Image Restoration using Bregman Iteration

The digital image processing refers to processing of two dimensional images by digital computer. The image restoration is the reconstruction process applied to the degraded images. The removal of noise from images can be done by filters. New techniques are involved in the field of restoration other than filters. Nonlocal Image representation has shown great potential in various low-level vision tasks. The usage of patches introduced many ideas of restoration process. The spatially adaptive iterative single value thresholding algorithm provides better results by using patches. The dictionary used by the algorithm is based on discrete cosine transform or principle component analysis. The patch based restoration suffers from two problems called computational complexity of dictionary learning and inaccurate sparse coding coefficients due to ignorance of relationship among patches. The above two problems are avoided by group based sparse representation of images. The group sparsity relaxes the complexity by having similar patches in the group. The dictionary used is self-adaptive exploiting selfsimilarity of images. The l0 minimization problem occurred during patch matching will be overcome by Split Bregman