The prior knowledge of compressed sensing is that signal is sparse or compressible. Compressed sensing can acquire the compressed form of sensing signal directly. Namely, the sampling and the compression of signal can be realized at the same time by the theory of compressed sensing. Therefor, the theory of compressed sensing breaks the Nyquist sampling theorem which require that sampling rate must depend on the maximum frequency of the signal. The sparsity of the signal is the premise of compressed sensing. The key of compressed sensing is measurement matrix which project high-dimensional signal to low dimensional space. We can obtain enough information from the measurement values to reconstruct the signal. The core of compressed sensing is the reconstruction algorithm which restore the origin signal by a small number of measurement values. So, it is significance to research the measurement matrix and reconstruction algorithm. In this paper, we focus on the research of key of measurement matrix and reconstruction algorithm. The purpose of this paper is to design efficient optimization method of measurement matrix and stable reconstruction algorithm,.this article main research contents are as follows:(1) A kind of measurement matrix optimization method is proposed based on the nature of the singular value and condition number of measurement matrix. The method reduce the singular value of random matrix to decrease the condition number of the observation matrix, thus it minish the mutual coherence between optimized measurement matrix and sparsity basis. Theoretical analysis and experimental results show that this method is simple, and it improve the quality of the reconstructed signal greatly.(2) The principle of selecting and removing atoms of CoSaMP is different in each iteration which lead to the estimation of the support set is inaccurate. This paper proposed a compressed sampling hard threshold reconstruction algorithm. It had the advantage of both algorithms and incorporated HTP in removing atoms to make the consistent principle, so its theoretical make a certain of selecting atoms more accurate in each iteration. Experimental results demonstrate that the algorithm in terms of accuracy of the reconstruction are superior to the above two algorithms. Compared with the other algorithms, it has the feature of high reconstruction accuracy, noise immunity and stability.(3) Block-based compressed sensing with variable sampling rate distributed the sampling rate for blocks according to the correlations between neighboring frames at encoding side. This approach increased the coding complexity of coding terminal undoubtedly. The search window size is fixed when the compressed sensing recovery of video sequences driven by multihypothesis predictions is considered. There is,however,no analysis of the relationship between inter-frame correlation and the search window size. Aiming at the above problems, it classifies blocks into different types depending on their inter-frame correlation,and adjusts the sampling rate at receiving terminal accordingly. It determines the size of search window according to the degree of regional change. The experimental results show that the proposed reconstruction algorithm can further reduce the complexity of collection terminal and transmission burden. Not only reduce the computational complexity and improve the quality of reconstruction.Researches in this paper provide new idea for the reconstruction of image and video which apply the compressed sensing theory, and achievements have an significant meaning and scientific value. |