IMAGE FILTERING BY SPARSE RECONSTRUCTION ON AFFINITY NET

摘要

PROBLEM TO BE SOLVED: To provide a method for despeckling synthetic aperture radar (SAR) images.SOLUTION: Multiplicative and additive noise in image pixels is reduced by clustering similar patches 202 of the pixels into clusters. The clusters form nodes in an affinity net composed of nodes 211 and vertices. From each cluster, a dictionary is learned by a sparse combination of corresponding atoms in dictionaries. The patches are aggregated collaboratively using the dictionaries to construct a denoised image.