Low-Rank Optimization for Image Denoising via Sparse Dictionary Representations

Denoise images by applying sparse coding to local patches using pre-trained overcomplete dictionaries, followed by patch aggregation through averaging. This technique employs sparse and redundant representations over learned dictionaries, as detailed in "Image Denoising Via Sparse and Redundant Representations over Learned Dictionaries." The algorithm involves dictionary initialization, patch extraction, L1-norm optimization for sparse coding (e.g., via Orthogonal Matching Pursuit), and weighted averaging to reconstruct the denoised image.