Compressed Sensing Image Reconstruction Algorithm Based On The Joint Statistical And Structural Priors

With rapid development of information technology,image as the most intuitive information carrier has become the mainstream form of data transmission.Due to the increasing demand for image quality,the traditional image compression and transmission technology has been difficult to meet the growing demand for data bandwidth.Compressed Sensing(CS)theory comes into being which breaks through the sampling rate of the traditional signal algorithm and has made great achievements in the field of image compression without the constraint of Nyquist sampling law.The image compression and reconstruction algorithms based on the compression sensing theory can reconstruct the original signal accurately(for the sparse signal)or approximately(for the compressible signal)from few sample measurements,so as to effectively save the resource cost of coding and compression and have wide application prospect in image data compression and transmission.Images with the sparsity or compressibility provide the premise guarantee for the application of CS theory.Traditional CS image reconstruction algorithms only use the sparse or compressible properties of images in wavelet transform domain,but do not fully consider the full use of statistical properties and structural characteristics.The wavelet sparse representation of the image has sparseness,but also has a strong cluster property,which appears the sparse representation of the tree structure between the layers and the statistically dependent distribution relationship between the wavelets coefficients.In this paper,we have proposed a CS image reconstruction algorithm based on the joint statistical and structural priors which can achieve efficient image reconstruction via a small amount of measurements.With the full use of inter-scale and intra-scale relations of the sparse coefficients,we optimize the iterative hard thresholding image CS algorithm specifically by building a Gaussian Scale Mixture model to constrain the local coefficient distribution statistically,and a tree model to constrain the global coefficients distribution structurally.Extensive simulations have been conducted and the results have shown that the proposed method has achieved a considerable promotion both on the speed and the PSNR gain,under the same compression ratio,compared with the traditional recovery algorithms.