Research On Image Compressed Sensing Based On Wavelet Transform

Image compression technology has gone through60years, emerged a lot of out-standing standards. And one of the most successful is image compression based onwavelet-transform. But all traditional image compression algorithms follow Nyquisttheorem, so all studies are under this premise, which restricts the development ofimage compression technology. Compressed Sensing is a new emerged hot theory inthe signal processing area. It breaks through the limitations of Nyqiust theorem. Itpoints out that, as long as the signal is sparse or compressible, the signal can below-speed sampled and at the same time data is compressed. The inherent simplicityand robustness for image representation make compressed sensing theory very suita-ble for the data compression. This paper mainly focuses on image compressed sensingbased on wavelet transform. The main works are as follows:(1) This paper introduces the wavelet transform and compressed sensing theory，describes traditional image compression technologies based on wavelet transform andsummarizes the shortcomings of such algorithms. Finally, we arrange and classifyexisting image compressed sensing.(2) In order to reduce the memory resource and computational complexity neededby performing image compressed sensing algorithm, a new image compressed sensingbased on joint reconstruction of high-frequency wavelet coefficients is proposed. Onthe basis of multi-scale compressed sensing algorithm, this algorithm samples thewavelet high-frequency coefficients of three directions separately exploiting com-pressed sensing. We propose using three-layer Bayesian Theory to exploit the corre-lation among three directions when reconstructing image. Experiments show that ourproposed method of image compressed sensing can not only reduce the requiredmemory and computational complexity, but also improve the quality of reconstructedimage.(3) Aiming at the problem of considerable required memory and computationalcomplexity needed by big image, we design a new block image compressed sensingbased on wavelet packet decomposition. This algorithm adopts a new image blockapproach-image block method based on wavelet packet decomposition which is verydifferent from the image traditional block method. In addition, through the analysis ofthe characteristics of the amount of information carried by the wavelet high-frequency coefficients of three directions, a new method of allocation samples is put forward.Experiments indicate that the proposed image compressed sensing method can be usedto compress large images. It can not only to eliminate the block effect brought by thetraditional image block method, but also improve the reconstruction quality of theimage.