Three-dimensional image reconstruction and device based on ct hardening artifact correction

Abstract

The application provides a three-dimensional image reconstruction method and device based on CT hardening artifact correction. The method comprises: obtaining a projection image set of a workpiece to be measured; wherein the projection image set is obtained by complete scanning of the workpiece to be measured by a scanner; obtaining a multi-energy attenuation model based on a fitting coefficient vector and a mass density vector; constructing a constraint regularization hardening correction model using a maximum likelihood principle based on Poisson characteristics of the projection image set and the multi-energy attenuation model; under a block coordinate descent framework, decomposing the constraint regularization hardening correction model into two sub-models to alternately solve the fitting coefficient vector and the mass density vector; and arranging data based on the obtained mass density vector to generate a three-dimensional reconstruction image of the workpiece to be measured. The method can correct CT hardening artifacts without additional devices and prior information, and can accurately realize three-dimensional image reconstruction.

Description

TECHNICAL FIELD

[0001] Embodiments of the present disclosure relate to the technical field of image reconstruction, in particular to three-dimensional image reconstruction based on CT hardening artifact correction and apparatus. BACKGROUND

[0002] In the process of X-ray CT imaging, hardening artifacts are often affected, which are mainly caused by the energy change of X-rays when penetrating different density materials, resulting in uneven brightness of the image.

[0003] At present, various correction techniques have been proposed, such as CT ray hardening artifact correction based on energy spectrum analysis. By using an energy-resolving detector to obtain the energy spectrum distribution data of different penetration thicknesses of X-rays penetrating the sample, the photon number corresponding to different penetration thicknesses is determined, and the photons in the preset energy range are intercepted, and then the penetration coefficient and the hardening coefficient are determined to obtain the fitting relationship between the two. The hardening correction relationship function is obtained by analyzing the energy spectrum of the X-ray scanning sample, so as to correct the hardening of the sample projection data.

[0004] However, this hardening artifact correction method requires the following prior information: energy spectrum distribution data of X-rays penetrating the sample, penetration coefficient and hardening coefficient corresponding to different penetration thicknesses, and energy range based on energy spectrum peak value corresponding to maximum penetration thickness of energy spectrum distribution curve to maximum ray energy value; and the energy-resolving detector is relatively expensive, and the spatial resolution is not as good as that of the non-energy-resolving detector, which may affect the clarity of the image and increase the difficulty of application.

[0005] Therefore, there is an urgent need for a method for correcting CT hardening artifacts without additional devices and prior information, and then accurately reconstructing three-dimensional images. SUMMARY

[0006] Therefore, the present application provides a three-dimensional image reconstruction method based on CT hardening artifact correction and an apparatus, which corrects the CT hardening artifact without additional devices and prior information, and then accurately reconstructs three-dimensional images.

[0007] To solve the above technical problems, the technical solution of the present application is as follows:

[0008] In one embodiment, a three-dimensional image reconstruction method based on CT hardening artifact correction is provided, the method comprising:

[0009] obtain a projection image set of the workpiece to be measured, wherein the projection image set is obtained by performing a complete scan on the workpiece to be measured by a scanner;

[0010] obtain a multi-energy attenuation model based on the fitting coefficient vector and the mass density vector, wherein the multi-energy attenuation model is used to represent a noiseless measurement value received by an energy integrating detector;

[0011] construct a constraint regularized hard correction model based on a Poisson characteristic of the projection image set and the multi-energy attenuation model using a maximum likelihood principle;

[0012] under a block coordinate descent framework, decompose the constraint regularized hard correction model into two sub-models to alternately solve the fitting coefficient vector and the mass density vector;

[0013] perform data arrangement based on the mass density vector obtained by solving to generate a three-dimensional reconstruction image of the workpiece to be measured.

[0014] wherein the constraint regularized hard correction model is:

[0015]

[0016] and

[0017] wherein s represents the fitting coefficient vector, and p represents the mass density vector; g(p) is a regularization function of the mass density vector, J is the number of elements of the vector p, N j represents a neighborhood of pixel j, and ψ(x) is a Huber potential function, and μ is a regularization parameter; is the multi-energy attenuation model, I is the projection image set, and f is a preset threshold.

[0018] wherein the two sub-models include:

[0019] a sub-model about the fitting coefficient vector s with a fixed mass density vector p is:

[0020]

[0021] wherein λ=[λ1 … λ R ] T is a Lagrange multiplier vector, and p (i-1) represents the mass density vector after the i-1th iteration;

[0022] and a sub-model about the mass density vector p with a fixed fitting coefficient vector s is:

[0023]

[0024] wherein s (i)a fitted coefficient vector after the i-th iteration.

[0025] wherein the decomposing the constraint regularized hardened correction model into two sub-models to solve the fitted coefficient vector and the mass density vector alternately comprises:

[0026] solving the fitted coefficient vector using a non-negative matrix solution algorithm for the sub-model about the fitted coefficient vector s;

[0027] solving the mass density vector using a non-negative matrix solution algorithm for the sub-model about the mass density vector p.

[0028] In another embodiment, a three-dimensional image reconstruction device based on CT hardened artifact correction is provided, the device comprising:

[0029] a first acquisition unit configured to obtain a projection image set of a workpiece to be measured; wherein the projection image set is obtained by a scanner performing a complete scan on the workpiece to be measured;

[0030] a second acquisition unit configured to acquire a multi-energy attenuation model based on a fitted coefficient vector and a mass density vector; wherein the multi-energy attenuation model is used to represent a noiseless measurement value received by an energy-integrating detector;

[0031] a construction unit configured to construct a constraint regularized hardened correction model based on a Poisson property of the projection image set and the multi-energy attenuation model using a maximum likelihood principle;

[0032] a solution unit configured to decompose the constraint regularized hardened correction model into two sub-models to solve the fitted coefficient vector and the mass density vector alternately under a block coordinate descent framework;

[0033] a generation unit configured to perform data arrangement based on the mass density vector obtained by the solution to generate a three-dimensional reconstruction image of the workpiece to be measured.

[0034] wherein,

[0035] the constraint regularized hardened correction model is:

[0036] and

[0037] wherein s represents a fitted coefficient vector, p represents a mass density vector; g(p) is a regularization function of the mass density vector, J is the number of elements of the vector p, N j represents a neighborhood of pixel j, ψ(x) is a Huber potential function, and μ is a regularization parameter; is the multi-energy attenuation model, I is the projection image set; f is a preset threshold.

[0038] The two sub-models include:

[0039] The sub-model of the fixed mass density vector with respect to the fitting coefficient vector s is:

[0040]

[0041] where λ = [λ1 … λ R ] T is a Lagrange multiplier vector, ρ (i-1) represents the mass density vector after the i-1th iteration;

[0042] and the sub-model of the fixed fitting coefficient vector with respect to the mass density vector ρ is:

[0043]

[0044] where s (i) represents the fitting coefficient vector after the i th iteration.

[0045] The two sub-models include:

[0046] In another embodiment, an electronic device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the three-dimensional image reconstruction method based on CT hardening artifact correction when executing the program.

[0047] In another embodiment, a computer readable storage medium is provided, which stores a computer program, and the program is executed by a processor to implement the three-dimensional image reconstruction method based on CT hardening artifact correction.

[0048] It can be seen from the above technical solution that, in the above embodiment, starting from the multi-energy attenuation model, based on the Poisson characteristics of the obtained projection image set, and using the maximum likelihood principle to construct the constraint regularization hard correction model of the multi-energy attenuation model; under the block coordinate descent framework, the constraint regularization hard correction model is decomposed into two sub-models to solve the fitting coefficient vector and the mass density vector alternately; and based on the mass density vector obtained by solving, the data arrangement is performed to generate the three-dimensional reconstruction image of the workpiece to be measured. The scheme realizes adaptive calculation of the X-ray energy attenuation model under unknown physical parameters, realizes beam hardening correction in a blind scene, so as to realize effective beam hardening correction under unknown physical parameters. Therefore, the CT hardening artifact is corrected without additional devices and prior information, and the three-dimensional image reconstruction can be accurately realized. 。 BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0050] Figure 1 A three-dimensional image reconstruction flowchart based on CT hardening artifact correction in the embodiments of the present application;

[0051] Figure 2 A three-dimensional image reconstruction flowchart based on CT hardening artifact correction in the embodiments of the present application;

[0052] Figure 3 A three-dimensional image reconstruction device structure diagram based on CT hardening artifact correction in the embodiments of the present application;

[0053] Figure 4 A physical structure diagram of an electronic device provided in the embodiments of the present application. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0055] The terms "first", "second", "third", "fourth" and the like in the description and in the claims of the present application, and above-mentioned drawings, if any, are used to distinguish between similar objects and are not necessarily used to describe a required or chronological order. It is to be understood that the use of the terms so-termed, where appropriate, can be interchanged, so that the embodiments of the application described herein can be carried out in other than the order shown or described herein. Furthermore, the terms "comprise" and "have", and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that comprises a list of steps or units is not necessarily limited to those steps or units that are clearly listed, but can include other steps or units not clearly listed or inherent to such processes, methods, products, or apparatuses.

[0056] The technical solutions of the present application will be described in detail below with specific examples. The following specific examples can be combined with each other, and the same or similar concepts or processes may not be described in detail in some examples.

[0057] In the X-ray CT imaging process, it is often affected by the hardening artifact, which is mainly caused by the energy change of X-rays when penetrating different density materials, resulting in uneven brightness of the image.

[0058] Currently, a variety of correction techniques have been proposed, such as X-ray spectrum estimation, nonlinear model parameter estimation, etc. X-ray spectrum estimation plays an important role in dual-energy X-ray CT imaging, CT image hardening correction, quantitative analysis of CT imaging, etc. It can provide important information about the energy distribution of the X-ray beam, which helps to improve image quality and more accurate material identification. However, the ill-posed nature of X-ray spectrum estimation can make it very sensitive to measurement errors. And when there is scatter in the CT scan data, it will increase the difficulty of spectrum estimation, because scatter will change the energy distribution of X-rays, affecting the accuracy of estimation. In actual experiments, in order to reduce the influence of scatter on spectrum estimation, a collimator is usually added in front of the X-ray source and the detector, thereby affecting the engineering applicability.

[0059] Nonlinear model parameter estimation can handle complex relationships between variables and provide more accurate predictions than linear models. However, nonlinear models are usually more complex than linear models and require more expertise to construct and interpret. Under the assumption of normally distributed errors, the estimation of nonlinear regression parameters can be done by maximum likelihood estimation or least squares method, which provides statistical properties of parameter estimation. Parameter estimation is not as intuitive as linear regression, and numerical optimization methods may be needed. Nonlinear optimization may have multiple local minima, causing the algorithm to fall into a non-global minimum.

[0060] The technical solution closest to the present application is a CT ray hardening artifact correction based on energy spectrum analysis. The energy spectrum distribution data of different penetration thicknesses when X-rays penetrate a sample is obtained by using an energy-resolving detector, the photon number corresponding to different penetration thicknesses is determined, the photons in a preset energy range are intercepted, the penetration coefficient and the hardening coefficient are determined, and the fitting relationship between the two is obtained. The hardening correction of the cone beam projection image of the sample is performed by using the fitting relationship, the projection data after hardening correction is obtained, and finally the ray hardening correction image is obtained by image reconstruction. This method analyzes the energy spectrum when the X-rays scan the sample, obtains the hardening correction relationship function, and thus performs hardening correction on the sample projection data.

[0061] However, this hardening artifact correction method requires the following prior information: the energy spectrum distribution data when the X-rays penetrate the sample, the penetration coefficient and the hardening coefficient corresponding to different penetration thicknesses, and the energy range preset based on the energy value corresponding to the energy spectrum peak in the energy spectrum distribution curve of the maximum penetration thickness to the maximum ray energy value; and the energy-resolving detector is relatively expensive and has a lower spatial resolution than the non-energy-resolving detector, which may affect the clarity of the image and increase the difficulty of application.

[0062] Therefore, there is an urgent need for a method for correcting CT hardening artifacts without additional devices and prior information, and thus accurately reconstructing a three-dimensional image.

[0063] Based on the above problems, the present application provides a three-dimensional image reconstruction based on CT hardening artifact correction and a device. In the embodiments of the present application, the physical model (multi-energy attenuation model) of X-ray CT imaging is used as a starting point, the Poisson characteristics of the obtained projection image set are used as a basis, and the maximum likelihood principle is used to construct a constrained regularization hardening correction model. In the block coordinate descent framework, the constrained regularization hardening correction model is decomposed into two sub-models to solve the fitting coefficient vector and the mass density vector alternately; and based on the mass density vector obtained by solving, the data is arranged to generate a three-dimensional reconstruction image of the workpiece to be measured. This scheme realizes adaptive calculation of the X-ray energy attenuation model under unknown physical parameters, realizes beam hardening calibration in a blind scene, and thus realizes effective beam hardening calibration under unknown physical parameters. Therefore, the present application corrects the CT hardening artifact without additional devices and prior information, and thus accurately reconstructs a three-dimensional image.

[0064] The process of the present application for three-dimensional image reconstruction based on CT hardening artifact correction will be described in detail below with reference to the accompanying drawings.

[0065] Referring to Figure 1 , Figure 1 A three-dimensional image reconstruction process based on CT hardening artifact correction is shown in the flowchart of the embodiments of the present application. The specific steps are:

[0066] In step 101, a projection image set of a workpiece to be measured is obtained; wherein the projection image set is a set of images obtained by a scanner performing complete scanning on the workpiece to be measured.

[0067] The workpiece to be measured in the embodiment of the present application is a workpiece which needs to be reconstructed in three dimensions.

[0068] The scanner can be set as follows:

[0069] The collected image gray scale range is set to 0-65535, the scanning voltage is 430kV, the scanning current is 1600μA, the source-to-detector distance is 1158mm, and the source-to-workpiece distance is 500mm.

[0070] In the specific implementation, the settings can be made according to the application environment and actual needs, and the embodiment of the present application does not limit this.

[0071] Before complete scanning on the workpiece to be measured, the workpiece to be measured is directly placed on the object table, and no device is added between the ray source, the workpiece to be measured and the detector, and then direct scanning is performed to obtain.

[0072] In step 102, a multi-energy attenuation model based on the fitting coefficient vector and the mass density vector is obtained; wherein the multi-energy attenuation model is used to represent the noiseless measurement value received by the energy integrating detector.

[0073] The multi-energy attenuation model in the embodiment of the present application is established for all workpieces, that is, the model can be used when all workpieces to be measured need to be reconstructed in three dimensions.

[0074] The multi-energy attenuation model is used to determine the noiseless measurement value, and can be represented as:

[0075]

[0076] Wherein s represents the fitting coefficient vector, and p represents the mass density vector. is an output basis function matrix obtained by stacking 1xR vectors column by column, and R is the number of basis functions.

[0077] The specific process of establishing the multi-energy attenuation model is given as follows:

[0078] The noiseless measurement value system collected by the energy integrating detector along the straight line l = l(x, y, z) in the Cartesian coordinate system has a superposition integral form, ∫ l dl represents the integral of the ray path, I out is a nonlinear function of the density map a(x, y, z), and represents the noiseless measurement value received by the energy integrating detector, and the initial physical model is:​

[0079] I out = 1 L (∫1α(x, y, z)dl);

[0080] where, l L (δ) = ∫l(k)e -δk dk is the Laplace transform of the mass attenuation spectrum l(k), representing the density of incident X-ray energy at attenuation coefficient k, after Laplace conversion:

[0081] ∫l(k)exp(-k∫1α(x, y, z)dl)dk;

[0082] In the spatial domain, it is discretized into J pixels, and the integral ∫1α(x, y, z)dl is approximated as ρ≥0 is a J×1 vector representing the three-dimensional image we want to reconstruct, is a J×1 vector of known weights, and is the length of the ray path intersecting the pixel, quantifying the contribution of each element of ρ to the X-ray attenuation of l along the straight path. Denote the total number of measurements collected by the energy-integrating detector array as N. For the nth measurement, define its discrete line integral as Stack all N such integrals into a vector to obtain Φρ. Approximate l(k) as a linear combination of R (R << N) basis functions, so that l(k) = b(k)s. Through the B1 spline fitting technique, the physical parameters in the X-ray attenuation process can be adaptively fitted:

[0083] b(k) is composed of first-order B-splines (referred to as B1 splines),

[0084] is an output basis function matrix obtained by stacking 1×R vectors column by column, and the multiple energy attenuation model is represented by the basis function matrix:

[0085]

[0086] Step 103, based on the Poisson characteristics of the projection image set, and the multiple energy attenuation model, a constrained regularization hard correction model is constructed using the maximum likelihood principle.

[0087] It is relatively complex to directly solve s and ρ in the multiple energy attenuation model. In the embodiment of the present application, starting from the multiple energy attenuation model, according to the Poisson statistical characteristics of the actual measurement data set (projection image set I) and the multiple energy attenuation model, a constrained regularization optimization problem is established using the maximum likelihood principle, that is, a constrained regularization hard correction model is used to solve the fitting coefficients s and the mass density ρ. ​​

[0088] Step 104, under the block coordinate descent framework, the constraint regularized hardened correction model is disassembled into two sub-models to alternately solve the fitting coefficient vector and the mass density vector.

[0089] Step 105, based on the mass density vector obtained by solving, data arrangement is performed to generate a three-dimensional reconstruction image of the workpiece to be measured.

[0090] In this embodiment, starting from the physical model (multi-energy attenuation model) of X-ray CT imaging, based on the Poisson characteristics of the obtained projection image set, and using the maximum likelihood principle, a constraint regularized hardened correction model is constructed; under the block coordinate descent framework, the constraint regularized hardened correction model is disassembled into two sub-models to alternately solve the fitting coefficient vector and the mass density vector; and based on the mass density vector obtained by solving, data arrangement is performed to generate a three-dimensional reconstruction image of the workpiece to be measured. This scheme realizes adaptive calculation of the X-ray energy attenuation model under unknown physical parameters, realizes beam hardening calibration in a blind scene, and thus realizes effective beam hardening calibration under unknown physical parameters. Therefore, the CT hardening artifact is corrected without additional devices and prior information, and the three-dimensional image reconstruction can be accurately realized.

[0091] In another example, the constraint regularized hardened correction model established is:

[0092]

[0093] and

[0094] wherein s represents the fitting coefficient vector, and p represents the mass density vector; g(p) is a regularization function of the mass density vector, J is the number of elements of the vector p, and is also the number of pixels, N j represents the neighborhood of pixel j, and ψ(x) is a Huber potential function, and μ is a regularization parameter; is a multi-energy attenuation model, I is a projection image set; f is a preset threshold, and the Huber potential function is represented by x as an unknown quantity, which is equivalent to x = p j - p j′ ; wherein f|x| = f x |x|.

[0095] In another example, the two sub-models in step 104 include:

[0096] The sub-model about the fitting coefficient vector s with the fixed mass density vector p is:

[0097]

[0098] wherein λ = [λ1 … λ R ]T is a Lagrange multiplier vector, p (i-1) denotes the mass density vector after the i-1th iteration;

[0099] and the sub-model of the fixed fitting coefficient vector with respect to the mass density vector p is:

[0100]

[0101] where S (i) denotes the fitting coefficient vector after the ith iteration;

[0102] and the fitting coefficient vector is solved using a non-negative matrix solution algorithm for the sub-model with respect to the fitting coefficient vector;

[0103] the mass density vector is solved using a non-negative matrix solution algorithm for the sub-model with respect to the mass density vector.

[0104] Referring to Figure 2 , Figure 2 is another three-dimensional image reconstruction flowchart based on CT hardening artifact correction in the embodiments of the present application. The specific steps are as follows:

[0105] Step 201, obtaining a projection image set of a workpiece to be measured; wherein the projection image set is a set of images obtained by a scanner performing complete scanning on the workpiece to be measured.

[0106] The workpiece to be measured in the embodiments of the present application is a workpiece that needs to be reconstructed into a three-dimensional image.

[0107] The scanner can be set as follows:

[0108] The acquisition image gray scale range is set to 0-65535, the scanning voltage is 430kV, the scanning current is 1600μA, the source-to-detector distance is 1158mm, and the source-to-workpiece distance is 500mm.

[0109] In specific implementation, the application environment and actual needs can be set, and the embodiments of the present application do not limit this.

[0110] Before complete scanning of the workpiece to be measured, the measured workpiece is directly placed on the object table, and no equipment is added between the ray source, the measured workpiece and the detector, and then direct scanning is performed to obtain.

[0111] Step 202, obtaining a multi-energy attenuation model established based on the fitting coefficient vector and the mass density vector; wherein the multi-energy attenuation model is used to represent a noiseless measurement value received by an energy-integrating detector.

[0112] The multi-energy attenuation model is used to determine the noiseless measurement value, and can be expressed as:

[0113]

[0114] where s denotes the fitting coefficient vector, and p denotes the mass density vector; is obtained by stacking 1 x R vectors column by column , R is the number of basis functions.

[0115] Step 203, based on the Poisson characteristics of the projection image set and the multi-energy attenuation model, a constrained regularization hard correction model is constructed using the maximum likelihood principle.

[0116] It is relatively complex to directly solve s and p in the multi-energy attenuation model. In the embodiment of the application, starting from the multi-energy attenuation model, according to the Poisson statistical characteristics of the actual measurement data set (the projection image set I) and the multi-energy attenuation model, a constrained regularization optimization problem is established using the maximum likelihood principle, that is, a constrained regularization hard correction model is established to solve the fitting coefficient s and the mass density p.

[0117] where the established constrained regularization hard correction model is:

[0118]

[0119] and

[0120] where s denotes the fitting coefficient vector, and p denotes the mass density vector; g(p) is a regularization function of the mass density vector, J is the number of elements of the vector p, N j denotes the neighborhood of pixel j, and ψ(x) is a Huber potential function, and μ is a regularization parameter; is a multi-energy attenuation model, and I is a projection image set; is a fidelity term, which is a weighted least square of the actual CT measurement data and the theoretical value; μg(p) is used to suppress noise; f is a preset threshold value, which is used to control the size of the discontinuity of the model; the Huber potential function is represented by x as an unknown quantity, which is equivalent to x = p j -p j′ ; where f|x| = f x |x|.

[0121] Step 204, under the block coordinate descent framework, the constrained regularization hard correction model is decomposed into a sub-model about the fitting coefficient vector with the fixed mass density vector, and a sub-model about the mass density vector with the fixed fitting coefficient vector.

[0122] The sub-model about the fitting coefficient vector s with the fixed mass density vector in this step is:

[0123]

[0124] where λ = [λ1… λ R ] T is the Lagrange multiplier vector, ρ (i-1) is the mass density vector after the i-1th iteration.

[0125] The sub-model of the fixed fitting coefficient vector with respect to the mass density vector ρ is:

[0126]

[0127] where s (i) is the fitting coefficient vector after the ith iteration.

[0128] Step 205, use the non-negative matrix to solve the fitting coefficient vector and the mass density vector respectively for the two sub-models.

[0129] When the mass density vector is fixed, the problem of solving with respect to s is linear, and we can use the non-negative matrix (NMF) to solve it, which is:

[0130]

[0131] Multiply both sides of the equation by After calculation, the iteration formula is:

[0132]

[0133] and respectively represent the fitting coefficient results after the i-1th and ith iterations.

[0134] When the fitting coefficient vector is fixed, the problem of solving with respect to ρ is nonlinear, and for this problem, we can also use NMF to solve it.

[0135]

[0136] Multiply both sides of the equation by After calculation, the iteration formula is:

[0137]

[0138]

[0139] and represent the mass density results after the i-1th and ith iterations, and k is the attenuation coefficient.

[0140] The relevant variables with * marks in the above formula represent the optimal solution of the corresponding variable.

[0141] Step 206, based on the obtained quality density vector, data arrangement is performed to generate the three-dimensional reconstruction image of the workpiece to be measured.

[0142] In the embodiment, based on the Poisson characteristics of the obtained projection image set and the multi-energy attenuation model, a constrained regularization hardened correction model is constructed using the maximum likelihood principle; under the block coordinate descent framework, the constrained regularization hardened correction model is decomposed into two sub-models, and the non-negative matrix method is used to solve the fitting coefficient vector and the quality density vector, respectively; and based on the obtained quality density vector, data arrangement is performed to generate the three-dimensional reconstruction image of the workpiece to be measured. The scheme realizes adaptive calculation of the X-ray energy attenuation model under unknown physical parameters, realizes beam hardening correction in a blind scene, and thus realizes effective beam hardening correction under unknown physical parameters. Therefore, the CT hardening artifact is corrected without additional devices and prior information, and thus accurate three-dimensional image reconstruction can be realized.

[0143] All the optional technical solutions described above can be combined to form optional embodiments of the disclosure, which will not be described here.

[0144] Based on the same inventive concept, the embodiment of the present application also provides a three-dimensional image reconstruction device based on CT hardening artifact correction. Referring to Figure 3 , Figure 3 The structure diagram of the three-dimensional image reconstruction device based on CT hardening artifact correction in the embodiment of the present application is shown in the figure. The device comprises:

[0145] The first acquisition unit 301 is configured to obtain a projection image set of a workpiece to be measured; wherein the projection image set is obtained by a scanner performing complete scanning on the workpiece to be measured;

[0146] The second acquisition unit 302 is configured to acquire a multi-energy attenuation model based on the fitting coefficient vector and the quality density vector; wherein the multi-energy attenuation model is used to represent the noiseless measurement value received by the energy integral detector;

[0147] The establishment unit 303 is configured to construct a constrained regularization hardened correction model based on the Poisson characteristics of the projection image set and the multi-energy attenuation model using the maximum likelihood principle;

[0148] The solving unit 304 is configured to decompose the constrained regularization hardened correction model into two sub-models under the block coordinate descent framework to alternately solve the fitting coefficient vector and the quality density vector;

[0149] The generation unit 305 is configured to perform data arrangement based on the obtained quality density vector to generate a three-dimensional reconstruction image of the workpiece to be measured.

[0150] In another example,

[0151] The constraint regularization hard correction model is:

[0152] And Where s represents a fitting coefficient vector, and p represents a mass density vector; g(p) is a regularization function of the mass density vector, J is the number of elements of the vector p, and N j Denotes a neighborhood of pixel j, and ψ(x) is a Huber potential function, and μ is a regularization parameter; Is a multi-energy attenuation model, I is a projection image set; f is a preset threshold.

[0153] In another example, the two sub-models include:

[0154] The sub-model about the fitting coefficient vector s with the fixed mass density vector p is:

[0155]

[0156] Where λ = [λ1 … λ R ] T Is a Lagrange multiplier vector, and p (i-1) Denotes the mass density vector after the i-1th iteration;

[0157] And the sub-model about the mass density vector p with the fixed fitting coefficient vector s is:

[0158]

[0159] Where s (i) Denotes the fitting coefficient vector after the i th iteration.

[0160] In another example, the constraint regularization hard correction model is disassembled into two sub-models to alternately solve the fitting coefficient vector and the mass density vector, including: using a non-negative matrix solution algorithm to solve the fitting coefficient vector for the sub-model about the fitting coefficient vector s; and using a non-negative matrix solution algorithm to solve the mass density vector for the sub-model about the mass density vector p.

[0161] The units of the above embodiments can be integrated or deployed separately; can be combined into one unit, or can be further split into multiple sub-units.

[0162] In another embodiment, an electronic device is also provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the three-dimensional image reconstruction method based on CT hardening artifact correction when executing the program.

[0163] In another embodiment, a computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement a three-dimensional image reconstruction method based on CT hardening artifact correction is also provided.

[0164] Figure 4 The physical structure of the electronic device is provided for the embodiments of the present application. As shown in the figure, the electronic device can include a processor (Processor) 410, a communications interface (Communications Interface) 420, a memory (Memory) 430, and a communications bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other through the communications bus 440. The processor 410 can invoke the logical instructions in the memory 430 to execute the following method: Figure 4

[0165] Obtaining a projection image set of a workpiece to be measured; wherein the projection image set is obtained by a scanner performing a complete scan on the workpiece to be measured;

[0166] Obtaining a multi-energy attenuation model based on a fitting coefficient vector and a mass density vector; wherein the multi-energy attenuation model is used to represent a noise-free measurement value received by an energy-integrating detector;

[0167] Based on the Poisson property of the projection image set and the multi-energy attenuation model, a constraint regularization hardening correction model is constructed using the maximum likelihood principle;

[0168] In the block coordinate descent framework, the constraint regularization hardening correction model is decomposed into two sub-models to alternately solve the fitting coefficient vector and the mass density vector;

[0169] Based on the obtained mass density vector, data is arranged to generate a three-dimensional reconstruction image of the workpiece to be measured.

[0170] In addition, the logical instructions in the memory 430 described above can be implemented in the form of a software function unit and sold or used as an independent product, which can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or parts of the present application that essentially contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the embodiments of the present application. The aforementioned storage medium includes a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.​

[0171] The apparatus embodiments described above are only illustrative, wherein the units described as separate components can or can not be physically separate, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0172] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software and the necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0173] The flowcharts and block diagrams in the drawings of the present application show the possible implementation architecture, function and operation of the system, method and computer program product according to various embodiments disclosed in the present application. In this regard, each block in the flowchart or block diagram can represent a module, a program segment, or a part of code, which contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions noted in the blocks can occur in different orders in different diagrams. For example, two connected blocks can actually be executed substantially in parallel, and sometimes they can be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagram or flowchart, and the combination of blocks in the block diagram or flowchart, can be realized by a dedicated hardware-based system that performs the specified functions or operations, or can be realized by a combination of special-purpose hardware and computer instructions.

[0174] Those skilled in the art can understand that the features disclosed in the various embodiments and / or claims of the present application can be combined and / or integrated in various combinations and / or integrations, even if such combinations or integrations are not explicitly described in the present application. In particular, the features disclosed in the various embodiments and / or claims of the present application can be combined and / or integrated in various combinations and / or integrations without departing from the spirit and teachings of the present application, and all such combinations and / or integrations fall within the scope of the present disclosure.

[0175] The principles and implementation manners of the present application are described by using specific examples in the present application, and the above examples are only used for helping to understand the method of the present application and its core idea, and are not used for limiting the present application. For those skilled in the art, according to the idea, spirit and principle of the present application, the specific implementation manners and application ranges can be changed, and any modification, equivalent replacement, improvement and the like made by those skilled in the art should be included in the protection scope of the present application.

Claims

1. A three-dimensional image reconstruction method based on CT hardening artifact correction, characterized in that, The method includes: Obtain a set of projected images of the workpiece to be tested; wherein, the set of projected images is obtained by a scanner performing a complete scan of the workpiece to be tested; A multi-energy decay model is established based on the fitting coefficient vector and the mass density vector; wherein, the multi-energy decay model is used to represent the noise-free measurement value received by the energy integration detector; Based on the Poisson properties of the projected image set, and the constrained regularization hardening correction model constructed using the maximum likelihood principle of the multi-energy decay model; Within the block coordinate descent framework, the constrained regularization hardening correction model is decomposed into two sub-models that alternately solve for the fitting coefficient vector and the mass density vector; Based on the mass density vector obtained by the solution, the data is arranged to generate a three-dimensional reconstructed image of the workpiece under test; The constraint regularization hardening correction model is as follows: ； and ; in, Represents the vector of fitted coefficients. Represents the mass density vector; Let J be the regularization function for the mass density vector, where J is a vector. The number of elements, where Nj represents the neighborhood of pixel j. Let H be the Huber potential function. For regularization parameters; For the aforementioned multi-energy decay model, The projected image set is defined as f; f is a preset threshold. The two sub-models include: The sub-model of the fixed mass density vector with respect to the fitting coefficient vector s is: ； in, Let Lagrange multiplier vectors be used. This represents the mass density vector after the (i-1)th iteration; and the mass density vector with a fixed fitting coefficient vector. The sub-model is: ； in, This represents the vector of fitting coefficients after the i-th iteration; The step of decomposing the constraint regularization hardening correction model into two sub-models and alternately solving for the fitting coefficient vector and the mass density vector includes: The fitting coefficient vector is solved using a nonnegative matrix solving algorithm for the sub-model with respect to the fitting coefficient vector s; For the mass density vector The sub-model uses a nonnegative matrix solving algorithm to solve for the mass density vector.

2. The method according to claim 1, characterized in that, The method further includes: The projected image set is obtained by placing the workpiece to be tested directly on the stage before the scanner performs a complete scan of the workpiece. No equipment is added between the X-ray source, the workpiece to be tested, and the detector, and then the scan is performed directly.

3. A three-dimensional image reconstruction device based on CT hardening artifact correction, characterized in that, The device includes: The first acquisition unit is used to acquire a set of projected images of the workpiece to be tested; wherein the set of projected images is obtained by a scanner performing a complete scan of the workpiece to be tested; The second acquisition unit is used to acquire a multi-energy decay model based on the fitting coefficient vector and the mass density vector; wherein, the multi-energy decay model is used to represent the noise-free measurement value received by the energy integration detector; Establishment unit, used to construct a constrained regularization hardening correction model based on the Poisson characteristics of the projected image set and the multi-energy decay model using the maximum likelihood principle; The solving unit is used to decompose the constraint regularization hardening correction model into two sub-models under the block coordinate descent framework and alternately solve the fitting coefficient vector and the mass density vector; The generation unit is used to arrange the data based on the mass density vector obtained by solving, and generate a three-dimensional reconstructed image of the workpiece to be tested; The constraint regularization hardening correction model is as follows: ; and ; ,in, Represents the vector of fitted coefficients. Represents the mass density vector; Let J be the regularization function for the mass density vector, where J is a vector. The number of elements, N j Describes the neighborhood of pixel j. Let H be the Huber potential function. For regularization parameters; For the aforementioned multi-energy decay model, The projected image set is defined as f; f is a preset threshold. The two sub-models include: The sub-model of the fixed mass density vector with respect to the fitting coefficient vector s is: ； in, Let Lagrange multiplier vectors be used. This represents the mass density vector after the (i-1)th iteration; and the mass density vector with a fixed fitting coefficient vector. The sub-model is: ； in, This represents the vector of fitting coefficients after the i-th iteration; The solution unit is specifically used to solve for the fitting coefficient vector using a nonnegative matrix solving algorithm on the sub-model with respect to the fitting coefficient vector s; for the mass density vector... The sub-model uses a nonnegative matrix solving algorithm to solve for the mass density vector.

4. The apparatus according to claim 3, characterized in that, The projected image set is obtained by placing the workpiece to be tested directly on the stage before the scanner performs a complete scan of the workpiece. No equipment is added between the X-ray source, the workpiece to be tested, and the detector, and then the image set is obtained directly by scanning.

5. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method of claim 1 or 2.

6. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method of claim 1 or 2.

Images