsive to a location of the at least one potential problem area. Alternatively or additionally, the method comprises defining areas to reconstruct differently, responsive to the at least one potential problem area. Alternatively or additionally, the method comprises detemiining a local confidence level responsive to the edges.

There is also provided in accordance with a preferred embodiment of the invention, a method of iterative image reconstmction, comprising: providing projection data of an object to be imaged, which object is of non-animal origin; first reconstmcting the object from the projection data; rejecting at least some of the data responsive to the first reconstruction; and repeating the first reconstmction, at least once, after the rejecting. There is also provided in accordance with a preferred embodiment of the invention, a method of iterative image reconstmction, comprising: providing projection data of an object to be imaged, which object is of non-animal origin; generating reconstruction constraints; first reconstructing the object from the projection data, responsive to the reconstruction constraints; rejecting at least some of the constraints responsive to the first reconstmction; and repeating the first reconstruction, at least once, after the rejecting. There is also provided in accordance with a preferred embodiment of the invention, a method of iterative image reconstruction, comprising: providing projection data of an object to be imaged, which object is of non-animal origin; generating reconstmction constraints; first reconstructing the object from the projection data, responsive to the reconstmction constraints; generating relaxation coefficients responsive to the first reconstruction; and repeating the first reconstmction, using the relaxation coefficients.

There is also provided in accordance with a preferred embodiment of the invention, a method of iterative image reconstruction, comprising: providing projection data of an object to be imaged, which object is of non-animal origin; generating reconstruction constraints; first reconstmcting the object from the projection data, responsive to the reconstruction constraints; varying values in the first reconstruction responsive to a majorant distribution function; and repeating the first reconstruction, at least once, after the varying. Preferably, the method comprises: determining the majorant distribution to be zero outside a convex object which is defined by all traces in the projection data which have zero projection values. Preferably, the majorant distribution function comprises at least three values.

In a preferred embodiment of the invention, the method comprises processing the data prior to repeating the first reconstmction. Alternatively or additionally, the first reconstmction comprises a plurality of iterations. Alternatively or additionally, the repeated reconstruction comprises a plurality of iterations. Altematively or additionally, the repeated reconstmction comprises applying a different pre-processing to the first reconstruction of the object, responsive to the first reconstmction. Alternatively or additionally, the repeated reconstruction comprises applying a different pre-processing to the projection data, responsive to the first reconstruction. Altematively or additionally, the repeated reconstmction comprises applying a different pre-processing to the first reconstruction of the object, responsive to an indication of an internal structure of the object. Alternatively or additionally, the repeated reconstmction comprises applying a different pre-processing to the projection data, responsive to responsive to an indication of an internal stmcture of the object.

There is also provided in accordance with a preferred embodiment of the invention, a method of image acquisition of an object, comprising: acquiring a set of projection data; reconstmcting a first image from the projection data using a first reconstmction method; analyzing the image to determine special treatment for portions of the image; reconstmcting a second image of the object, using the analysis, where the second reconstmction is a different reconstruction method from the first reconstruction. Preferably, the method comprises acquiring data for the second reconstmction, responsive to the analysis. Preferably, the data is acquired responsive to a desired image quality in the second image. Alternatively or additionally, the data is acquired responsive to a desired analysis of the second image.

In a preferred embodiment of the invention, the method comprises varying an intensity of ionizing radiation used for the data acquisition, responsive to the analysis. Alternatively or additionally, the method comprises varying a wavelength of ionizing radiation used for the data acquisition, responsive to the analysis. Preferably, the ionizing radiation comprises x-ray radiation. Alternatively, the ionizing radiation comprises gamma radiation.

In a preferred embodiment of the invention, the data is acquired using non-ionizing electro-magnetic radiation. In a preferred embodiment of the invention, the method comprises varying at least one parameter of a detection circuit, responsive to the analysis. Preferably, the at least one parameter comprises a gain.

In a preferred embodiment of the invention, the method comprises selecting at least one element of a detection system, from a plurality of available elements, responsive to the analysis. Preferably, the element comprises a detector. Alternatively or additionally, the element comprises a filter. Alternatively or additionally, the element comprises a collimator.

In a preferred embodiment of the invention, the second reconstruction method comprises an algebraic reconstruction method. Preferably, the algebraic reconstmction method comprises an ART-like reconstruction method.

In a preferred embodiment of the invention, the second reconstmction method comprises a Baysian reconstmction method. Alternatively, the second reconstruction method comprises a Fourier transform reconstruction method. Alternatively, the second reconstruction method comprises a Maximum Likelihood reconstmction method. Alternatively, the second reconstmction method comprises a Maximum Entropy reconstruction method. Alternatively, the second reconstmction method comprises a backprojection reconstruction method.

In a preferred embodiment of the invention, the second reconstruction method uses a different resolution grid than the first reconstruction method. Alternatively or additionally, the first reconstmction method comprises a backprojection method. Alternatively, the first reconstruction method comprises a Baysian method. Alternatively, the first reconstruction method comprises a Maximum Likelihood method. Alternatively, the first reconstruction method comprises a Maximum Entropy method. Alternatively, the first reconstruction method comprises a Fourier transform method.

In a preferred embodiment of the invention, the first reconstmction failed to achieve a satisfactory convergence for at least a portion of the image. Preferably, the failure is determined from an error level. Alternatively or additionally, the failure is determined from unexpected values for reconstmcted pixel values.

There is also provided in accordance with a preferred embodiment of the invention, a method of image acquisition of an object, comprising: acquiring a set of projection data; reconstmcting a first image from the projection data using a first reconstruction method; analyzing the image to determine special treatment for portions of the image; acquiring data for a second reconstmction, utilizing a different data acquisition configuration from a first configuration used for the acquiring a set of projection data, responsive to the analysis, which data acquisition configuration comprises selected elements from an available set of functionally equivalent elements; and reconstmcting a second image of the object. Preferably, the different configuration uses a different optical detector from the first configuration. Alternatively or additionally, the different configuration uses a different filter from the first configuration. Alternatively or additionally, the different configuration uses a different detector circuit from the first configuration.

In a preferred embodiment of the invention, the method comprises post-processing the reconstmcted image using an image-processing method. Preferably, the image processing method is adapted to enhance measurement of features in the reconstructed image.

In a preferred embodiment of the invention, the object is a manufactured object. Preferably, the object is a cast object. Alternatively, the object is an extruded object.

There is also provided in accordance with a preferred embodiment of the invention, CT imaging apparatus, comprising: a source of x-ray radiation; a data acquisition system for acquiring attenuation data corresponding to an x-ray density of an object placed in the system, where the data acquisition system comprises at least one set of functionally equivalent elements; and a controller which selectively selects a particular one of the elements to be used in the data acquisition system, responsive to an analysis performed by the controller, of data acquired through the system, with a different one of the elements.

There is also provided in accordance with a preferred embodiment of the invention, a method of algebraic CT image reconstruction, comprising: providing projection data; generating a constraint on a moment of the data; and reconstmcting an image from the data using the moment constraint. Preferably, the moment comprises a first-order moment. Alternatively, the moment comprises a second-order moment. Alternatively, the moment comprises a higher than second-order moment. There is also provided in accordance with a preferred embodiment of the invention, a method of generating relaxation coefficients for an iterative algebraic reconstmction method, comprising: providing a preliminary image reconstructed with a set of constraints during a given iteration; and setting relaxation coefficients for a subsequent iteration responsive to deviations between the constraints and values in the image. Preferably, setting relaxation coefficients comprises: comparing the deviations to a threshold; and setting relaxation coefficients responsive to the comparison. Preferably, the threshold is a function of statistical properties of the deviations.

There is also provided in accordance with a preferred embodiment of the invention, a method of CT image reconstmction, comprising: providing projection data of an object from a plurality of projection angles, which object is of non-animal origin; back-projecting the data into a data stmcture representing a varying resolution grid. Preferably, the data-structure comprises a hierarchical data structure.

There is also provided in accordance with a preferred embodiment of the invention, an industrial inspection system comprising: a feeder of one or more objects to be imaged, which objects are of non-animal origin; and a CT imager, mechanically coupled to the feeder, which images the one or more objects. Preferably, the source comprises a conveyer belt conveying a plurality of objects. Alternatively or additionally, the object is assembled from sub-components. Alternatively, the object is cast. Alternatively, the object is rolled. Alternatively, the object is injection molded.

In a preferred embodiment of the invention, the feeder comprises a take-up device.

In a preferred embodiment of the invention, the object is machined.

In a preferred embodiment of the invention, the feeder comprises an extruder which extrudes an object in a form of a continuous profile. Preferably, the extmder comprises an extrusion nozzle. Alternatively or additionally, the extruder comprises a shaper.

In a preferred embodiment of the invention, the object consists essentially of one material. Alternatively, the object consists essentially of two materials. Alternatively, the object is composed of a composite material. Alternatively, the object consists of more than two materials.

In a preferred embodiment of the invention, the CT imager is synchronized to image an object responsive to a provision of the object by the feeder. Alternatively or additionally, the CT imager images the objects using a spiral imaging method. In a preferred embodiment of the invention, the CT imager utilizes a method as described herein. There is also provided in accordance with a preferred embodiment of the invention, a

CT imager for industrial imaging comprising: a detector; an x-ray source; an imaging area; and a manipulator which lifts and conveys objects to be imaged to and from the imaging area. Preferably, the manipulator comprises a robotic arm. Alternatively, the manipulator comprises a winch. There is also provided in accordance with a preferred embodiment of the invention, a method of manufacturing quality assurance, comprising: manufacturing an object; imaging the object with a CT imager; measuring features on the image to detect deviations in an internal stmcture of the object from design specifications; and rejecting the object if it does not meet the design specifications. Preferably, the object is cast. Alternatively, the object is extmded. Alternatively, the object is mo Id- formed.

In a preferred embodiment of the invention, the method comprises: analyzing the image to detect at least one material defect; and rejecting the object responsive to the detected defects. Preferably, the at least one defect comprises a bubble. Alternatively or additionally, the at least one defect comprises a void. Alternatively or additionally, the at least one defect comprises a variation in density. Alternatively or additionally, the at least one defect comprises a crack. Alternatively or additionally, the at least one defect comprises a cmst. There is also provided in accordance with a preferred embodiment of the invention, a method of tomographic reconstruction, comprising: acquiring image data from a plurality of projection angles of an object, which object is of non-animal origin; first reconstmcting an image from the image data, using a tomographic reconstruction method; and applying a discretization to the image to convert image values of the reconstructed image into a limited set of allowed image values, where the discretization maintains at least one image property, which at least one image property comprises a moment of the image.

Preferably, the moment is a first-order moment. Alternatively, the moment is a second- or higher- order moment.

In a preferred embodiment of the invention, the limited set of values comprises only two values. Alternatively or additionally, the method comprises second reconstmcting the image, in an iterative manner after the applying a discretization. Alternatively or additionally, the limited set of values substantially corresponds in number of value and in relative values to expected density values in the image.

There is also provided in accordance with a preferred embodiment of the invention, a method of tomographic reconstruction, comprising: providing an object having at least one sub-component with a known geometry, which object is of non-animal origin; acquiring image data of the object from a plurality of projection angles; reconstructing an image of the object using the known geometry of the at least one subcomponent. Preferably, reconstmcting an image comprises: generating a low quality reconstruction of the image; identifying the sub-component in the image; and further reconstructing the image using the identification, as a basis. Preferably, identifying the sub-component comprises identifying a position of the sub-component. Alternatively or additionally, identifying the sub-component comprises identifying an orientation of the sub-component.

In a preferred embodiment of the invention, the further reconstmcting comprises fixing values for pixels in the image, which pixels correspond to portions of the at least one-sub component, which fixing is responsive to the known geometry.

In a preferred embodiment of the invention, the object comprises essentially of the at least one sub-component. Alternatively or additionally, the at least one sub-component comprises two sub-components having known geometries.

BRIEF DESCRIPTION OF THE DRAWINGS The present invention will be more clearly understood from the following detailed description of the preferred embodiments of the invention and from the attached drawings, in which:

Fig. 1 is a schematic illustration of a slice of a spherical object, showing deviations from a design specification;

Fig. 2 is a flowchart of a process of CT image reconstmction, in accordance with a preferred embodiment of the invention; Fig. 3 is a flowchart of a method of reducing reconstruction complexity, in accordance with a preferred embodiment of the invention;

Fig. 4 is a flowchart of a method of selecting projection angles, in accordance with a preferred embodiment of the invention; Fig. 5A is a flowchart of a method of defining a multi-resolution grid, in accordance with a preferred embodiment of the invention;

Fig. 5B is a schematic illustration of an image of an object slice, illustrating a multi- resolution grid defined in accordance with the method of Fig. 5 A; Fig. 6A is a flowchart of a method of generating band limitations, in accordance with a preferred embodiment of the invention;

Fig. 6B is a schematic illustration of a banded image of an object slice, determined in accordance with the method of Fig. 6 A;

Fig. 7 is a flowchart of a method of iterative reconstruction, in accordance with a preferred embodiment of the invention;

Fig. 8A is a flowchart of a method of setting relaxation coefficients, for the iterative reconstmction of Fig. 7;

Fig. 8B is a flowchart of a method of creating moment constraints, in accordance with a preferred embodiment of the invention; Fig. 9 is a schematic illustration of an imaging device in accordance with a preferred embodiment of the invention;

Fig. 10 is a schematic illustration of an embodiment of the invention for quality control of extrusion; and

Fig. 11 is a schematic illustration of a conveyer belt embodiment of the invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

Fig. 1 is a schematic illustration of a slice of a spherical object 20, showing deviations from a design specification. Object 20 includes a wall 22 surrounding a cavity 24. A dotted line 28 indicates a design specification of object 20, while a line 26 indicates a deviation during manufacture. In cast objects, there is no way of accessing cavity 24 for direct measurements, short of destmctively slicing open object 20. A significant fraction of the cost of cast objects is a result of uncertainty in the thickness of inner walls. Since such walls require a minimum thickness, cast objects are often manufactured using a greater than required wall-thickness, to compensate for possible errors. It should be noted that the thicknesses and deviations involved are often small - relevant deviations being less than a tenth of a millimeter, for an object whose cross-section may include several centimeters of material.

In a preferred embodiment of the invention, objects such as object 20 are inspected using X-Ray computerized tomography (CT), which is a non-destructive method of inspection. Preferably, a dedicated CT inspection system is provided. Fig. 2 is a flowchart of a process of CT image reconstmction, in accordance with a preferred embodiment of the invention. The method is first described in general terms, with a more detailed description given with reference to Figs. 3-8.

First, an estimated image of the imaged slice of object 20 is obtained (30). In a prefeπed embodiment of the invention, the estimated image is produced by a low quality imaging of object 20. In one example, a small number of projections, for example 16, is used to reconstruct the image slice. In one example, a filtered backprojection reconstruction method is used. Preferably when using backprojection reconstruction techniques, an edge detection method is performed on the image, since backprojection often generates blurred images.

Additionally or alternatively, it is noted that object 20 is essentially a known object, for example manufactured according to design specifications. In a preferred embodiment of the invention, the design specifications and/or CAD files are analyzed to provide the estimated image. Additionally or alternatively, a low quality data acquisition and/or reconstruction are used to judge a similarity between the design specification and the imaged object, to determine whether the design specifications may be used as a rough estimation of the object slice. In a preferred embodiment of the invention, the sinogram of the acquired data is analyzed to determine a similarity to the design specifications and/or CAD files, and no reconstruction is necessary.

In a preferred embodiment of the invention, a low quality data acquisition and/or reconstmction is used identify which object is being imaged. In a prefeπed embodiment of the invention, such an identification may be performed by matching an acquired sinogram to a virtual sinogram of the design specifications, without reconstmcting an estimated image.

As will be clear from the description of some prefeπed embodiments of the invention with reference to Fig. 2, the greater the known similarity between the "estimated" image and the actual image slice, the less complex the image reconstruction may be made in some case. The estimated image is then preferably analyzed (32) to determine the possibility and/or extent of applying complexity reduction methods, in accordance with a prefeπed embodiment of the invention. In a preferred embodiment of the invention, the analysis includes edge detection of the object boundaries.

Additional projection data may be acquired (34). Preferably the projection angles and/or other parameters of the acquired projections are determined responsive to analysis (32) and/or the currently acquired data. In one example, projection angles at which a radiation trace passes through a large amount of material and is attenuated to a system noise level, are not used and/or data not acquired. In another example, a sensitivity of a detector array may be preset to an expected radiation transmission level. In another example, a source radiation level may be set responsive to an expected attenuation.

After the data is acquired, the data may be analyzed again, so that parameters of reduction in reconstmction complexity (36) may be determined. Preferably, the reduction is achieved by reducing the number of pixels to be reconstmcted and/or their resolution and/or accuracy of reconstruction.

The acquired projection data is then reconstmcted as a reconstmction "A" (38). In a prefeπed embodiment of the invention, an iterative reconstmction method is used, preferably ART or an ART-like reconstmction method. Preferably, the estimated image is used as a starting point for the iterations. Alternatively, a starting point is generated responsive to the estimated image and/or a-priori knowledge. Preferably, a discrete reconstmction method is used. Alternatively, a continuous reconstmction method is used. A description of ART may be found in GT Herman, "Image Reconstmction from Projections", in the "The Fundamentals of Computerized Tomography", 1980 or in Y. Censor, "Finite Series-expansion Reconstruction Methods", Proceedings of the IEEE, vol. 71, No. 3 Mar. 1983, pp. 409-419, the disclosures of which are incorporated herein by reference.

In a prefeπed embodiment of the invention, after reconstmction "A" is completed, an optional second reconstruction step is applied. Preferably, the second reconstruction step includes rejecting data and/or reconstmction constraints which appear problematic (40) and then performing a reconstruction "B" (42). Preferably, the second reconstruction is also an ART reconstruction. It is noted that in some prefeπed embodiments of the invention one or both of the reconstructions may use non- ART or even non-iterative reconstmction techniques. However, one benefit of using ART-like techniques is their suitability for parallel processing. Additionally, in some ART-like technique, there is freedom in selecting the reconstruction starting point.

Alternatively or additionally, to rejecting data and/or reconstruction constraints, other pre-processing operations, especially clean-up operations may be performed on the partially reconstmcted image and/or on the projection data. In one example, a smoothing filter may be applied. Possibly, the pre-processing is responsive to the partially reconstmcted image, for example a determination of problematic areas in the image. For example smoothing may not be applied to these areas. Alternatively or additionally, the pre-processing is responsive to the earlier provided indication of the imaged object. For example, smoothing may not be applied to complex portions of the object or to data from which these portions are reconstructed. When the image slice is reconstructed, edges are preferably determined in the image slice to reconstruct an outline of a slice of object 20 (44). The reconstructed image may then be used (46), preferably, by performing measurements on the image, to determine deviations of the reconstmcted object from design specifications. Fig. 3 is a flowchart of a method of reducing reconstruction complexity (36 in Fig. 2), in accordance with a preferred embodiment of the invention. Although the steps in Fig. 3 are shown in a particular order, it will be clear from the following discussion that these steps may be performed in any order or even that two steps combined into a single step. Also, some or all of these steps may be performed in conjunction with analyzing the estimated image (32 in Fig. 2) or even before an estimated image is obtained (30), for example if an object specification is entered into the system. It should be noted that each of the steps in Fig. 3 may be individually applied. In particular, in some embodiments of the invention, only one or even none of the steps of Fig. 3 are applied.

The steps in Fig. 3 include: (a) selecting only a limited number of projection angles (50);

(b) using multi-resolution grid to set non-constant pixels sizes (52);

(c) generating a band image which is a subset of the complete image slice (54); and

(d) setting reconstruction parameters for various areas of the image slice (56), based for example on a trade-off between image quality and reconstruction complexity or based on an expected local eπor.

Fig. 4 is a flowchart of a method of selecting projection angles, in accordance with a prefeπed embodiment of the invention (50 of Fig. 3). Selecting projection angles may serve two ends, (i) to reduce the number of projections used in a reconstmction, possibly reducing memory and/or CPU requirements; and/or (ii) to use more suitable projections, reducing eπor and possibly assuring a faster convergence. Preferably, some or all of the projection data is acquired only after projection angles are selected. Additionally or alternatively, the projection data may first be acquired and only that data which matches the desired projection angles is used for reconstmction.

In a step 60, the sinogram and/or estimated image and/or a-priori knowledge, such as design data, are analyzed to determine suitable and/or unsuitable angles. In a prefeπed embodiment of the invention, angles at which significant traces are completely attenuated by object 20, are considered to be unsuitable. Preferably, significant traces are those which pass through important pixels (e.g., based on user indication and/or automatic analysis) and/or traces which pass through pixels which do not have many associated traces. Additionally or alternatively, the angles are selected so that the dynamic range of expected values in a single projection is minimized and/or matches system limitations. Alternatively or additionally, the angles are selected so that the traces do not pass through areas which are difficult to reconstmct. Alternatively or additionally, the angles are selected so that portions of the image which are of interest are "illuminated" with substantially independent traces, to whatever extent possible that allows a reasonable image to be reconstructed.

In a prefeπed embodiment of the invention, a preset number of projections is used, for example 64 or 32. The possible projections are preferably ranked (62), for example according to the criteria of step 60. Then, the 32 or 64 best angles are preferably selected (64). An exemplary prefeπed selection criterion is that the selected angles match a certain angular distribution function over the range of possible projection angles. Additionally or alternatively, a criterion is that the projection angles are not clumped together.

Additionally or alternatively, the number of projections may be determined according to which object is being imaged. One method of selecting a number of projection angles is by simulating reconstruction with varying numbers of projections and/or particular projection angles and selecting a set of angles which yields satisfactory results and/or minimizes computation time. Alternatively or additionally, the projections may be selected by analyzing a geometrical shape of the object and/or the imaged slice. In a prefeπed embodiment of the invention, an object to be imaged may be placed in the system at a random orientation and a low resolution acquired image and/or sinogram are used to estimate its orientation, for example, facilitating selection of projection angles and/or registration with an a-priori shape, for example a CAD drawing.

In a prefeπed embodiment of the invention, when a particular projection angle is desired, a plurality of projections may be acquired at near angles, and the best and/or most angularly exact projection used for reconstruction. Thus, a lower angular accuracy is needed.

In a prefeπed embodiment of the invention, the slice to be imaged is preferably selected to match a slice in a design specification. Thus, a smallest amount of deviation should be expected. Additionally, any deviation determined may be more meaningful if the deviation is indicated on a design drawing of the slice. In a prefeπed embodiment of the invention, a report of the results of imaging an object include a visual, statistical and/or location specific analysis of any deviations found.

The other steps (52, 54 and 56) of Fig. 3 are all associated with identifying and/or selectively dealing with particular portions of the image slice. In a prefeπed embodiment of the invention, the selection of projection angles is made responsive to these steps. For example, the above described data acquisition dynamic range may be determined only for traces which pass through important areas. Additionally or alternatively, projection angles may be selected based on how data for these particular portions is acquired, and not necessarily with respect to the entire image.

Fig. 5 A is a flowchart of a method of defining a multi-resolution grid (52 of Fig. 3), in accordance with a prefeπed embodiment of the invention. Fig. 5B is a schematic illustration of an image of an object slice 70, illustrating a multi-resolution grid 72 defined in accordance with the method of Fig. 5 A. In a prefeπed embodiment of the invention, smaller pixels are assigned to portions of the image slice in which a higher imaging resolution is required. Preferably, the pixel size is determined responsive to an expected local eπor and/or range of density values. In a prefeπed embodiment of the invention, object 20 is being imaged to detect deviations from a known design. Thus, a higher imaging resolution is typically required at or near the outline of object slice 70. Alternatively, if a design of object 20 is not known, certain heuristics may be applied to the image, based on task parameters. For example, in some object slices it is not expected there to be an island smaller than a certain size. Thus, based on a rough reconstmction at a resolution of that size, it is possible to determine all of those image areas in which there might be a non-empty pixel. It is noted that in typical object inspection applications, there are only two or three materials in object 20, e.g., metal, air and possibly spacers. Preferably, pixels whose values are relatively certain remain large, while pixels whose values are expected to change are made smaller. Typically, pixels at the boundaries of the object are expected to change while pixels which are further away from the boundaries are expected to remain constant. Fig. 5 A illustrates a method of providing small pixels at object slice boundaries. In addition, these small pixels are preferably preset with an estimated density value, based on the estimated image. Alternatively or additionally, the pixel values may take expected deviations in the object manufacture into account. Minimum and maximum pixel (cell) sizes are set (80). The estimated image slice is divided into pixels of the maximum size and each large pixel is assigned an initial density value (82). For each pixel, if the pixel is not at the minimum allowed size and its (normalized) density is neither 0 or 1, the pixel is split into smaller pixels, preferably four or nine (84). If the pixel size is larger than a certain threshold size Tl, each pixel is assigned a calculated "real" density value (86). Preferably the density value is the amount of "black" (attenuation) in the area covered by the pixel divided by the size of the pixel.

If the pixel size is smaller than or equal to Tl, a virtual density is calculated.

Preferably, the virtual density is the density of the pixel from which the pixel was split off. Alternatively, other density functions may be used, for example, the pixel value being a function of a distance from the boundary. Preferably, Tl is the minimum pixel size.

Additionally or alternatively, Tl may be spatially varying, for example, being a function of a distance from a boundary (or band - see Fig. 6A)

The process is preferably repeated until all the pixels have values of 0 or 1 or are at a minimum size. In a prefeπed embodiment of the invention, all pixels that have the density values of 0 or 1 and are not minimal in size are not changed during reconstmction. The projection data is preferably sensitivity-corrected to accommodate ignoring the pixels, for example, by reducing the attenuation values of the projection data.

In a prefeπed embodiment of the invention, Tl and the min and max pixels sizes are the same over the entire image. Alternatively or additionally, different values are used for different parts of the image.

In Fig. 5B, a grid 72 is shown (for only a portion of the image slice), which is a result of applying the method of Fig. 5 A. Pixels 74 are the original maximum size, pixels 76 are an intermediate size and pixels 78 are a minimum size. In the method of Fig. 5 A, the pixels on the boundary are automatically determined to be of a minimal size. However, in a preferred embodiment of the invention, also other pixels may be set to be the minimal size. Furthermore, in some embodiments and situations, a pixel on the boundary may be assigned a size larger than the minimum size, for example if only a low-quality reconstmction is required at that point. Other methods alternative to that of Fig. 5 A may be used to define the pixel size, in some prefeπed embodiments of the invention. In one example, the pixel size may be defined to be a monotonically increasing function of the pixel's distance, starting for example at a distance of 1 standard deviation (of the expected or allowed errors) from a boundary.

Alternatively or additionally, the method of Fig. 5A may be modified so that a pixel is split into smaller pixels responsive to a required reconstruction quality and/or interest in the pixel and/or in a pixel whose reconstmction is significantly affected by that pixel.

Preferably, the pixels are square pixels and at each step of the method of Fig. 5 A they are split into four equal parts. Alternatively or additionally, non-square pixels may be used, for example hexagons or other polygons. Alternatively or additionally, a five- or nine- way split and/or a non-symmetric split may be used. Alternatively or additionally, the pixels are not aligned with a regular grid. Alternatively or additionally, the pixels have a non-isotropic resolution, for example being higher along one axis of the object than along a different axis. Alternatively or additionally, different pixel grids may be used for different parts of the image, the grids preferably being selected to match image reconstmction requirements. Alternatively or additionally, the pixels may be overlapping. Alternatively or additionally, the pixels are stored as pixel center values, in which pixel values are determined by interpolation between values at pixels centers and/or derivatives.

Fig. 6A is a flowchart of a method of generating band limitations (steps 54 of Fig. 3), in accordance with a prefeπed embodiment of the invention. Fig. 6B is a schematic illustration of a banded portion 100 of an image slice, in accordance with the method of Fig. 6 A. As mentioned above, many deviations in an object manufacture process occur at boundaries of the object. Assuming the deviations to be bounded by some amount, some pixels of an image slice are not expected to include any material (empty pixels). Alternatively or additionally, other portions of the image may be expected to include a full pixel. In Fig. 6B, two bands 104 and 102 are defined around portion 100. The density of a pixel (noting that a boundary pixel may have an intermediate density) is only in doubt in this band. In a central area 103 (bounded by dashed lines), it is expected that all pixels will be full. In an external, distant area 105, it is expected that all the pixels will be empty. These pixels (in regions 103 and 105) are preferably fixed in value and are not modified during reconstmction.

It should be noted that non-zero density values of pixels are allowed outside the band. However, in many cases, only zero values will exist outside the bands. Alternatively or additionally, the allowed values may be a continuous, not necessarily limited to 0 or 1, both inside and outside the band. It is noted that if the full range of deviation is found in a part of the image, the quality of reconstmction at that point may not be high. However, such an object is probably defective, so that exact level of defect may not be important. Thus, larger pixels may be tolerated at the outer reaches of bands, in some embodiments of the invention.

Referring to Fig. 6 A, a first step is to determine "eπor locations", in which deviations are expected or may occur (106). As can be appreciated, different amount of deviations (or density values) may be expected or allowed on different parts of the imaged object. Alternatively or additionally, certain parts of the object may be pre-defined to be especially problematic and requiring a higher quality reconstmction and/or a greater allowed deviation range. In some cases, certain pixels may be determined to have a known density and thus are not reconstmcted. In one example, the imaged object may include a jig having exactly known dimensions, The pixels of the jig are preferably identified and removed from consideration. Alternatively or additionally, certain pixels may be reconstructed even through they are outside defined bands. In one example, in casting, if air bubbles (voids) of a certain size are expected, any wall having a thickness of less than twice that certain size (for example) may be reconstructed to make sure it does not include bubbles. Such reconstruction possibly being less limited with respect to density values and attenuation.

In a prefeπed embodiment of the invention, an expert system is provided which analyses the object slice (estimated image), the manufacturing method and/or previous detected deviations and generates a map of areas of the object which are prone to be problematic. Alternatively or additionally, such a map may be provided by a human expert.

Alternatively or additionally, the band sizes and/or certainty levels (described below) are generated by simulated imaging of the object. In a preferred embodiment of the invention, the system includes a simulator in which different eπor and imaging conditions may be tested to determine, for example, the effect of imaging angle and pixel size of the detection of deviations.

In a preferred embodiment of the invention, the reconstruction and/or small pixel sizes may be limited to a portion of the object on which measurements are performed. Preferably, the projection data for radiation traces which do not pass through these areas is ignored.

In a prefeπed embodiment of the invention, when a multi-resolution grid as in Figs. 5 A and 5B is applied, the band is extended to be aligned with a pixel boundaries and/or pixels of at least a minimal size.

In a preferred embodiment of the invention, the spatial resolution of reconstmction is varied responsive to determination of problematic areas. Alternatively or additionally, the gray-level resolution is enhanced for problematic areas, possibly by increasing the detection sensitivity and/or gain for traces which pass through the problematic areas.

In step 108, each pixel is assigned an a-priori certainty level regarding its value, possibly based on an estimated image. Generally, as shown in Fig. 6B, pixels which are closer to an object have a greater probability of being full than pixels which are further away. The form of this dependency may vary over the image, preferably taking into account local eπor conditions. In addition it may be different for pixels inside the object and for pixels outside the object, for example reflected a greater probability of material being missing outside the nominal edge than there being extra material present. In a preferred embodiment of the invention, a linear dependency on distance from the edge is used to relate a distance with a certainty value. Alternatively, a square or an exponential dependency is used for the relationship between distance and certainty. Alternatively or additionally, a plurality of (possibly nested) bands may be defined, each band having its own certainty level and/or zero- order approximation. In a prefeπed embodiment of the invention, the certainty levels are used as a constraint in the reconstruction process, so that it is more difficult to "fill" a pixel which has a low certainty of being "full".

In some preferred embodiments of the invention, the certainty levels are only used for the first reconstruction, for example to set initial density values. Alternatively or additionally, the certainty levels are changed during the reconstmction. Alternatively or additionally, the same certainty levels are used throughout the reconstruction. In a prefeπed embodiment of the invention, the certainty values are used as a probability field, so that a calculated density of a pixel is a product of a reconstmcted density value and a certainty value. Such a product may be applied to an estimated image between reconstmction iterations. Alternatively or additionally, the certainty level may be a set of values for each pixel, indicating the probability of the density value being within a certain range. Thus, one pixel may have a high certainty of being "1", another pixel a high certainty of being "0.5" (e.g., a different material) and another pixel a high certainty of being "0". In some cases, the certainty levels may comprise a multi-modal distribution function, whereby a pixel is more likely to have a value between 0.1 and 0.2 or 0.7 and 0.8 than any intermediate value.

In a prefeπed embodiment of the invention, different radiation levels are determined for different projections, for example to maximize a detector sensitivity. In a prefeπed embodiment of the invention, the radiation level optimization and/or other imaging parameter optimizations only take into account the imaging of pixels which are inside the bands.

Fig. 7 is a flowchart of a method of iterative reconstmction, in accordance with a prefeπed embodiment of the invention. In a prefeπed embodiment of the invention, some of the projection data is first removed from consideration. Preferably, a hull is generated (110). The hull is preferably defined as a convex geometrical object which suπounds the object and outside of which no pixel of the object can exist. Preferably, projection data from traces which do not intersect the hull are removed from consideration. Preferably, the hull is defined for the object including a band. Alternatively or additionally, the hull is defined only for the object itself, excluding the band, preferably by directly analyzing projection data. Alternatively or additionally, the hull is defined to include a different width of band than used for reconstmction. In a prefeπed embodiment of the invention, the hull is used to define a determinate majorant distribution function for the reconstmcted density. During reconstmction, in each iteration, each pixel may be assigned a minimum between the pixel's reconstmcted density value and the hull value, preferably before preceding with the iteration. Alternatively or additionally, each pixel value may be clipped to an expected value range (e.g., between 0 and 1). In a prefeπed embodiment of the mvention, the hull may include a plurality of values, for different pixels, thus, one pixel may be limited (clamped) to a maximum value of "0.5" and another pixel to the value "0.9". The number of clamping values may depend on the number of materials and/or on the certainty levels. In a prefeπed embodiment of the invention, pixels which fall on (or very near) a boundary in the reconstmcted image are not clamped.

In a prefeπed embodiment of the invention, initial values are set for each pixel which is under consideration (112). Some of these values may be set using the method of Fig. 5A. In a prefeπed embodiment of the invention, the initial values are selected so that a moment of at least some of the projections is maintained. Alternatively or additionally, the moment of all the projections are maintained. The maintained moment is preferably a zero-order moment. However, first, second or higher order moments may also be maintained.

In a prefeπed embodiment of the invention, an iterative reconstructive technique such as ART is used. In a prefeπed embodiment of the invention, each radiation trace is coπected for beam hardening, for example using a look-up table (and/or a previous estimate of the image). Preferably, the look-up table is calculated. Alternatively or additionally, the look-up table is empirically derived during a calibration of the imaging device. In the ART reconstmction technique, relaxation coefficients are preferably used to help the iterations converge to a final image. In a prefeπed embodiment of the invention, every one or more iterative steps (114) may be followed by a step of re-setting relaxation coefficients.

Alternatively or additionally, an intermediate reconstruction may be binarized (or converted to an imager with more than two discrete values) between iterations. Preferably, as described below, such binarization uses a threshold which preferably preserves a moment of the reconstmcted image and/or of the projection data.

In a preferred embodiment of the invention, the iterations of a reconstruction step are repeated until an exit condition is met. In a prefeπed embodiment of the invention, the exit condition is that an absolute eπor between a reconstmcted image (or a binarized version thereof) and the projection data is smaller than a predetermined eπor level. Alternatively or additionally, the exit condition comprises a time and/or computer operation limitation on the reconstruction. Alternatively or additionally, the exit condition comprises determining that the amount of change in eπor over previous reconstmction iterations is smaller than a predetermined amount and/or bounded. The above predetermined amounts may be functions, for example of the count of the reconstmction step being performed, the complexity of the image and/or reconstmction parameters.

In a preferred embodiment of the invention, a measure of local (or global) image complexity is the number of edges crossed by a trace. Another possible measure is the density of comers in an area. Another possibly measure is the average feature size, where a feature may be an edge or a line. Another possible measure is a standard deviation of local image properties. Possibly, an image complexity measure is a combination of one or more of the above measures and/or of image complexity measures which are known in the field of image processing. In a preferred embodiment of the invention, the reconstruction for one part of the image may be paused or stopped while reconstmction continues for a second part of the image, for example based on the error in one part of the image being acceptable and the eπor in the second part being unacceptable. Preferably, the image is partitioned so that the traces intersect only one part and not the other. Alternatively or additionally, different reconstruction techniques may be applied to different parts of the image. Alternatively or additionally, different techniques may be applied to different image slices and/or different iterations.

As described with reference to Fig. 2, the reconstruction may comprise two or more sets of iterations, in-between which problematic constraints may be removed (40 in Fig. 2). In a prefeπed embodiment of the invention, after a certain number of iterations, (e.g., 20) certain constraints may be identified as being problematic. In a prefeπed embodiment of the mvention, a constraint is so identified if it deviates from the reconstructed image by more than a threshold value. Preferably, the threshold value is a function of the number of iterations and/or the mean and/or other statistical properties of the deviations. In a preferred embodiment of the invention, statistical properties are calculated over the deviations for this (for which the constraint is determined) and/or other parts of the image or for the entire image. Alternatively or additionally, the statistics are of the projection data for this and/or other parts of the image or for data for the entire image. Alternatively or additionally, the statistics are of previously reconstructed image values in part or all the image. Alternatively or additionally, to identifying problematic constraints at a certain iteration, the identification may be a function of statistical properties of deviations for that constraint over several iterations. For example, a constraint for which the deviations cycle between extreme values, may be a problematic constraint. In a preferred embodiment of the invention, if problematic constraints do not comprises a significant portion of the total path of a certain projection direction, the constraints are removed. Preferably, a significant part is defined to be more than 30, 20 or 10% of the data. Alternatively or additionally, if sufficient data is available at nearby projection angles, the constraints may be removed even if they form more than 30% of the data. Alternatively or additionally, in some cases a problematic constraint may be provisionally removed and if this removal does not aid convergence, the removal may be reversed.

Alternatively or additionally, problematic constraints may be identified based on the existence of unexpected patterns in a partially reconstmcted image. In one example, an island may be an unexpected feature. In another example, a spike i.e., a pixel with value 1 surrounded by pixels with value zero may be assumed to be an eπor. In another example, a sharp comer is often unrealistic. Alternatively or additionally, the unexpected patterns may be determined based on the manufacturing technique and/or other a-priori knowledge, such as CAD drawings

In a preferred embodiment of the invention, the system of constraints is adjusted by eliminating the strongest constraint (e.g., the most limiting one). Alternatively or additionally, more projection data may be acquired to generate more constraints. Alternatively or additionally, the reconstruction for at least that area (of the constraints) may be repeated, for example if an area is identified to include a comer, special reconstruction parameters for a comer may be used, so that no constraints may need to be removed. Some causes of problematic constraints include detector noise, detector blurring, indeterminate detector response, source noise, scattering and/or diffraction, thickness resolution limitations, very high or very low absorption (thick slabs or comers), non-uniform matter (for example caused by bubbles), poor calibration, poor object positioning and/or orientations, vibration and/or motion. Alternatively or additionally, to removing constraints, other clean-up operations may be performed between image reconstruction stages, for example image smoothing, averaging of pixel values, edge detection and enhancement and thresholding, for example, to assist in convergence.

In a preferred embodiment of the invention, the size of pixels does not change over the course of iterations. Alternatively or additionally, the pixels size and/or the width of a band may be changed. In one example, if the value in a pixel reaches 1 (or 0) (to within a given precision) and/or if the pixel is surrounded by pixels of density value 1 (or 0), the pixel and/or its neighbors are removed from consideration in further iterations. Alternatively or additionally, the band may be widened if several non-zero valued pixels encroach on the edge of the band. Alternatively or additionally, pixels may be reduced in size if they fail to converge on a value of 0 or 1.

Fig. 8A is a flowchart of a method of setting relaxation coefficients, for the iterative reconstmction of Fig. 7 (step 116). Relaxation coefficients are often used in ART techniques, to aid convergence. It is known in the art to use a relaxation value of γ=0.05 to obtain satisfactory convergence. In a prefeπed embodiment of the invention, different values of γ are used for different constraints. Preferably, these different values are related to a size of deviation between the constraint and a projection of the reconstmcted value. In a prefeπed embodiment of the invention, these relaxation coefficients may be set anew every one or more iteration steps.

First, the deviations of reconstmcted values from the constraints are determined (120). These deviations are preferably compared to each other and/or to their historical values (122). New relaxation coefficients are preferably determined (124). Preferably, the new coefficients are constrained to the range 0 to 1, preferably, 0.05 to 1. In a preferred embodiment of the invention, deviations are considered problematic only if they are outside a certain gate, for example a 3σ gate (around the expected value). In a preferred embodiment of the invention, the relaxation coefficients are an increasing linear function of the deviations. Alternatively to a linear relationship, a quadric relationship is used. Alternatively, a power or an exponential relationship are used. Possibly, different relationships between deviations and relaxation coefficients may be defined for different portions of the image.

In a prefeπed embodiment of the invention, if changing a relaxation coefficient for a particular constraint does not improve its convergence, the relaxation coefficient may be maintained at its previous value or set to a predetermined value, for example 0.05.

Referring to Fig. 8B, in a prefeπed embodiment of the invention, a moment based constraint may be defined, for example, to constraint the total area (mass) of the image. In a prefeπed embodiment of the invention, a zero order moment is used. Alternatively or additionally, a higher order moment, such as 1 st or 2nd order moment is used. In a prefeπed embodiment of the mvention, the moment is an average of moments for several projection directions, especially if non-parallel beams are used or in the presence of noise. Thus, a moment of the projection(s) is calculated (126) and a constraint derived from its moment is generated (128), which limits the reconstmcted image as a whole rather than individual pixels. Alternatively or additionally, moment constraints or other multi-pixel constraints may be defined for portions of an image. Often, the reconstmction is completed at this point. However, in some prefeπed embodiments of the invention additional post reconstruction steps may be performed, for example to clean-up the image, as described above with reference to clean-up operations between reconstmction steps. In a prefeπed embodiment of the invention, the image is checked after reconstmction to determine that it matches the projection data. Preferably, the reconstructed image is projected at the projection angles and the amount of discrepancy between the projection data and the generated projections is noted. In a prefeπed embodiment of the invention, the amount of discrepancy is used to indicate an expected eπor level in the image and/or measurements, to an end user. Alternatively or additionally, if the discrepancies are higher than a predetermined amount, the reconstructed image may be rejected and further data acquisition may be required. In a prefeπed embodiment of the invention, only certain portions of the image are checked. Alternatively or additionally, different amounts of discrepancy are allowed for different portions of the image. In a prefeπed embodiment of the invention, the image is binarized or otherwise thresholded to include more than two discrete values) so that it includes only a discrete set of values, preferably, each value coπesponding to a material in the object. In a prefeπed embodiment of the mvention, the threshold is selected so that a zero (and/or higher) moment of the image is not changed by the thresholding operation (to within a preset precision level). In a prefeπed embodiment of the invention, a proper threshold is found by performing a binary search over a range of threshold values between 0+ε and 1-ε.

In a prefeπed embodiment of the invention, edges are detected on the resulting image. Preferably, these edges are used to perform measurements and/or for other inspection tasks. A prefeπed type of edge detection is described in J.C. Russ, The Image Processing Handbook, p. 674, 2nd Ed., CRC Press, 1994, the disclosure of which is incorporated herein by reference. Alternatively or additionally, the image may be analyzed using other tools, for example pattern recognition tools. Alternatively or additionally, the image may be used for non-inspection tasks, for example for promotional or instructional purposes.

In a prefeπed embodiment of the invention, one or more of the following measurement techniques are applied to a reconstmcted image:

(a) Mechanical dimensions. For example, distances between entities and features. Generally, these measurements require edge determination. (b) Geometric fitting. For example to a circle, or a cube. Generally, these measurements also require edge detection and may also require higher quality/resolution reconstruction in some parts of the slice.

(c) Density non-uniformities. In a prefeπed embodiment of the invention, the reconstmcted image is used as a starting point for further reconstruction, where the parts of images which are suspected to include bubbles or non-uniform densities are allowed to vary their density values, during reconstruction.

In a prefeπed embodiment of the invention, various reconstmction parameters may be varied and/or relaxed, responsive to the required measurements, so that a faster and/or more efficient reconstmction is possible. In one example, if measurement of comers is not required in the final image, a lower pixel resolution may be used, at least at comers. Alternatively or additionally, there may be an interaction between the measurement and the imaging device. For example, the imaging device may be calibrated so that the x-ray beam is collimated best at those portions of projection angles which generate traces which pass through pixels of interest, possibly at the expense of other pixels in the image slice. In some cases this may require translating the collimator perpendicular to the x-ray beam, so that a higher quality portion of the collimator interacts with a desired portion of the beam.

In one type of measurement, what is required is a simple yes/no answer to the following question: "is the object within allowed tolerances of the design or not?". This question can often be answered using a less than maximum quality reconstruction, by tailoring the reconstmction quality/precision so that the sum of the reconstmction eπor and the real error is less than the allowed eπor, for the object as a whole or for a portion of interest of the object. If the answer is yes, than it usually does not matter how much deviation there is in the manufactured object. If the answer to the question is no with a low enough probability it may pay to divert the object to a testing laboratory where more exact testing may be applied.

Alternatively or additionally, the reconstructed image may be analyzed to identify problematic portions thereof, for example inconsistent or unexpected absorbency. Such an portion may require more attention. In a prefeπed embodiment of the invention, the extra attention is provided by repeating the reconstmction of the portion, possibly requiring new data to be acquired, possibly at higher resolution or with better statistics. In the example of bubbles, large bubbles can be viewed by using a high resolution. Smaller bubbles, which cannot be viewed, affect the average density, which can be measured. Other problematic areas include areas in which two materials or two sources are blended, for example plastic and additives such as color or plastisizer or two separate aluminum billets. In some cases, intermediate density values may be allowed (for reconstmction) in some parts of the object and not allowed in others. Alternatively, such intermediate values indicate a potential problematic area.

In a prefeπed embodiment of the invention, an additional reconstmction step may be required for some portions of the image. In the case of comers of complex portions of an object, it may be desirable to acquire data at additional and/or different projection angles. Alternatively or additionally, very dense portions or portions with a very low density (for example comers) might be outside the dynamic range of the imaging system. In a prefeπed embodiment of the invention, these portions are imaged again with higher (or lower) intensity x-rays, so that data is acquired within the dynamic range.

In a preferred embodiment of the invention, the interface area between two materials is analyzed to detect the position of the boundary. Alternatively or additionally, the interface area is analyzed to determine the presence of a c st which forms or separates between the two materials. As can be appreciated, the boundary thickness is typically much smaller than the typical pixel sized used for image reconstruction. In a preferred embodiment of the invention, a smaller pixel size is used for the boundary area. Alternatively or additionally, at least one projection angle is selected so that the x-ray beam is substantially parallel to at least a portion of the boundary area. In a prefeπed embodiment of the invention, an increased data acquisition resolution is only utilized along part of the boundary. Preferably, that part of the boundary comprises a plurality of boundary areas which are selected as a statistical sample of the entire boundary area of interest. Preferably, by analyzing these areas it is possible to determine a probability of their being defects along the boundary and/or its average thickness. Alternatively or additionally, such a statistical analysis approach is applied over the entire image to look for defects and/or determine other types of inconsistencies between the reconstmcted image and the object design

Fig. 9 is a schematic illustration of an imaging device 200 in accordance with a prefeπed embodiment of the invention. As in most CT imagers, device 200 utilizes an x-ray tube (or other source) 202 and some type of detector 216. In a prefeπed embodiment of the invention, device 200 uses very highly collimated beams, to achieve a high spatial resolution. Alternatively or additionally, scintillator detectors do not provide the required resolutions. In the embodiment of Fig. 9, device 200 may comprise three x-ray collimators and an optical collimator, for a CCD based detector 216. Preferably, device 200 includes a computer 218 which controls device 200 and/or reconstructs images from data acquired thereby. A first x-ray collimator 204 is placed at an exit of tube 202. A second x-ray collimator 206 is placed between tube 202 and an object 208 to be imaged. A third x-ray collimator 210 is preferably placed after object 208, to remove scattering and diffraction in object 208.

The patterned x-ray beam is preferably detected using a screen-camera combination. A screen 212 is preferably iπadiated by the beam. Light exiting the screen is preferably collimated by an optical collimator 214 (which preferably absorbs x-ray radiation) and is detected by a linear CCD camera 216. In a prefeπed embodiment of the invention, the screen is read out using laser beam. Alternatively or additionally, the screen is a florescent screen, which emits light in the pattern of the x-ray beam for a short time after the irradiation. Alternatively or additionally, the screen acts as a scintillator crystal, and the object is iπadiated again for each data line acquired.

In a preferred embodiment of the invention, the CCD camera has a high spatial resolution, such as 20 micrometers and a large number of elements, such as 10,000. In a prefeπed embodiment of the invention, a TDI type CCD camera is used, to compensate for any motion in the image and/or synchronize with motion transverse to the camera and/or increase the cameras light sensitivity. One cause of motion is vibration. Alternatively or additionally, a two dimensional CCD camera may be used. In some embodiments of the invention, camera 216 may have a smaller field of view than screen 212, so that camera 216 is required to mechanically or optically scan across the surface of screen 212. Preferably the scanning is along one axis only. Alternatively or additionally, the scan is two-dimensional. Alternatively or additionally, camera 216 includes a zoom lens, to allow control over object size and data acquisition resolution.

In a preferred embodiment of the invention, object 208 is brought into place using a conveyer belt, a crane, a robotic arm and/or other actuators known in the art. Preferably, the movement of the object is synchronized to the imaging, such that as soon as an object is brought it is imaged. This may be achieved using central control of both the feeder and the object. Alternatively, a sensor on the imager may activate the imager when an object trips the sensor. In a preferred embodiment of the invention, the projection angles are achieved by moving the detector and/or the x-ray tube. Alternatively or additionally, the object is moved and/or rotated.

In a preferred embodiment of the invention, device 200 is checked and/or calibrated using x-rays. Preferably a phantom is used. Alternatively or additionally, device 200 may be tested using an optical test pattern instead of screen 212. Alternatively or additionally, a light source may be used instead of object 208. In a prefeπed embodiment of the invention, device 200 uses a parallel beam of x-ray radiation. Alternatively or additionally, device 200 uses a fan-beam configuration. Alternatively or additionally, device 200 utilizes a scanning pencil beam.

As can be appreciated, the calibration and/or operation of device 200 may interact with the reconstruction method chosen. In one example, if only a small part of object 218 is being imaged, the collimator may be closed, so that only a narrow x-ray beam is generated and less diffraction and scattering is generated. Conversely, if the collimation of the beam is limited in quality, the smallest pixel size may be made larger. Alternatively a smaller pixel size may be used to compensate. In a preferred embodiment of the invention, device 200 includes multiple detection path elements, each optimized for particular data acquisition conditions. In a prefeπed embodiment of the invention, the path element is automatically selected by the device, responsive to data acquisition requirements. In one example, device 200 includes multiple collimators, for example higher and lower resolution collimators. In another example, device 200 includes a plurality of optical detectors of different spatial, temporal and/or gray level resolution and/or of different sensitivity response, gain control, sensitivity range, spectral response and/or other detection parameters. In another example, device 200 includes multiple optical or x-ray wave-length filters (not shown). In another example, device 200 includes multiple signal processing circuits (for example incorporated in computer 218). In another example, device 200 includes multiple scintillation or solid-state radiation detectors 216.

Fig. 10 is a schematic illustration of an embodiment of the invention for quality control of extmsion. In a preferred embodiment of the invention, an extruded object 222 is extruded by a nozzle 224 connected to a material source 220. In a preferred embodiment of the invention, a CT imager is positioned as shown by reference number 228, so that a cross- section of object 222 is imaged. Alternatively or additionally, the CT imager is positioned at reference number 226, so that the both object 222 and nozzle 224 are imaged simultaneously. Thus, problems in the extrusion process itself (e.g., bubbles) and/or wearing out of nozzle 224 may be detected.

In a prefeπed embodiment of the invention, object 222 is in continuous motion, and helical CT reconstruction techniques are used. In one such reconstmction method, virtual projection data is generated by interpolation between nearby projection angles and/or positions of object 222. In position 226, the nozzle does not move, so helical scanning may not be applicable. Alternatively or additionally, in the device of Fig. 9, helical scanning may be utilized to image object 208 while it is being placed inside device 200. Such helical scanning is useful for more efficient volume imaging. Alternatively or additionally, helical scanning and/or reconstmction may be used to abolish the need to stop the movement of object 208 (due to its being brought in) and/or its internal components before imaging. Alternatively or additionally, to helical scanning, it may be assumed that a defect in object 222 has a considerable axial dimension, thus, motion of object 222 may be ignored during reconstmction. In particular, inconsistencies during reconstruction may indicate a defect, in and of themselves.

In a prefeπed embodiment of the invention, an image of object 222 is reconstructed in substantially real-time, for example, in less than 10, 5, 1, or 0.1 seconds. Additionally, the number of pixels in the reconstmcted image is at least 1,000,000, 20,000,000 or 100,000,000 pixels. This type of resolution and reconstmction time may also be used for imaging moving objects, such as a watch, in operation. It is noted that using the methods described herein and noting that major portions of the imaged object have nearly or exactly known geometries, the actual amount of processing requirement is preferably reduced as compared to prior art methods.

In a preferred embodiment of the invention, the CT imager is used to image objects constmcted of one or more of metal, glass, plastic, mbber, concrete, wood, composite materials and/or other industrial materials. Alternatively or additionally, the imager is used to image objects comprising a plurality of materials, for example aluminum with plastic spacers. In prefeπed embodiments of the invention, the imaged objects are of non-animal origin, thus excluding living and dead human bodies. In a prefeπed embodiment of the invention, composite materials are imaged in a way which reveals information about their compositing materials. In one example, a higher spatial resolution may be provided (e.g., pixel size) in a direction of a grain of the material. Alternatively or additionally, a very small pixel size is provided in at least parts of the object, to better view the grain of the individual component materials. In a prefeπed embodiment of the invention, especially where polymer materials are imaged, the radiation used for imaging may also be used for radiation treatment, for example to generate cross-linking in the polymer material and/or to cure it.

Objects imaged by the CT imager described herein may be manufactured using substantially any known manufacturing method, including, but not limited to: casting, extrusion, micro-extmsion, roller-milling, laminating, machining and/or assembling with and/or without connectors. In particular, some prefeπed embodiments of the invention are suitable for continuous processes, such as extrusion, as well as discrete processes, such as casting. In a prefeπed embodiment of the invention, a cast object may be imaged while it is still in a cast. Also, it is noted that, in some casting processes, imaging of a single slice may be sufficient to approve a specimen.

The imaged objects may have a cross-section between 10 and 100 cm, or even larger, for example being larger than 20, 60 or 100 centimeters. The resolution is preferably between 0.001 and 0.1 mm, for example being better than 50, 20, 7 or even 1 micrometers. In some cases, an object may be too large to image at one time and overlapping portions of the object may be imaged. In one example, the object is rotated (rather than the detector and/or the x-ray source. Possibly, knowledge of the internal stmcture of the object is used to estimate the orientation of the object. Alternatively or additionally, a rotation sensor on a rotational actuator used to rotate the object is used to estimate its angular position. In some embodiments of the invention, a plurality of axial slices of the object are acquired using multiple axially displaced arrays. Alternatively or additionally, a plurality of trans-axially overlapping detector arrays may be used to receive radiation form the object or to read a phosphor display which receives radiation from the object. Alternatively or additionally, to imaging relatively large objects, small objects may be imaged, including for example minerals, mineral samples, gem stones (cut or uncut) and/or artificial gems stones. Such analysis is preferably used to find hidden flaws, bubbles and/or structures.

Fig. 11 is a schematic illustration of a conveyer belt embodiment of the invention. In this embodiment, a conveyer 240 conveys objects 242 to be imaged to a CT imager 230. In a prefeπed embodiment of the invention, all the objects on the conveyer are imaged. Alternatively or additionally, only a sample of the objects are imaged, for example responsive to a reconstmction time constraint of imager 230. Alternatively or additionally, only a sample of the manufactured objects are passed along conveyer 240 to be imaged. In a prefeπed embodiment of the invention, fast imaging of an object 242 is made possible by utilizing multiple radiation sources (not shown) and/or multiple detectors and/or multiple detector rows. Alternatively or additionally, fast imaging utilizes an electron beam- CT imager, in which a ring shaped target is scanned using an electron-beam, to generate a fast moving source. For example in Fig. 11, electron beams from a source at a location 232 are sent to a target 234 which is located along a half ring (shown as part of the dotted line). Such an image may also be used in the embodiment of Fig. 10.

In a preferred embodiment of the invention, the operation of the CT imager may be controlled by a user. Examples of input which may be provided by a user, include: number of projections, required image quality, portions of the image which are of interest, estimated image, coπections to an estimated image, identification of problem portions of an object, corrections to a reconstructed image and/or answers to questions posed by the CT imager. Such questions may be posed by the CT imager, for example in case the imager stalls or cannot reconstruct an image. Answers to such questions may include, for example a selection between two alternate reconstmctions or resetting of certain reconstruction parameters.

In a prefeπed embodiment of the invention, each object to be imaged may utilize different reconstmction and/or acquisition parameters. In a preferred embodiment of the invention, the parameters are selected based on a simulation n on the object design. The simulation may be used, for example, to determine which parameters yield a fastest reconstruction, highest probability of detecting eπors, assuring convergence and/or require a shortest data acquisition time.

Alternatively or additionally, such parameters may be determined by applying heuristics to the image. One example of such a heuristic is to set a width of a band to be lower at straight areas of an image than at curved areas of an image. In some cases, the imaged object may be blurred as a result of motion artifacts. In a prefeπed embodiment of the invention, such motion artifacts may be coπected using the knowledge that the imaged object is a rigid object, making assumptions of the motion vector and shifting the projection data to take the motion vector into account. Possibly, the motion vector is determined by analyzing the image and/or coπespondence between sets of projection data from different angles. Alternatively or additionally, the direction of the motion may be known, for example as being a result of motion of the CT imager and/or the object while it is being imaged. Alternatively or additionally, a motion vector may be determined by comparing projection data from opposite sides of the object. Alternatively or additionally, a motion vector may be determined utilizing the estimated image and/or other a-priori knowledge. Altematively or additionally, motion caused by vibration may be coπected for by measuring the vibration and estimating the motion blur caused thereby. In some cases, such a coπection may take into account the different resonance properties of different parts of the object. It is noted, that such an analysis can usually be performed on a single object and then applied to all the manufactured objects. In one type of non-destructive testing, an object is analyzed to determine vibration modes of its sub-components. To test an object specimen, the object is stmck and/or otherwise vibrated and one or more images are acquired. Differences in manufacturing between objects (deviations) will often cause a variation in the resonance characteristics of portions of the object. Thus, the acquired image may be different than for a "coπect" object. The correct image is preferably used as an estimated image for the tested image.

The present invention has been described mainly as using 2D X-ray CT imaging. However, it should be appreciated that the above reconstmction techniques may be used for other imaging techniques. In one example, the above reconstruction technique may be used to generate cine images of the operation of an object. In this context it is noted that one type of a- priori knowledge in a manufacture object is the existence of certain objects (sub-components), usually rigid, therein. If the layout of each of these objects is known, even if their exact position is not, the objects may usually be identified from a low resolution reconstructed image. Once the objects are identified, it may be possible to more exactly specify their position and/or orientation in the image, for example using pattern matching, possibly generating strong constraints on the image. In manufactured objects it is often the case that the component objects (usually sub-components of the manufactured object) are known, even if their exact position is not. Additionally, in an operating machine, (some of) the component elements are in motion, but their exact shape may be known (for example from a previous image). Alternatively or additionally, a cap on the amount of motion of the elements may be known (thereby possibly defining a band size). Alternatively or additionally, the imaging may be synchronized to the motion of the sub-components. In some cases, the periodicity or other regularity of motion of sub-components is determined by analyzing vibration vectors of the object or by Fourier analyzing a sequence of images to determine motion frequencies in certain parts of the object. Alternatively or additionally to cine imaging of an operating machine, a stressed object may be imaged to track its failure modes. In all of the above cases, it is possible to restrict the number of pixels which need to be actually reconstmcted.

Alternatively or additionally to acquiring cine images, 3D images of an object may be acquired. One method of acquiring 3D images is helical scanning, described above. Alternatively or additionally, a multi-slice image may be acquired, possibly using a multi-row detector and/or by moving the object and/or the imager relative to each other and/or using a cone beam. In a prefeπed embodiment of the invention, bands and/or multi-resolution grids defined for a 3D image are 3D. For example, the band portion in one slice may be dependent on the object profile in nearby slices. Such a dependency may also be used when determining band widths and expected eπors in a 2D imaging.

Alternatively or additionally, a 3D image of an object may be acquired by imaging the object from angles which do not all lie in a single plane. In an extreme example, an object may be imaged from three perpendicular directions, as well as from intermediate directions, possibly providing hemi-spherical coverage of the object.

In a prefeπed embodiment of the invention, the imaged object may be specially prepared for imaging, for example by impregnation or immersion in a contrast media, preferably a liquid or a gas. This may be useful when imaging a very transparent or a porous structure. Alternatively or additionally, only one of a plurality of source materials which are mixed together to form a final material is tagged using a contrast media. Thus, an even dispersion of the source material in the mixture may be determined.

Alternatively or additionally, to imaging using x-ray CT, some of the above described embodiments may be used for imaging using gamma radiation. Alternatively or additionally, to transmission imaging, emission imaging may also be applied, for example by using a radioactive source material instead of a contrast media. Alternatively or additionally, SPECT imaging may be applied, using the above reconstmction techniques. Alternatively or additionally, multi-wavelength and/or combined gamma radiation and x-ray radiation may be applied. Alternatively or additionally, optical tomography may be performed. Generally however, when the imaged object is made radioactive it cannot be sold. Thus, this type of imaging is usually reserved for test objects and not for production objects.

Some embodiments of the present invention are especially suited to industrial imaging, due to the types of limitations, requirements, allowed radiation energies, type of motion, predictability of structure and/or pre-knowledge in that field. However, it should be appreciated that some of the above features and/or prefeπed embodiments of the invention may also be applied to medical imaging and/or other types of tomographic reconstruction, for example, electron microscopy.

It will be appreciated that the above described methods of CT image reconstruction may be varied in many ways, including, changing the order of steps, which steps are performed on-line and which steps are performed off-line. In addition various distributed and/or centralized configurations may be used to implement the above invention, preferably utilizing a variety of software tools. In addition, a multiplicity of various features, both of methods and of devices have been described. It should be appreciated that different features may be combined in different ways. In particular, not all the features shown above in a particular embodiment are necessary in every similar prefeπed embodiment of the invention.

Further, combinations of the above features are also considered to be within the scope of some prefeπed embodiments of the invention. Also within the scope of the invention are computer readable media on which software, for performing part or all of a preferred embodiment of the invention, are written. It should also be appreciated that many of the embodiments are described only as methods or only as apparatus. The scope of the invention also covers hardware and/or software adapted and/or designed and/or programmed to carry out the method type embodiments. In addition, the scope of the invention includes methods of using, constructing, calibrating and/or maintaining the apparatus described herein. When used in the following claims, the terms "comprises", "comprising", "includes", "including" or the like mean "including but not limited to".

Claims

1. A method of image reconstmction for tomographic imaging, comprising: providing an indication of an internal stmcture of an object to be imaged, which object is of non-animal origin; selecting projection angles for reconstmction responsive to said indication; and reconstructing an image from data acquired at said selected projection angles.

2. A method according to claim 1, comprising: acquiring data at a plurality of projection angles; and reconstructing a low quality image from said acquired data, wherein said indication comprises said low quality image.

3. A method according to claim 2, wherein said data acquired at selected projection angles comprises a sub-set of data acquired at said plurality of projection angles.

4. A method according to claim 2, wherein said selected projection angles and said plurality of projection angles each include at least one projection angles not found in the other.

5. A method according to claim 2, wherein said data acquired at said plurality of projection angles comprises a sub-set of all data acquired at said selected plurality of projection angles.

6. A method according to any of claims 1-5, wherein selecting projection angles comprises selecting an estimated minimum number of projection angles which would yield a suitable reconstmction.

7. A method according to any of claims 1-6, wherein selecting projection angles comprises selecting projection angles responsive to an object complexity.

8. A method according to any of claims 1-7, wherein selecting projection angles comprises selecting projection angles responsive to indications of heavy absorption in said indication of internal stmcture.

9. A method according to any of claims 1-7, wherein selecting projection angles comprises selecting projection angles responsive to indications of low absorption in said indication of internal structure.

10. A method according to any of claims 1-9, wherein selecting projection angles comprises selecting projection angles responsive to a particular feature in the reconstructed image.

11. A method according to claim 10, wherein said particular feature comprises at least one sharp comer.

12. A method according to claim 10, wherein said particular feature comprises a boundary area of said object.

13. A method according to claim 12, wherein said boundary area comprises a boundary between two materials which comprise the object.

14. A method according to any of claims 1-8, wherein selecting projection angles comprises selecting projection angles responsive to an incidence angle of traces of said projection angles with features of said indicated internal structure.

15. A method of image reconstruction for tomographic imaging, comprising: providing an indication of an internal structure of an object to be imaged, which object is of non-animal origin; selecting at least one resolution for data acquisition responsive to said indication; and reconstmcting an image from data acquired using said at least one resolution.

16. A method according to claim 15, wherein said at least one resolution comprises at least two different resolutions for two different projection angles.

17. A method according to claim 15 or claim 16, wherein said at least one resolution comprises at least two different for data acquired at a single projection angles.

18. A method according to any of claims 15-17, wherein said resolution comprises a spatial resolution.

19. A method according to any of claims 15-18, wherein said resolution comprises a grayscale resolution.

20. A method according to any of claims 15-19, wherein selecting at least one resolution comprises selecting an estimated minimum resolution which would yield a suitable reconstmction.

21. A method according to any of claims 15-20, wherein selecting at least one resolution comprises at least one resolution responsive to an object complexity.

22. A method according to claim 21, wherein said object complexity comprises a local object complexity.

23. A method according to any of claims 15-22, wherein selecting at least one resolution comprises selecting at least one resolution responsive to indications of heavy absorption in said indication of internal stmcture.

24. A method according to any of claims 15-22, wherein selecting at least one resolution comprises selecting at least one resolution responsive to indications of low absorption in said indication of internal stmcture.

25. A method according to any of claims 15-24, wherein selecting at least one resolution comprises selecting at least one resolution responsive to a particular feature of said reconstmcted image.

26. A method according to claim 25, wherein said particular feature comprises a reference feature used for measurement on the image.

27. A method according to claim 25, wherein said particular feature comprises a boundary area.

28. A method according to claim 25, wherein said particular feature comprises a feature which is itself measured on the image.

29. A method according to any of claims 1-28, wherein said indication comprises an estimation of internal structure of said object.

30. A method according to claim 29, wherein said estimation comprises a design specification of said object.

31. A method according to claim 29, wherein said estimation comprises an at least two- dimensional representation of said object.

32. A method according to claim 29, wherein said estimation comprises a previously reconstructed image of said obj ect.

33. A method according to claim 29, wherein said estimation comprises an image of a previously imaged object of similar manufacture.

34. A method according to any of claims 1-33, wherein said indication comprises a possible deviation from a desired internal structure.

35. A method according to any of claims 1-34, wherein reconstructing an image comprises reconstructing only a portion of said object, responsive to said indication.

36. A method of image reconstmction, comprising: providing projection data of an object, which object is of non- animal origin; and reconstructing an image from said projection data, wherein a different reconstruction treatment is applied to at least one portion of said image, wherein said different reconstruction treatment comprises a different reconstmction method for said at least one portion.

37. A method of image reconstruction, comprising: providing projection data of an object, which object is of non-animal origin; and reconstmcting an image from said projection data, wherein a different reconstruction treatment is applied to at least one portion of said image, wherein said different reconstmction treatment comprises using a different value for at least one reconstruction parameter of a reconstruction method used for said at least one portion.

38. A method of image reconstruction, comprising: providing projection data of an object, which object is of non-animal origin; and reconstmcting an image from said projection data, wherein a different reconstmction treatment is applied to at least one portion of said image, wherein said different reconstruction treatment comprises reconstructing said at least one portion at a different spatial resolution.

39. A method according to any of claims 36-38, wherein said at least one portion comprises at least two portions, each receiving different treatment from each other and from at least a third portion of said image.

40. A method of image reconstruction, comprising: providing projection data of an object, which object is of non-animal origin; and reconstructing an image from said projection data, wherein a different reconstruction treatment is applied to at least one portion and at least a second portion of said image, such that at least three different treatments are applied to said image, one for each of said portions and for at least another portion of said image.

41. A method according to any of claims 36-37 or 40, wherein applying a different reconstmction treatment comprises reconstmcting at a different resolution.

42. A method according to any of claims 36-41, comprising: providing an indication of an internal stmcture of said object, wherein said different reconstruction treatment is applied responsive to said indication.

43. A method of image reconstmction, comprising: providing projection data of an object, which object is of non-animal origin; providing an indication of an internal structure of said object; and reconstructing an image from said projection data, wherein a different reconstmction treatment is applied to at least one portion of said image, wherein said different reconstmction treatment is applied responsive to said indication.

44. A method according to claim 42 or claim 43, wherein providing said indication comprises reconstructing a low quality image of said object.

45. A method according to claim 44, wherein said low-quality image is reconstructed using a backprojection method.

46. A method according to claim 44, wherein said low-quality image is reconstructed using a Baysian method.

47. A method according to claim 44, wherein said low-quality image is reconstmcted using a Fourier transform method.

48. A method according to claim 44, wherein said low-quality image is reconstmcted using a Maximum Likelihood method.

49. A method according to claim 44, wherein said low-quality image is reconstructed using a Maximum Entropy method.

50. A method according to any of claims 44-49, wherein said low-quality image is reconstmcted using a different resolution grid than used for said reconstructing an image.

51. A method according to any of claims 42-50, wherein said reconstmcting an image comprises reconstmcting using an algebraic reconstmction method.

52. A method according to any of claims 42-50, wherein said reconstmcting an image comprises reconstructing using a Baysian method.

53. A method according to any of claims 42-50, wherein said reconstmcting an image comprises reconstmcting using a Fourier transform method.

54. A method according to any of claims 42-50, wherein said reconstmcting an image comprises reconstructing using a Maximum Likelihood method.

55. A method according to any of claims 42-50, wherein said reconstructing an image comprises reconstructing using a Maximum Entropy method.

56. A method according to any of claims 42-50, wherein said reconstructing an image comprises reconstructing using a backprojection method.

57. A method according to any of claims 42-56, wherein said reconstmcting an image comprises reconstmcting using a finer resolution grid than used for said indication.

58. A method according to any of claims 42-57, wherein said special treatment comprises setting up an initial estimate of said image, responsive to said indication.

59. A method according to any of claims 42-58, wherein providing said indication comprises retrieving an indication generated responsive to an imaging of a similar object.

60. A method according to any of claims 42-59, wherein providing said indication comprises providing a design specification of said object.

61. A method according to any of claims 42-60, wherein providing said indication comprises providing a manufacturing specification of said object.

62. A method according to any of claims 42-61, wherein said special treatment is responsive to an estimated distance, of said at least one portion, from at least one edge of said object, wherein said estimation is based on said indication.

63. A method according to any of claims 42-62, wherein said special treatment is responsive to an expected reconstmction eπor, of said at least one portion, wherein said expected eπor is based on said indication.

64. A method according to any of claims 42-63, wherein said special treatment is responsive to an expected manufacturing eπor, of said at least one portion, wherein said expected manufacturing eπor is based on said indication.

65. A method according to any of claims 42-64, wherein said special treatment is responsive to a confidence in an internal structure of said object, wherein said confidence is based on said indication.

66. A method according to any of claims 36-65, wherein said at least one portion is at least one band surrounding said object.

67. A method according to claim 66, wherein said band comprises a region that overlaps at least an outside edge of said object.

68. A method according to claim 66, wherein said band comprises a region that surrounds an aperture of said object.

69. A method according to any of claims 36-65, wherein said at least one portion encompasses substantially only a particular feature of said object.

70. A method according to claim 69, wherein said feature comprises an area having a particular density.

71. A method according to any of claims 36-70, wherein said special treatment comprises utilizing an estimate for selected pixels inside said at least one portion, instead of reconstmcting said selected pixels.

72. A method according to any of claims 36-71, wherein said special treatment comprises utilizing an estimate for selected pixels outside said at least one portion, instead of reconstmcting said selected pixels.

73. A method according to any of claims 36-71, wherein said special treatment comprises providing a lower quality reconstruction outside said at least one portion.

74. A method according to claim 73, wherein said lower quality reconstmction comprises a reconstruction with a greater eπor.

75. A method according to claim 73, wherein said lower quality reconstmction comprises a reconstmction with a lower spatial resolution.

76. A method according to any of claims 36-75, wherein a same pre-processing is applied to pixels inside and outside of said at least one portion.

77. A method according to any of claims 36-75, wherein a different pre-processing is applied to pixels inside and outside of said at least one portion.

78. A method according to claim 76 or claim 77, wherein said pre-processing comprises filtering.

79. A method according to any of claims 36-78, wherein said special treatment is responsive to a level of detail required in said at least one portion.

80. A method according to any of claims 36-79, wherein said special treatment is responsive to measurements to be performed on said at least one portion.

81. A method according to any of claims 36-80, wherein said special treatment is responsive to a material composition of said object.

82. A method according to claim 81, wherein said object comprises a composite material and said special treatment is responsive to at least one characteristic of said composite material.

83. A method according to claim 82, wherein said at least one characteristic comprises a fiber direction of said material.

84. A method according to claim 82, wherein said at least one characteristic comprises a cell size of said material.

85. A method of image reconstruction, comprising: providing projection data of an object, which object is of non-animal origin; providing an indication of an internal stmcture of said object; and reconstmcting an image from said projection data, wherein said reconstmcting comprises only reconstructing pixels from said data in at least one certain region of said image, which region has a shape determined responsive to said indication.

86. A method according to claim 85, wherein said at least one certain region comprises at least two non-contiguous regions.

87. A method according to claim 85 or 86, wherein said shape comprises a band shape enclosing at least one area of non-reconstructed pixels.

88. A method according to any of claims 85-87, wherein said shape is determined from boundaries of said object which boundaries are indicated by said indication.

89. A method of image reconstmction, comprising: providing projection data of an object, which object is of non-animal origin; providing an indication of an internal stmcture of said object; and reconstructing an image from said projection data, responsive to at least one potential- problem area in said indication.

90. A method according to claim 89, wherein said at least one potential-problem area comprises an edge.

91. A method according to claim 89 or claim 90, wherein said at least one potential- problem area comprises a suspected crack area.

92. A method according to claim 89 or claim 90, wherein said at least one potential- problem area comprises a suspected void area.

93. A method according to any of claims 89-92, wherein reconstructing comprises varying a spatial resolution of said reconstmction, responsive to a location of said at least one potential problem area.

94. A method according to any of claims 89-93, wherein reconstructing comprises varying a gray-level resolution of said reconstruction, responsive to a location of said at least one potential problem area.

95. A method according to any of claims 89-94, comprising defining areas to reconstmct differently, responsive to said at least one potential problem area.

96. A method according to any of claims 89-95, comprising determining a local confidence level responsive to said edges.

97. A method of iterative image reconstruction, comprising: providing projection data of an object to be imaged, which object is of non-animal origin; first reconstructing said object from said projection data; rejecting at least some of said data responsive to said first reconstmction; and repeating said first reconstruction, at least once, after said rejecting.

98. A method of iterative image reconstruction, comprising: providing projection data of an object to be imaged, which object is of non- animal origin; generating reconstmction constraints; first reconstructing said object from said projection data, responsive to said reconstmction constraints; rejecting at least some of said consfraints responsive to said first reconstruction; and repeating said first reconstruction, at least once, after said rejecting.

99. A method of iterative image reconstmction, comprising: providing projection data of an object to be imaged, which object is of non-animal origin; generating reconstmction constraints; first reconstmcting said object from said projection data, responsive to said reconstruction constraints; generating relaxation coefficients responsive to said first reconstruction; and repeating said first reconstruction, using said relaxation coefficients.

100. A method of iterative image reconstmction, comprising: providing projection data of an object to be imaged, which object is of non-animal origin; generating reconstruction constraints; first reconstructing said object from said projection data, responsive to said reconstruction constraints; varying values in said first reconstmction responsive to a majorant distribution function; and repeating said first reconstruction, at least once, after said varying.

101. A method according to claim 100, comprising: determining said majorant distribution to be zero outside a convex object which is defined by all traces in said projection data which have zero projection values.

102. A method according to claim 101, wherein said majorant distribution function comprises at least three values.

103. A method according to any of claims 97-102, comprising processing said data prior to repeating said first reconstmction.

104. A method according to any of claims 97-103, wherein said first reconstmction comprises a plurality of iterations.

105. A method according to any of claims 97-104, wherein said repeated reconstmction comprises a plurality of iterations.

106. A method according to any of claims 97-105, wherein said repeated reconstmction comprises applying a different pre-processing to said first reconstruction of said object, responsive to said first reconstruction.

107. A method according to any of claims 97-106, wherein said repeated reconstruction comprises applying a different pre-processing to said projection data, responsive to said first reconstruction.

108. A method according to any of claims 97-107, wherein said repeated reconstmction comprises applying a different pre-processing to said first reconstmction of said object, responsive to an indication of an internal structure of said object.

109. A method according to any of claims 97-108, wherein said repeated reconstmction comprises applying a different pre-processing to said projection data, responsive to responsive to an indication of an internal structure of said object.

110. A method of image acquisition of an object, comprising: acquiring a set of projection data; reconstmcting a first image from said projection data using a first reconstruction method; analyzing said image to determine special treatment for portions of said image; reconstructing a second image of said object, using said analysis, wherein said second reconstmction is a different reconstmction method from said first reconstmction.

111. A method according to claim 110, comprising, acquiring data for said second reconstmction, responsive to said analysis.

112. A method according to claim 111, wherein said data is acquired responsive to a desired image quality in said second image.

113. A method according to claim 111 or claim 112, wherein said data is acquired responsive to a desired analysis of said second image.

114. A method according to any of claims 111-113, comprising varying an intensity of ionizing radiation used for said data acquisition, responsive to said analysis.

115. A method according to any of claims 111-113, comprising varying a wavelength of ionizing radiation used for said data acquisition, responsive to said analysis.

116. A method according to claim 114 or 115, wherein said ionizing radiation comprises x- ray radiation.

117. A method according to claim 114 or 115, wherein said ionizing radiation comprises gamma radiation.

118. A method according to any of claims 111-113, wherein said data is acquired using non- ionizing electro-magnetic radiation.

119. A method according to any of claims 111-118, comprising varying at least one parameter of a detection circuit, responsive to said analysis.

120. A method according to claim 119, wherein said at least one parameter comprises a gain.

121. A method according to any of claims 111-120, comprising selecting at least one element of a detection system, from a plurality of available elements, responsive to said analysis.

122. A method according to claim 121, wherein said element comprises a detector.

123. A method according to claim 121, wherein said element comprises a filter.

124. A method according to claim 121, wherein said element comprises a collimator.

125. A method according to any of claims 110-124, wherein said second reconstmction method comprises an algebraic reconstruction method.

126. A method according to claim 125, wherein said algebraic reconstmction method comprises an ART-like reconstmction method.

127. A method according to any of claims 110-124, wherein said second reconstmction method comprises a Baysian reconstmction method.

128. A method according to any of claims 110-124, wherein said second reconstruction method comprises a Fourier transform reconstmction method.

129. A method according to any of claims 110-124, wherein said second reconstmction method comprises a Maximum Likelihood reconstmction method.

130. A method according to any of claims 110-124, wherein said second reconstruction method comprises a Maximum Entropy reconstruction method.

131. A method according to any of claims 110-124, wherein said second reconstmction method comprises a backprojection reconstruction method.

132. A method according to any of claims 110-131, wherein said second reconstmction method uses a different resolution grid than said first reconstruction method.

133. A method according to any of claims 110-132, wherein said first reconstruction method comprises a backprojection method.

134. A method according to any of claims 110-132, wherein said first reconstruction method comprises a Baysian method.

135. A method according to any of claims 110-132, wherein said first reconstruction method comprises a Maximum Likelihood method.

136. A method according to any of claims 110-132, wherein said first reconstmction method comprises a Maximum Entropy method.

137. A method according to any of claims 110-132, wherein said first reconstmction method comprises a Fourier transform method.

138. A method according to any of claims 110-137, wherein said first reconstruction failed to achieve a satisfactory convergence for at least a portion of said image.

139. A method according to claim 138, wherein said failure is determined from an eπor level.

140. A method according to claim 138, wherein said failure is determined from unexpected values for reconstmcted pixel values.

141. A method of image acquisition of an object, comprising: acquiring a set of proj ection data; reconstructing a first image from said projection data using a first reconstmction method; analyzing said image to determine special freatment for portions of said image; acquiring data for a second reconstmction, utilizing a different data acquisition configuration from a first configuration used for said acquiring a set of projection data, responsive to said analysis, which data acquisition configuration comprises selected elements from an available set of functionally equivalent elements; and reconstructing a second image of said object.

142. A method according to claim 141, wherein said different configuration uses a different optical detector from said first configuration.

143. A method according to claim 141, wherem said different configuration uses a different filter from said first configuration.

144. A method according to claim 141, wherein said different configuration uses a different detector circuit from said first configuration.

145. A method according to any of claims 1-144, comprising post-processing said reconstructed image using an image-processing method.

146. A method according to claim 145, wherein said image processing method is adapted to enhance measurement of features in said reconstmcted image.

147. CT imaging apparatus, comprising: a source of x-ray radiation; a data acquisition system for acquiring attenuation data corresponding to an x-ray density of an object placed in said system, wherein said data acquisition system comprises at least one set of functionally equivalent elements; and a controller which selectively selects a particular one of said elements to be used in said data acquisition system, responsive to an analysis performed by said controller, of data acquired through said system, with a different one of said elements.

148. A method of algebraic CT image reconstmction, comprising: providing projection data; generating a constraint on a moment of said data; and reconstructing an image from said data using said moment constraint.

149. A method according to claim 148, wherein said moment comprises a first-order moment.

150. A method according to claim 148, wherein said moment comprises a second-order moment.

151. A method according to claim 148, wherein said moment comprises a higher than second-order moment.

152. A method of generating relaxation coefficients for an iterative algebraic reconstmction method, comprising: providing a preliminary image reconstructed with a set of constraints during a given iteration; and setting relaxation coefficients for a subsequent iteration responsive to deviations between said constraints and values in said image.

153. A method according to claim 152, wherein setting relaxation coefficients, comprises: comparing said deviations to a threshold; and setting relaxation coefficients responsive to said comparison.

154. A method according to claim 153, wherein said threshold is a function of statistical properties of said deviations.

155. A method of CT image reconstmction, comprising: providing projection data of an object from a plurality of projection angles, which object is of non-animal origin; back-projecting said data into a data structure representing a varying resolution grid.

156. A method according to claim 155, wherem said data-stmcture comprises a hierarchical data stmcture.

157. An industrial inspection system comprising: a feeder of one or more objects to be imaged, which objects are of non- animal origin; and a CT imager, mechanically coupled to said feeder, which images said one or more objects.

158. A system according to claim 157, wherein said source comprises a conveyer belt conveying a plurality of objects.

159. A system according to claim 158, wherein said object is assembled from subcomponents.

160. A system according to claim 158, wherein said object is cast.

161. A system according to claim 158, wherein said object is rolled.

162. A system according to claim 158, wherein said object is injection molded.

163. A system according to claim 158, wherein said feeder comprises a take-up device.

164. A system according to claim 158, wherein said object is machined.

165. A system according to claim 157, wherem said feeder comprises an extmder which extrudes an object in a form of a continuous profile.

166. A system according to claim 165, wherein said extmder comprises an extrusion nozzle.

167. A system according to claim 165, wherein said extmder comprises a shaper.

168. A system according to any of claims 157-167, wherein said object consists essentially of one material.

169. A system according to any of claims 157-167, wherein said object consists essentially of two materials.

170. A system according to any of claims 157-167, wherein said object is composed of a composite material.

171. A system according to any of claims 157-167, wherein said object consists of more than two materials.

172. A system according to any of claims 157-171, wherein said CT imager is synchronized to image an object responsive to a provision of said object by said feeder.

173. A system according to any of claims 157-172, wherein said CT imager images said objects using a spiral imaging method.

174. A system according to any of claims 157-173, wherein said CT imager utilizes a method accordmg to any of claims 1-147.

175. A CT imager for industrial imaging comprising: a detector; an x-ray source; an imaging area; and a manipulator which lifts and conveys objects to be imaged to and from said imaging area.

176. An imager according to claim 175, wherein said manipulator comprises a robotic arm.

177. An imager according to claim 175, wherein said manipulator comprises a winch.

178. A method of manufacturing quality assurance, comprising: manufacturing an object; imaging said object with a CT imager; measuring features on said image to detect deviations in an internal stmcture of said object from design specifications; and rejecting said object if it does not meet said design specifications.

179. A method according to claim 178, wherein said object is cast.

180. A method according to claim 178, wherein said object is extruded.

181. A method according to claim 178, wherein said object is mold-formed.

182. A method according to any of claims 178-181, comprising: analyzing said image to detect at least one material defect; and rejecting said object responsive to said detected defects.

183. A method according to claim 182, wherein said at least one defect comprises a bubble.

184. A method according to claim 182, wherein said at least one defect comprises a void.

185. A method according to claim 182, wherein said at least one defect comprises a variation in density.

186. A method according to claim 182, wherein said at least one defect comprises a crack.

187. A method according to claim 182, wherein said at least one defect comprises a cmst.

188. A method according to any of claims 1-146, wherein said object comprises a manufactured object.

189. A method according to claim 188, wherein said object is a cast object.

190. A method according to claim 188, wherein said object is an extruded object.

191. A method of tomographic reconstruction, comprising: acquiring image data from a plurality of projection angles of an object, which object is of non-animal origin; first reconstructing an image from said image data, using a tomographic reconstruction method; and applying a discretization to said image to convert image values of said reconstmcted image into a limited set of allowed image values, wherein the discretization maintains at least one image property, which at least one image property comprises a moment of said image.

192. A method according to claim 191, wherein said moment is a first-order moment.

193. A method according to claim 191, wherein said moment is a second- or higher- order moment.

194. A method according to any of claims 191-193, wherein said limited set of values comprises only two values.

195. A method according to any of claims 191-194, comprising second reconstmcting said image, in an iterative manner after said applying a discretization.

196. A method according to any of claims 191-195, wherein said limited set of values substantially corresponds in number of value and in relative values to expected density values in said image.

197. A method of tomographic reconstruction, comprising: providing an object having at least one sub-component with a known geometry, which object is of non-animal origin; acquiring image data of said object from a plurality of projection angles; reconstructing an image of said object using said known geometry of said at least one sub-component.

198. A method according to claim 197, wherein reconstructing an image comprises: generating a low quality reconstmction of said image; identifying said sub-component in said image; and further reconstructing said image using said identification, as a basis.

199. A method according to claim 198, wherein identifying said sub-component comprises identifying a position of said sub-component.

200. A method according to claim 198, wherein identifying said sub-component comprises identifying an orientation of said sub-component.

201. A method according to any of claim 198-200, wherein said further reconstmcting comprises fixing values for pixels in said image, which pixels correspond to portions of said at least one-sub component, which fixing is responsive to said known geometry.

202. A method according to any of claim 197-201, wherein said object comprises essentially of said at least one sub-component.

203. A method according to any of claims 197-202, wherein said at least one subcomponent comprises two sub-components having known geometries.