Image Compression And Reconstruction Based On Sparse And Redundant Representations

Digital image compression has been a widespread concern in the field of imageprocessing, the core of the problem is if the representation of image data could be moresparsely. A good method of image representation has become the focus of theresearchers.The over-complete Sparse representation of digital signals is a new digital signalrepresentation theory that is developed and became mature in recent years, theresearchers used this theory to various aspects of digital image processing, andsuccessfully used for digital image compression. The research in this paper is based onsparse representation compression.The sparse representation of the image can be divided into two processes: Sparsedecomposition and dictionary learning: sparse decomposition is a process to get therepresent coefficient with a known over-complete dictionary. The process of dictionarylearning which is adverse to sparse decomposition update the over-complete dictionaryby the represent coefficient. The image sparse decomposition results will be more inline with the image features if these two processes can combine effectively, therebyimproving the quality of the sparse representation of the image. Based on the content ofthe two processes, the paper analyzes a variety of traditional sparse decomposition anddictionary learning algorithm, then introduces and analyzes the core idea of thesealgorithms and the performance difference.There are form a compression algorithms bycombine OMP algorithms with K-SVD dictionary. Then the algorithm is comparedwith some traditional image compression algorithm. The experimental results showthat: this method is more effective for digital image compression in the highcompression ratio.In addition, in order to improve the computational efficiency of image sparserepresentation compression algorithm, this paper presents a new image compressionalgorithm. Based on the K-SVD algorithm, a new orthogonal basis union dictionary isused to form the over-complete dictionary, and the BCR algorithm on the basis of thenew dictionary is used to replace OMP algorithms for sparse decomposition of images.Finally, this new algorithm is compared with adaptive K-SVD image compressionalgorithm. The experimental results show that: Although the new algorithm and theK-SVD algorithm is compared to the compressed image quality showed no advantages, but the new algorithm still has good compression at high compression rates, and certainimprovements on the optimization of the algorithm time.