Sparse-view CT reconstruction based on total variation with alternating gradient constraint edge-preserving algorithm

Abstract Sparse-view computed tomography (CT) offers a wide range of options for medical diagnosis. However, conventional CT reconstruction algorithms cannot meet the requirements for medical diagnosis in the presence of incomplete scan data. According to the compressed sensing theory, this problem can be solved by the introduction of regular terms. Using total variation (TV) regularization to reconstruct the CT image can effectively improve the quality of the image reconstruction to some extent, but excessive smoothing occurs at the edges of the image. Therefore, in this study, the problems of noise reduction and edge preservation in sparse-view reconstruction based on TV were investigated. The alternating edge-preserving diffusion and smoothing (AEDS) algorithm achieves the performance of edge detail preservation using high sparsity l 0-norm. Then, we introduce a gradient prior term in the AEDS algorithm to fully utilize the information before and after iteration to better preserve the edge information of the image, named alternating gradient constraint edge-preserving. Based on the above approach, we propose a double-constrained sparse-view CT reconstruction algorithm with coupled denoising and edge preservation performances. And we verify the convergence of the model. Compared to other algorithms, the experimental results show that the proposed algorithm preserves more image details while suppressing noise and improving the quality of the reconstructed image. The algorithm cleverly combines l 0-norm and l 1-norm, to solve the problem of using only one constraint in sparse-view CT reconstruction, which does not balance denoising and edge protection well.