Image Compression Based On Compressed Sensing

In this era of information explosion, information transmission and sharing is an important link, the storage and transmission of massive image data has become a serious problem in the field of digital image processing. The traditional image compression technology is limited by the Nyquist sampling theorem, which leads to a relatively large amount of sampling data, cannot solve this problem. Combining compressed sensing theory with sparse representation theory can overcome the limit of the Nyquist sampling theorem, and it can also realize the compression and reconstruction of image when the number of feature data gotten by low sampling rate is small. The research on image compression technology, which combines compressed sensing theory with sparse representation theory, has theoretical value and practical application significance.Based on the study of image compression principle and several common image compression methods, this paper studies on image sparse representation, compressive sensing theory and morphological component analysis theory based on sparse representation and so on. The main work includes the following three aspects:1. Common compression coding methods are expounded respectively, and two classic image compression methods are studied. Moreover, this paper realizes these two classic methods and shows the experiment results.2. Sparse representation and compressive sensing theory are studied, and several kinds of compression method that combine sparse representation with compressed sensing theory are realized, and the experiment results are given.3. A new image compression method, which combines morphological component analysis (MCA) theory with compression sensing theory, is put forward. Firstly, a pair of sparse decomposition dictionaries rendundant discrete wavelet transform (RDWT) and wave atoms transform(WAT)is used to capture image structure and texture to realize the optimal image sparse decomposition. Then, the structure and texture are measured by compressed sensing methods respectively. Finally, the image is reconstructed. The experiment results show that the new method proposed in this paper is able to reconstruct images with high quality. This method has certain practice guidance for the storage and transmission of data at the request of less storage and high quality.