Total variation with modified group sparsity for CT reconstruction under low SNR

Area of Science:

• Medical Imaging
• Image Reconstruction
• Signal Processing

Background:

• Stair artifacts in non-destructive testing (NDT) can impede diagnosis.
• Classical total variation (TV) algorithms struggle with stair artifacts and edge preservation.
• Low signal-to-noise ratio (SNR) poses a significant challenge in image reconstruction.

Purpose of the Study:

• To introduce and evaluate a novel method for addressing stair artifacts and low SNR in image reconstruction.
• To develop a model that can effectively suppress noise while preserving image edges.
• To improve the accuracy of NDT and diagnostic imaging.

Main Methods:

• Proposed a weighted total variation with overlapping group sparsity (GOGS-TV) model.
• Integrated a Gaussian kernel and overlapping group sparsity into the TV model.
• Utilized structured sparsity and Gaussian kernel weighting for image reconstruction.

Main Results:

• Numerical simulations confirmed the superiority of the GOGS-TV model over classical simultaneous algebraic reconstruction technique (SART) and TV algorithms.
• The GOGS-TV model significantly improved SNR in reconstruction images.
• Demonstrated enhanced noise suppression and edge preservation capabilities.

Conclusions:

• The GOGS-TV model effectively reduces stair artifacts, particularly in low SNR conditions.
• The model leverages both gradient sparsity and structured sparsity for improved reconstruction.
• Gaussian kernel weighting adapts to global image distribution, enhancing artifact reduction.