Research On Compressed Sensing Algorithm Based On Image Multiscale Geometric Analysis

We obtain external information about sixty percent comes from vision,image contains a lot of information we need.With the rapid developmenting of information technology,image has become one of the most important means of information transmission.The amount of information increasing in all aspects of the massive image information storage,transmission and processing has been an important issue for us.How to effectively solve the problems caused by the explosion of digital image information is the key to the acquisition and compression of images.The compressed sensing theory adopts the measurement matrix satisfying the equidistant constraint condition to compress and measure the sparse signal,and then accurately reconstruct the original signal by solving the optimization problem.Compressed sensing can compress the image with a far lower sampling frequency than the existing sampling frequency,and effectively compress the image.The research of compressed sensing theory mainly focuses on three core problems,such as measurement matrix,sparse representation and reconstruction algorithm.In this paper,compressed sensing is introduced and studied from the aspects of measurement matrix,sparse representation and reconstruction algorithm.The current compressed sensing algorithms are compared and analyzed.In order to improve the reconstruction quality of image compressed sensing,some new improvement methods are proposed.The structure measurement matrix in compressed sensing has a very important influence on the performance of signal acquisition and reconstruction.Firstly,the Gauss measurement matrix is optimized,and a new method of measurement matrix optimization is proposed.The orthogonal equalization of Gauss random matrix is applied to improve the orthogonality and column irrelevance of Gauss random measurement matrix,while ensuring that the measurement matrix can satisfy the constraint equidistance condition.Taking the optimized matrix as the measurement matrix,the K-svd training dictionary is used as the sparse basis and the OMP algorithm is applied to the image compression sensing experiment,which verifies the effectiveness of the matrix optimization method.Aiming at the minimum total variation method of image compressed sensing algorithm with low sampling rate lack of texture image reconstruction problems,starting from the angle of image multiscale geometric analysis,advantages to reconstruct the image texture features and effective use of wave atoms transform proposed multi-scale improved total variation method of compressed sensing algorithm.Then the group sparse compressed sensing algorithm in the low sampling rate of the reconstructed image texture of chaotic shortcomings,on the basis of the image wave atom transform coefficients of the optimal characteristics of a suppression matrix and proposes a kind of optimization,finally after the group sparse compressed sensing algorithm.The algorithm is verified by simulation experiments which results show the improved algorithms are better than the original algorithm in terms of reconstruction quality.