6 March 2026 in IEEE Signal Processing Letters
To address the issues of edge blurring and texture loss in image compressed sensing (CS) reconstruction, this letter proposes an edge-preserving reconstruction method (EP-CS). Building on the observation that reconstruction errors of the L1-TV model mainly occur in edge regions, EP-CS incorporates an edge detection mechanism to locate key image structures and imposes stronger sparsity constraints on the corresponding wavelet coefficients, thereby embedding edge structural prior into the reconstruction model. The proposed model achieves simultaneous edge enhancement and artifact suppression while mitigating the under-estimation of large-magnitude coefficients. Meanwhile it remains globally convex, an efficient solver can be derived within the framework of the alternating direction method of multipliers (ADMM). Experimental results show that the proposed method outperforms state-of-the-art baselines in both visual quality and objective metrics, significantly reducing artifacts and structural blurring while improving detail fidelity.