Supplementary Material to ”Patch Group Based Nonlocal Self-Similarity Prior Learning for Image Denoising”

In this supplementary material, we provide: 1. The closed-form solution of the proposed weighted sparse coding model in the main paper. 2. Comparison of the Patch Prior based Denoising (PPD) method and the proposed Patch Group Prior based Denoising (PGPD) method. 3. More denoising results on the 20 widely used natural images. 4. The denoising results of the competing methods on the Berkeley Segmentation Data Set [1]. 1. Closed-Form Solution of the Weighted Sparse Coding Problem The weighted sparse coding problem in the main paper is: min α ‖y −Dα‖22 + ‖wTα‖1. (1) Since D is an orthonormal matrix, problem (1) is equivalent to mi