Research On Efficient Image Compression Algorithm Based On Compressed Sensing Theory

Compressive sensing(Compression sampling,Sparse sampling,Compression sensing)is a new signal processing theory proposed by E.J.Candes,J.Romberg,T.Tao and D.L.Donoho in 2004.It indicates that even if the sampling rate of the signal doesn't satisfy the Nyquist sampling theorem,the original signal can still be reconstructed from a small number of samples through the optimization equation as long as the signal is sparse.In the theory of compressive sensing,the sampling and computing costs of the sensor can be greatly reduced if the sampling and compression of signals are carried out simultaneously.It also points out that as long as a suitable signal representation space is found,all signals can be sparsely represented and they can be compressed and reconstruction effectively.In this thesis,we study the basic framework of compressed sensing theory firstly.According to the characteristics of images,we study efficient image compression algorithm.We also combine existing image downsampling algorithms and image interpolation algorithms to explore efficient image compression algorithms.The main contributions of this article are as follows:1.We study the basic principle and algorithms of traditional compressed sensing theory,and analyze the key steps in compressed sensing theory,such as image signal sparse design,image signal observation design,image signal reconstruction design.In addition,we propose an improved spatial compression-based image compression sensing algorithm and a visual perception-based adaptive image compression sensing algorithm to improve the classical algorithm.A large number of experiments have been conducted to verify that the image compression algorithm proposed in this thesis is highly efficient and feasible.2.This thesis proposes an image compression sensing framework based on spatial downsampling,and analyze the influence of downsampling algorithms and interpolation algorithms on the improved algorithm.Moreover,the image segmentation algorithm is applied to the image compression based on spatially downsampling algorithm,and an image compressive sensing algorithm guided by the minimum mean square error criterion is proposed to achieve high compression sensing reconstruction of the image.3.This thesis proposes a visual perception-based image compression sensing sampling to learn various image block classification criteria.We also propose a feedback-based adaptive image compression sensing algorithm based on visual perception and a step-wise adaptive image compression sensing algorithm based on visual perception.In this thesis,the minimum mean square error,structural similarity and minimum observable difference are used to guide the redistribution of the sample.Finally,all experiments prove that the proposed visual perception-based image compression is efficient and feasible in different image block classification criteria and different sampling rates.