WCSE

Efficient Compressed Sensing MR Image Reconstruction Using Anisotropic Overlapping Group Sparsity Total Variation

Abstract— Compressed sensing magnetic resonance imaging has been proved to be an efficient method to reconstruct MR images from highly under-sampled k-space data. Total variation (TV) and wavelet transform are two main sparsity expressions used as prior information in image recovery. In traditional TV, it just considers the sparse characteristic and ignores the group sparsity feature. In this paper, an extension of TV, named anisotropic overlapping sparse total variation (AOGSTV) is applied in the CS-MRI reconstruction process to reduce the staircase artifacts that always exist in TV model. A fast composite splitting algorithm (FCSA) is used to solve the AOGSTV problem. Radial sampling trajectory is used to under-sample the kspace data. Experimental results demonstrate that our proposed method can achieve better quality than the other state-of-the-art methods.

Index Terms— Compressed Sensing, MRI, Anisotropic Overlapping Group Sparsity Total Variation, FCSA

Linna Wu, Hongwei Du
Centers for Biomedical Engineering, University of Science and Technology of China, CHINA
Yingpin Chen
School of Optoelectronic Information, University of Electronic science and Technology of China, CHINA

[Download]

Cite: Linna Wu, Yingpin Chen, Hongwei Du, "Efficient Compressed Sensing MR Image Reconstruction Using Anisotropic Overlapping Group Sparsity Total Variation," Proceedings of 2017 the 7th International Workshop on Computer Science and Engineering, pp. 288-293, Beijing, 25-27 June, 2017.