Group-Based Sparse Representation Based on

Sparse coding has been applied in various domains, especially in image restoration. Most methods depend on the 1  -norm optimization techniques and the patch-based sparse representation models, but they suffer from two limits: one is the high computational complexity in dictionary learning; the other is ignoring the relationship among patches which influences the accuracy of sparse coding coefficients. In this paper, we choose the group-based sparse representation models to simple calculation process and realize the nonlocal self-similarity of images by graphbased transform dictionary design. Besides, we utilize p  norm minimization to solve nonconvex optimization problems on the basis of the weighted schatten p-norm minimization, which can make the optimization model more flexible. Through reasonable parameters selection, experimental results show that our proposed method has a better performance on the compressive sensing than many current state-of-the-art schemes in both peak signal-to-noise ratio and visual perception.