Research On Compressive Sensing Based Image Coding

The continuous increase of image data places higher demands on storage resources and bandwidth resources.Compressive sensing is a new sampling method that can break through the Nyquist sampling theorem,and provides new directions and ideas for data collection and storage,but the reduction in the number of samples does not mean that the storage space is reduced.This article mainly studies image compressive sensing from two perspectives,which are sampling rate and bitrate.Besides,it also discusses the spatial domain adaptive sampling and reconstruction algorithm.Firstly,the image compressive sensing reconstruction algorithm is studied from the perspective of sampling rate,and a joint prior model is proposed.The nuclear norm of the compressive sensing initial degraded image will increase obviously.Based on the sparse representation compressed sensing reconstruction model,the weighted nuclear norm minimization model based on non-local low rank is introduced.A compressed sensing reconstruction algorithm with local sparse and non-local low rank prior is proposed.The joint prior model is divided into multiple sub-problems by alternately updating variables.The experimental results verify the effectiveness of the model.Spatial domain sampling is a classic image sampling method,but the uniform sampling scheme is redundant in the flat area and insufficient in the detail area.Through image texture analysis,a spatial domain adaptive sampling scheme is proposed,which uses iterative nearest-neighbor interpolation and adaptive filtering to reconstruct the sampled image.On this basis,an adaptive sampling reconstruction model based on the joint prior is proposed to further improve the image reconstruction quality.At this stage,the research on compressed sensing focuses on image reconstruction from the perspective of sampling rate.However,from the perspective of data storage,the reduction in sampling dimension does not increase the actual storage efficiency.Therefore,this article studies compressive sensing from the perspective of bitrate.Firstly,the compressed sensing quantization coding is analyzed,and the measurement matrix,quantization scheme and reconstruction algorithm are discussed.On this basis,the compressive sensing quantization coding structure is further designed,and the implicit correlation between the compressive sensing measurements is used to propose a grouped progressive prediction compressed sensing image coding scheme to improve the rate-distortion performance.