Research On Image Restoration Based On Compressed Sensing Theory

The theory of compressed sensing breaks the traditional Shannon-Nyquist framework,which brings a new idea to the field of imaging.When acquiring target information,it only need acquire a small number of observations by means of compressed sampling,and then reconstruct the target image through the sparse reconstruction algorithm and measurement data.Signal sparsity,measurement matrix and reconstruction algorithm are the most important components of compressed sensing theory.This paper studies the impact of common measurement matrices on the compression sampling and reconstruction process through a large number of experiments.Meanwhile,this paper proposes an improved measurement matrix.This paper uses the construction method of Bernoulli matrix intervening the construction process of part Hadamard matrix to obtain the new measurement matrix.In the process of image reconstruction,the new matrix can improve the quality and the stability of the reconstruction effectively.In the part of compressed sensing reconstruction algorithm,this paper deeply studies the greedy algorithms,and gives a detailed improvement idea for the deficiencies of the greedy algorithms in the process of reconstruction.This paper proposes a Modified Sparsity-Adaptive Subspace Pursuit algorithm.Compared with other algorithms,it can be found that this modified algorithm can effectively improve the reconstruction quality and reconstruction stability of the image.Although the theory of compressed sensing can reduce the number of measurements,it also faces the problem of large-scale matrix operation.The paper studies the Block Compressed Censing(BCS)which can reduce the matrix size and improve the computing speed.Block Compressed Censing is troubled with block-effect.This paper proposed some methods to solve this problem.Simple median filter can remove the block-effect,but the picture will be obscure.This paper proposed an improved TV model to wipe out the block-effect,experiment results show that this method is effective and the picture will not be obscure.