A CT image ring artifact correction method and device based on double-domain regularization

Abstract

This invention discloses a method and apparatus for correcting ring artifacts in CT images based on dual-domain regularization. The method includes: Step 1, acquiring a CT scan dataset; Step 2, initialization; Step 3, using the image reconstruction operator R... G Update reconstructed image x (k) The intermediate image x is obtained. (k+1 / 2) Step 4: Using the operator T that constrains the gradient sparsity of the image, process the intermediate image x. (k+1 / 2) Constraints are applied to obtain the reconstructed image x. (k+1) Step 5: Using the operator Q that constrains the gradient sparsity and group sparsity of the vertical stripe artifact image, process the vertical stripe artifact image S... (k) Constraints are applied to obtain the vertical stripe artifact image S. (k+1) Update the detector's response inconsistency compensation coefficient; Step 6, determine if the upper limit of the number of iterations has been reached, and if so, output the reconstructed image x. (k+1) and vertical stripe artifact image S (k+1) If not, proceed to step 3. This invention overcomes the limitation of existing algorithms that assume the response inconsistency compensation coefficient is independent of the angle, and better corrects ring artifacts in CT images.

Description

Technical Field

[0001] This invention relates to the field of X-ray CT (Computed Tomography) image processing technology, and in particular to a method and apparatus for correcting ring artifacts in CT images based on dual-domain regularization. Background Technology

[0002] The basic process of X-ray CT imaging involves placing the object under test on a sample stage, turning on the X-ray source to generate an X-ray beam, and simultaneously turning on the detector to detect the X-rays that interact with the sample. The sample is then rotated one revolution relative to the X-ray source and detector to acquire CT scan data. The scan data is processed, and a reconstruction algorithm is used to obtain a CT image reflecting the internal structure of the object. During CT imaging, various factors affect the quality of the CT image, such as the X-ray energy spectrum distribution, the stability of the X-ray tube current, the filter material, the detector response efficiency, and the reconstruction algorithm.

[0003] Photon-counting computed tomography (PCD-CT) has become a research hotspot in X-ray imaging in recent years. Compared with traditional CT, PCD-CT has great potential in providing more image information and is of great significance for contrast imaging and soft tissue imaging. The flat-panel photon-counting detector (PCD) is an important component of PCD-CT, possessing many advantages such as no dark current and a high signal-to-noise ratio. However, its clinical application still faces significant challenges. One of the most common defects of photon-counting detectors is the non-ideal response of the detector pixels, which leads to vertical stripe artifacts in the projected data, resulting in concentric ring artifacts or semi-ring artifacts in the reconstructed images. These artifacts degrade CT images, severely affecting the quantitative analysis of the images and reducing diagnostic accuracy.

[0004] Existing methods for correcting ring artifacts can be broadly categorized into hardware correction methods and algorithmic correction methods. Hardware correction methods primarily reduce ring artifacts by continuously translating the X-ray source or detector to ignore the effects of detector response inconsistencies. However, this method is complex to implement and cannot completely eliminate ring artifacts. Algorithmic correction methods can be further divided into image domain post-processing methods and projection domain pre-processing methods. Image domain post-processing methods are typically performed in polar coordinates. A polar coordinate transformation converts the rectangular image to a polar image, then image processing methods are used to locate and eliminate vertical stripe artifacts in the image. Finally, the corrected polar image is transformed back to rectangular coordinates, but coordinate system transformations often lead to more complex calculations. Projection domain pre-processing methods typically first use image processing methods to locate the positions of vertical stripes in the sine wave, then use interpolation or filtering methods to correct the sine wave. The correction effect depends on the accuracy of stripe artifact localization and may result in the reappearance of ring artifacts in the reconstructed image or incomplete compensation of damaged low-frequency information. Recently, some scholars have proposed optimization models that simultaneously constrain the projection domain and the image domain. However, assuming that the response inconsistency compensation coefficient is independent of the projection angle, it will lead to insufficient or excessive compensation for the same detector pixel at certain angles.

[0005] In summary, existing correction methods suffer from drawbacks such as incomplete removal of ring artifacts, easy introduction of new artifacts, difficulty in guaranteeing image spatial resolution, and incorrect assumptions regarding response inconsistency compensation coefficients. Therefore, a technical solution is desired to overcome or at least mitigate at least one of the aforementioned drawbacks of existing technologies. Summary of the Invention

[0006] The purpose of this invention is to provide a method and apparatus for correcting ring artifacts in CT images based on dual-domain regularization, which can effectively remove ring artifacts in reconstructed images while preserving the structure and details of the images.

[0007] To achieve the above objectives, the present invention provides a method for correcting ring artifacts in CT images based on dual-domain regularization, comprising:

[0008] Step 1: Obtain the CT scan dataset p, and set the CT scan geometric parameter set G and the upper limit of the number of iterations N;

[0009] Step 2: Initialize the reconstructed image of the estimated acyclic artifact and the estimated vertical stripe artifact image;

[0010] Step 3, using the image reconstruction operator R G Update the reconstructed image x after the k-th iteration (k) The intermediate image x is obtained. (k +1 / 2) =R G (p,x (k)), R G Related to the set of scanning geometry parameters G;

[0011] Step 4: Using the operator T that constrains the image gradient sparsity, process the intermediate image x obtained in Step 2. (k+1 / 2) By applying constraints, we obtain the reconstructed image x after the (k+1)th iteration. (k+1) ;

[0012] Step 5: Using the operator Q that constrains the gradient sparsity and group sparsity of the vertical stripe artifact image, the vertical stripe artifact image S after the kth iteration is processed. (k) By applying constraints, we obtain the vertical stripe artifact image S after the (k+1)th iteration. (k +1) , using S (k+1) Update the detector's response inconsistency compensation coefficient S;

[0013] Step 6: Determine if the maximum number of iterations N has been reached. If so, output the reconstructed image x. (k+1) and vertical stripe artifact image S (k+1) If not, proceed to step 3.

[0014] Furthermore, the operator T is defined by the optimization problem of equation (1) as follows:

[0015]

[0016] In equation (1), Ax (k) To reconstruct image x (k) The estimated projection data obtained after forward projection, λ1 is a pre-set adjustment parameter used to control the weights of the regularization term and the data fidelity term, D H x、D V x represents the gradients in the horizontal and vertical directions of the reconstructed image, respectively, ||·||2 represents the l2 norm, and ||·||1 represents the l1 norm.

[0017] Furthermore, the operator Q is defined by the optimization problem of the following equation (2):

[0018]

[0019] In equation (2), Ax (k+1) To reconstruct image x (k+1) The estimated projection data obtained after forward projection, where λ2 is the pre-set adjustment parameter used for the weights of the gradient constraint regularization term and the data fidelity term, and D V S represents the gradient of the vertical stripe artifact image S in the vertical direction, ||·||1 represents the l1 norm, λ3 is a pre-defined adjustment parameter for the weights of the group sparsity constraint regularization term and the data fidelity term, Φ(S) is the set of vertical stripes, ||·||2,1 Indicate l 2,1 Norm.

[0020] Furthermore, the image reconstruction operator R G It can be obtained using iterative reconstruction algorithms or analytical reconstruction algorithms.

[0021] The present invention also provides a CT image ring artifact correction device based on dual-domain regularization, comprising:

[0022] The data input module is used to input the CT scan dataset p, the CT scan geometric parameter set G, and the upper limit of the number of iterations N;

[0023] An initialization module is used to initialize the reconstructed image of the estimated acyclic artifact and the estimated vertical stripe artifact image.

[0024] Iterative processing estimation module for image reconstruction operator R G Update the reconstructed image x after the k-th iteration (k) The intermediate image x is obtained. (k+1 / 2) =R G (p,x (k) ), R G Related to the set of scanning geometry parameters G;

[0025] The image gradient sparsity constraint module is used to constrain the image gradient sparsity using the operator T, which is used to process the intermediate image x obtained in step 2. (k+1 / 2) By applying constraints, we obtain the reconstructed image x after the (k+1)th iteration. (k+1) ;

[0026] The group sparsity constraint module is used to apply the operator Q, which constrains the gradient sparsity and group sparsity of the vertical stripe artifact image, to the vertical stripe artifact image S after the kth iteration. (k) By applying constraints, we obtain the vertical stripe artifact image S after the (k+1)th iteration. (k+1) , using S (k+1) Update the detector's response inconsistency compensation coefficient S;

[0027] The judgment module is used to determine whether the upper limit of the number of iterations N has been reached. If so, the reconstructed image x is output. (k+1) and vertical stripe artifact image S (k+1) If not, the iterative estimation module will continue to obtain the intermediate image x. (k+1 / 2) .

[0028] Furthermore, the operator T is defined by the optimization problem of equation (1) as follows:

[0029]

[0030] In equation (1), Ax(k) To reconstruct image x (k) The estimated projection data obtained after forward projection, λ1 is a pre-set adjustment parameter used to control the weights of the regularization term and the data fidelity term, D H x、D V x represents the gradients in the horizontal and vertical directions of the reconstructed image, respectively, ||·||2 represents the l2 norm, and ||·||1 represents the l1 norm.

[0031] Furthermore, the operator Q is defined by the optimization problem of the following equation (2):

[0032]

[0033] In equation (2), Ax (k+1) To reconstruct image x (k+1) The estimated projection data obtained after forward projection, where λ2 is the pre-set adjustment parameter used for the weights of the gradient constraint regularization term and the data fidelity term, and D V S represents the gradient of the vertical stripe artifact image S in the vertical direction, ||·||1 represents the l1 norm, λ3 is a pre-defined adjustment parameter for the weights of the group sparsity constraint regularization term and the data fidelity term, Φ(S) is the set of vertical stripes, ||·|| 2,1 Indicate l 2,1 Norm.

[0034] Furthermore, the image reconstruction operator R G It can be obtained using iterative reconstruction algorithms or analytical reconstruction algorithms.

[0035] This invention considers both the sparse prior of the reconstructed image in the image domain and the sparse prior of vertical fringe artifacts in the projection domain. It penalizes ring artifacts in the image domain by leveraging the sparse prior of the reconstructed image, and updates the detector's response inconsistency compensation coefficient by applying group sparsity and sparsity constraints along the vertical fringe direction to the vertical fringe artifact image, thereby correcting the projection image. Unlike existing methods, this invention assumes that the detector's response inconsistency compensation coefficient is related to the projection angle, thus enabling a more reasonable correction of response inconsistencies between detectors. Attached Figure Description

[0036] Figure 1 This is a flowchart of a CT image ring artifact correction method based on dual-domain regularization, as described in an embodiment of the present invention.

[0037] Figure 2 The image shown is a scanned phantom image from an embodiment of the present invention.

[0038] Figure 3 This is a reconstructed image of SART in an embodiment of the present invention;

[0039] Figure 4 The image is a reconstructed image obtained using an embodiment of the present invention. Detailed Implementation

[0040] In the accompanying drawings, the same or similar reference numerals are used to denote the same or similar elements or elements having the same or similar functions. The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0041] like Figure 1 As shown, the CT image ring artifact correction method based on dual-domain regularization provided in this embodiment of the invention includes:

[0042] Step 1: Obtain the CT scan dataset p, and set the CT scan geometric parameter set G, the upper limit of the number of iterations N, and the upper limit of the number of sub-iterations K. The CT scan geometric parameter set G includes the distance from the X-ray source to the detector center (SDD), the distance from the X-ray source to the turntable center (SOD), the number of detector elements, the detector element size, the number of scanning angles, and the angle sampling interval.

[0043] Step 2: Initialize the reconstructed image with estimated acyclic artifacts and the estimated vertical stripe artifact image. The initialized reconstructed image is represented as x. (0) The initialized vertical stripe artifact image is represented as S. (0) Generally, all pixel values ​​are set to 0.

[0044] Step 3, iteratively process and estimate the image x (k) k = 0, 1, ..., N, specifically including:

[0045] Image reconstruction operator R G Update the reconstructed image x after the k-th iteration (k) The intermediate image x is obtained. (k+1 / 2) =R G (p,x (k) ), R G The set of scanning geometric parameters G can be obtained using either iterative or analytical reconstruction algorithms. Specifically, the iterative reconstruction algorithms include ART, SART, and EM; the analytical reconstruction algorithms include FBP and BPF.

[0046] Step 4: Using the operator T that constrains the image gradient sparsity, process the intermediate image x obtained in Step 2. (k+1 / 2) By applying constraints, we obtain the reconstructed image x after the (k+1)th iteration. (k+1) .

[0047] Step 5: Using the operator Q that constrains the gradient sparsity and group sparsity of the vertical stripe artifact image, the vertical stripe artifact image S after the kth iteration is processed. (k)By applying constraints, we obtain the vertical stripe artifact image S after the (k+1)th iteration. (k +1) , using S (k+1) Update the detector's response inconsistency compensation coefficient S.

[0048] Step 6: Determine if the maximum number of iterations N has been reached. If not, go back to step 3; if yes, terminate the iteration and output the reconstructed image x. (k+1) and vertical stripe artifact image S (k+1) .

[0049] The method provided in this embodiment considers both the sparse prior of the reconstructed image and the sparse prior of the vertical stripe artifact image. The sparse prior of the reconstructed image is used to penalize ring artifacts in the image domain. The detector's response inconsistency compensation coefficient is updated by applying group sparsity and image gradient sparsity constraints along the vertical stripe direction to the vertical stripe artifact image, thereby correcting the projected image. Unlike existing methods, we assume that the detector's response inconsistency compensation coefficient is related to the projection angle, thus enabling a more reasonable correction of response inconsistencies between detectors.

[0050] In the above embodiments, the operator T is defined by the optimization problem of the following equation (1), which can be solved by methods such as the ADMM method:

[0051]

[0052] In equation (1), For data fidelity items, Ax (k) To reconstruct image x (k) The estimated projection data obtained after forward projection, λ1(||D H x||1+||D V x||1) represents the regularization term, and λ1 is a pre-set adjustment parameter used to control the weights of the regularization term and the data fidelity term. Its specific value is obtained experimentally, for example, 0.0008, D H x、D V x represents the gradients in the horizontal and vertical directions of the reconstructed image, respectively, ||·||2 represents the l2 norm, and ||·||1 represents the l1 norm.

[0053] In this embodiment, sparse prior constraints are applied to the reconstructed image using equation (1) to achieve the effects of denoising the image and removing some ring artifacts, thereby obtaining an estimated reconstructed image without ring artifacts.

[0054] Of course, in addition to the methods provided in the above embodiments, the operator T can also be obtained using existing methods that constrain image sparsity, such as the three-dimensional block matching filter algorithm (bm3d) or the nonlocal mean filter algorithm.

[0055] In one embodiment, in step 5, the operator Q is defined by the optimization problem of equation (2) as follows:

[0056]

[0057] In equation (2), λ2‖D v S‖1+λ3||Φ(S)|| 2,1 As a regularization term, Ax (k+1) To reconstruct image x (k+1) The estimated projected data obtained after forward projection, where λ2 is a pre-set adjustment parameter for the weights of the gradient constraint regularization term and the data fidelity term, and λ3 is a pre-set adjustment parameter for the weights of the group sparsity constraint regularization term and the data fidelity term, with specific values ​​obtained experimentally. In a specific example, the values ​​are λ2 = 1, λ3 = 0.12, and D... V S represents the gradient of the vertical stripe artifact image S in the vertical direction, ||·||1 represents the l1 norm, and Φ(S) is the set of vertical stripes. 2,1 Indicate l 2,1 Norm.

[0058] In this embodiment, equation (2) is used to apply sparse prior constraints to the vertical stripe artifact image on the projection domain, which achieves the effect of extracting the vertical stripe artifact image in this invention. Equation (2) is used to obtain the estimated vertical stripe artifact image.

[0059] This invention also provides a CT image ring artifact correction device based on dual-domain regularization, which includes a data input module, an initialization module, an iterative processing estimation module, an image gradient sparsity constraint module, a group sparsity constraint module, and a judgment module, wherein:

[0060] The data input module is used to input the CT scan dataset p, the CT scan geometric parameter set G, the upper limit of the number of iterations N, and the upper limit of the number of sub-iterations K.

[0061] The initialization module is used to initialize the reconstructed image with estimated acyclic artifacts and the image with estimated vertical stripe artifacts.

[0062] The iterative processing estimation module is used to reconstruct the image using the operator R. G Update the reconstructed image x after the k-th iteration (k) The intermediate image x is obtained. (k+1 / 2) =R G (p,x (k) ).

[0063] The image gradient sparsity constraint module is used to constrain the image gradient sparsity using the operator T, which constrains the image gradient sparsity, for the intermediate image x obtained in step 2. (k+1 / 2)By applying constraints, we obtain the reconstructed image x after the (k+1)th iteration. (k+1) .

[0064] The group sparsity constraint module is used to constrain the gradient sparsity and group sparsity of the vertical stripe artifact image S after the kth iteration by using the operator Q. (k) By applying constraints, we obtain the vertical stripe artifact image S after the (k+1)th iteration. (k+1) , using S (k+1) Update the detector's response inconsistency compensation coefficient S.

[0065] The decision module determines whether the maximum number of iterations N has been reached. If so, the iteration is terminated, and the reconstructed image x is output. (k +1) and vertical stripe artifact image S (k+1) If not, the iterative estimation module will continue to obtain the intermediate image x. (k+1 / 2) .

[0066] In the above embodiments, the l1 norm represents the sum of the absolute values ​​of the parameters, which can prevent overfitting. The l1 norm does not make the parameters equal to 0 but rather close to 0, thus smoothing the parameters. The optimized parameter vector is often relatively sparse. The l2 norm represents the square root of the sum of the squares of the parameters, which is the Euclidean distance. The smaller the l2 norm, the smaller each element of the parameter is, close to 0, and the regularization term is differentiable everywhere. 2,1 The norm represents the l2 norm of each column. For a fringe artifact image, if the horizontal axis represents the column number, the vertical axis represents the l2 norm value of each column of the fringe component. Most values ​​are distributed near 0, indicating locations where fringe artifacts do not exist. Values ​​deviate from 0 for locations where fringe artifacts exist. Treating each column as a group, using l... 2,1 The norm describes the sparsity of the group to characterize this feature of vertical stripe artifact images.

[0067] To better demonstrate the advantages of the dual-domain regularized ring artifact correction method provided by this invention in removing ring artifacts, the following specific embodiment will be used to compare the method described in this invention with the existing typical algorithm SART.

[0068] Data acquisition in this embodiment was performed using an industrial CT system equipped with a Hamamatsu 130kV X-ray source and a Dectris Eiger 2 photon counting detector. This detector consists of 2068 × 512 detector units, each 75µm in size. In this experiment, the energy thresholds were set to 50keV and 60keV. The experimental voltage and current were 100kV and 300µA, respectively, with 720° of scanning angle. Projection data corresponding to the central slice was extracted to simulate fan-beam scanning. The scanned sample was an equivalent water-bone phantom, placed on the scanning stage. The image is shown below. Figure 2 As shown.

[0069] The scan data were reconstructed using SART and the method of this invention, respectively, and the reconstruction results are as follows: Figure 3 , Figure 4 As shown. Wherein: Figure 3 This is the result of SART reconstruction; Figure 4 This is the reconstruction result of the method of the present invention. It can be seen that the SART reconstruction result has severe ring artifacts; while in the result of the present invention, the ring artifact removal effect is obvious, and all the details of the image are clearly reconstructed.

[0070] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Those skilled in the art should understand that modifications can be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for correcting ring artifacts in CT images based on dual-domain regularization, characterized in that, include: Step 1: Obtain the CT scan dataset p, and set the CT scan geometric parameter set G and the upper limit of the number of iterations N; Step 2: Initialize the reconstructed image of the estimated acyclic artifact and the estimated vertical stripe artifact image; Step 3, by the image reconstruction operator R G Update the reconstructed image x after the kth iteration (k) , to obtain an intermediate image x (k+1 / 2) = R G (p, x (k) ), R G is related to the set of scan geometry parameters G; Step 4, using the operator T that constrains the sparsity of the image gradient, on the intermediate image x obtained in step 2 (k+1 / 2) constraining, to obtain the reconstructed image x after the k+1 iteration (k+1) ; Step 5, using the operator Q which constrains the sparsity of the vertical stripe artifact image gradient and the group sparsity, update the vertical stripe artifact image S (k) with the constraint of the sparsity of the vertical stripe artifact image gradient and the group sparsity, to obtain the vertical stripe artifact image S (k+1) , update the detector response inconsistency compensation coefficient S (k+1) ; Step 6: Determine if the maximum number of iterations N has been reached. If so, output the reconstructed image x. (k+1) and vertical stripe artifact image S (k+1) If not, proceed to step 3.

2. The CT image ring artifact correction method based on dual-domain regularization as described in claim 1, characterized in that, Operator T is defined by the optimization problem of the following equation (1): In equation (1), Ax (k) To reconstruct image x (k) The estimated projection data obtained after forward projection, λ1 is a pre-set adjustment parameter used to control the weights of the regularization term and the data fidelity term, D H x、D V x represents the gradients in the horizontal and vertical directions of the reconstructed image, respectively, ||·||2 represents the l2 norm, and ||·||1 represents the l1 norm.

3. The CT image ring artifact correction method based on dual-domain regularization as described in claim 1 or 2, characterized in that, Operator Q is defined by the optimization problem of the following equation (2): In equation (2), Ax (k+1) To reconstruct image x (k+1) The estimated projection data obtained after forward projection, where λ2 is the pre-set adjustment parameter used for the weights of the gradient constraint regularization term and the data fidelity term, and D V S represents the gradient of the vertical stripe artifact image S in the vertical direction, ||·||1 represents the l1 norm, λ3 is a pre-defined adjustment parameter for the weights of the group sparsity constraint regularization term and the data fidelity term, Φ(S) is the set of vertical stripes, ||·|| 2,1 Indicate l 2,1 Norm.

4. The CT image ring artifact correction method based on dual-domain regularization as described in claim 1 or 2, characterized in that, Image reconstruction operator R G It can be obtained using iterative reconstruction algorithms or analytical reconstruction algorithms.

5. A device for correcting ring artifacts in CT images based on dual-domain regularization, characterized in that, include: The data input module is used to input the CT scan dataset p, the CT scan geometric parameter set G, and the upper limit of the number of iterations N; An initialization module is used to initialize the reconstructed image of the estimated acyclic artifact and the estimated vertical stripe artifact image. Iterative processing estimation module for image reconstruction operator R G Update the reconstructed image x after the k-th iteration (k) The intermediate image x is obtained. (k+1 / 2) =R G (p,x (k) ), R G Related to the set of scanning geometry parameters G; The image gradient sparsity constraint module is used to constrain the image gradient sparsity using the operator T, which is used to process the intermediate image x obtained in step 2. (k+1 / 2) By applying constraints, we obtain the reconstructed image x after the (k+1)th iteration. (k+1) ; The group sparsity constraint module is used to apply the operator Q, which constrains the gradient sparsity and group sparsity of the vertical stripe artifact image, to the vertical stripe artifact image S after the kth iteration. (k) By applying constraints, we obtain the vertical stripe artifact image S after the (k+1)th iteration. (k+1) , using S (k+1) Update the detector's response inconsistency compensation coefficient S; The judgment module is used to determine whether the upper limit of the number of iterations N has been reached. If so, the reconstructed image x is output. (k+1) and vertical stripe artifact image S (k+1) If not, the iterative estimation module will continue to obtain the intermediate image x. (k+1 / 2) .

6. The CT image ring artifact correction device based on dual-domain regularization as described in claim 5, characterized in that, Operator T is defined by the optimization problem of the following equation (1): In equation (1), Ax (k) To reconstruct image x (k) The estimated projection data obtained after forward projection, λ1 is a pre-set adjustment parameter used to control the weights of the regularization term and the data fidelity term, D H x、D V x represents the gradients in the horizontal and vertical directions of the reconstructed image, respectively, ||·||2 represents the l2 norm, and ||·||1 represents the l1 norm.

7. The CT image ring artifact correction device based on dual-domain regularization as described in claim 5 or 6, characterized in that, Operator Q is defined by the optimization problem of the following equation (2): In equation (2), Ax (k+1) To reconstruct image x (k+1) The estimated projection data obtained after forward projection, where λ2 is the pre-set adjustment parameter used for the weights of the gradient constraint regularization term and the data fidelity term, and D V S represents the gradient of the vertical stripe artifact image S in the vertical direction, ||·||1 represents the l1 norm, λ3 is a pre-defined adjustment parameter for the weights of the group sparsity constraint regularization term and the data fidelity term, Φ(S) is the set of vertical stripes, ||·|| 2,1 Indicate l 2,1 Norm.

8. The CT image ring artifact correction device based on dual-domain regularization as described in claim 5 or 6, characterized in that, Image reconstruction operator R G It can be obtained using iterative reconstruction algorithms or analytical reconstruction algorithms.

Images