Spectral CT Image Reconstruction Based on Similarity Tensor and Hyper-Laplacian with Overlapping Group Sparsity Prior

Ziwen Qi

doi:10.1016/j.amc.2026.130225

Spectral CT Image Reconstruction Based on Similarity Tensor and Hyper-Laplacian with Overlapping Group Sparsity Prior

2026-07-08

0

PRE

AI

Z

Ziwen Qi

H

Huihua Kong *

X

Xiaoshuang Du

J

Jiaotong Wei

J

Jinxiao Pan

DOI：10.1016/j.amc.2026.130225

Original

Original request for help

Save

Abstract

En 中文

• A novel spectral CT reconstruction algorithm HLOGS-TD is proposed, which synergistically integrates non-local similarity tensor decomposition, hyper-Laplacian prior, and overlapping group sparsity to simultaneously exploit inter-channel spectral correlation and image gradient sparsity. • The method effectively suppresses noise and artifacts in narrow energy channels while preserving fine edges, textures, and structural details, outperforming FBP, SART, TV, TD, SISTER, and HLOGS in both numerical simulations and preclinical mouse experiments. • The proposed framework combines tensor decomposition for spatial-spectral correlation modeling and overlapping group sparse optimization for edge-preserving denoising, offering a new mathematical tool for high-quality photon-counting spectral CT reconstruction.

Spectral CT Image Reconstruction Based on Similarity Tensor and Hyper-Laplacian with Overlapping Group Sparsity Prior

Abstract

Journal

Organization

Cited Papers

Citing Papers