The Research Of Image Reconstruction Algorithm Based On Compressive Sensing

In traditional image sampling and compression model,the signal is first sampled at Nyquist rate, then a large number of redundant data is discard via complex compressive algorithm. This method not only causes the huge waste of sampling resource, but also takes a great challenge to the system processing power and hardware.Compressive sensing theory implements sampling and compressing simultaneously, which greatly reduces the sampling resource. Reconstruction algorithm is the major issue of compressive sensing, this thesis focuses on the research of image compressive sensing reconstruction algorithm, how to use the prior information and the observation data obtained by sampling to reconstruct the signal is the research emphasis. The major work and research results of this thesis are as follows:1. By deeply research, it is found that threshold parameter used in two-dimensional projected gradient(2DPG) algorithm has a great effect on the reconstruction performance. Based on the texture features of natural images, this thesis proposes an adaptive two-dimensional projected gradient algorithm(A-2DPG) combining sampling rate with total-variation of wavelet coefficients. Experimental results show that comparing with 2DPG, the proposed A-2DPG provides superior performance on both the image reconstruction quality and visual effect.2. The reconstruction process of two-dimensional projected gradient(2DPG) algorithm includes iterative convex projection and bivariate shrinkage, iterative convex projection operation guarantees that the signal always satisfies the basic conditions of reconstruction process, bivariate shrinkage is used to make signal sparse and be filtered. As the iterative process goes, the filtering performance of bivariate shrinkage suffer certain restriction, signal reconstruction performance tends to keep steady. Based on Gaussian process regression at wavelet domain, this thesis proposes a post-processing method for compressive sensing reconstruction, in order to achieve the purpose of more accurate reconstruction signal.