Research On SAR Images Reducing Speckle Algorithm Based On Sparse Decomposition

Speckle reduction, edge detection, segmentation are the most basic hot contents in SAR image processing for SAR image, and SAR image speckle reduction is one of the most basic and most important issue because SAR image speckle influences its subsequent applications directly. Research of this thesis is SAR image reducing speckle methods based on sparse decomposition, focuses on the images sparse decomposition in the SAR image reducing speckle applications. The contributions of the thesis include:1. Research on images sparse representation theory, methods, and application of images processing based on images sparse representation, focused on the theory and application of over-complete sparse representation based on redundant dictionary. Then by analyzing the different performances of SAR image and speckle in the sparse decomposition, identify the difference between of the image content and image noise in sparse decomposition, putting forward SAR image decomposition and reconstruction methods based on image sparse decomposition.2. SAR image reducing speckle algorithm based on image sparse decomposition. Firstly, SAR image multiplicative convers to additive noise through logarithmic conversion. Then over-complete dictionary adopts adaptive atoms dictionary, and SAR images sparse decomposition is based on OMP (orthogonal matching pursuit) algorithm. Atomic dictionary updates by SVD (singular value decomposition), we use simulated annealing algorithm to quickly find optimal atom at every step of iteration in the OMP process, which will greatly reduce the speed of sparse decomposition. Finally, by image sparse representation achieves SAR image speckle reduction.