Image Denoising Via Sparse and Redundant Representations Over Learned Dictionaries

Description

Elad, Michael, and Michal Aharon. “Image denoising via sparse and redundant representations over learned dictionaries.” IEEE Transactions on Image processing 15.12 (2006): 3736-3745.

This work has presented a simple method for image denoising, leading to state-of-the-art performance, equivalent to and sometimes surpassing recently published leading alternatives. 

The proposed method is based on local operations and involves sparse decompositions of each image block under one fixed over-complete dictionary, and a simple average calculations. The content of the dictionary is of prime importance for the denoising process—we have shown that a dictionary trained for natural real images, as well as an adaptive dictionary trained on patches of the noisy image itself, both perform very well.

Image Denoising via Dictionary Learning and Structural Clustering