Immune Optimization Reconstruction Of Compressive Sensing With Prior Model

Recent years, a new data sampling theory, named compressive sensing(CS) that can achieving compression during the sampling, is appeared in the field of signal processing, CS is a new theory that breaking the constraints of traditional Nyquist Sampling Theorem for data acquisition and making a revolutionary change. There are wide application prospects for CS in the fields of compressed imaging system, analog/information biology, and biology-sensing and so on. There are many issues deserve to study, and compressed sensing reconstruction is an important part of the CS theoretical framework.Currently, the main CS reconstruction algorithms are the minimum l\norm algorithm, series of matching pursuits algorithm, and Bayesian compressive sensing and so on, these reconstruction algorithms pay less attention on the structure of the original image; however our method consider the structure of the original image during the reconstruction, obtaining the measurements of image’s wavelet coefficients via wavelet transform, in view of Intra-scale dependencies between image wavelet coefficients, the immune optimization CS reconstruction framework is proposed with a prior model in this paper. The innovation work of this paper includes:1. The immune optimization reconstruction framework with a prior model in l0-norm is put forward in this paper. The reconstruction framework of CS is of two parts: firstly, using the prior model, the locations of the sparse coefficients can be located; secondly, the values of the sparse coefficients can be computed on those locations of the sparse coefficients.2. By the immune genetic algorithm, the locations of the sparse coefficients are fixed on. A fuzzy image is gained by an inverse wavelet transform of those low frequency coefficients, and detect the edges of fuzzy image, then get the vaccine via the prior knowledge of intra-scale dependencies between image wavelet coefficients inside each sub-band, finally get the prior model. At last, search for the locations of high frequency sub-band wavelets.3. Using the modified clonal selection algorithm to attain the values of high frequency sub-band coefficients that corresponding to those locations of sparse coefficients, ultimately, get the best reconstruction image. The experimental results show that our results have better visual effect and smaller error than the results of some methods.