Research Of Image Compression Based On Compressed Sensing

Compressed sensing theory is a subversion of the traditional theory It brokethrough the restrictions of the Nyquist sampling theorem, it is possible to randomsampling with less data sampling points (much lower than the Nyquist samplingtheorem minimum quantity required) to accurately restore the original signal. Thebiggest difference compared with the traditional signal compression technology isthat compressed sensing technology compresses the signals,while sampling thedata.In the conventional signal compression techniques, a large number of thesampled data obtained in the compression process were discarded, thus resulting ina waste of resources. Compressed sensing theory put forwar to overcome it.The study of this paper is mainly about compressed sensing theory in imagecompression technical applications. Focuses on the analysis and comparison of theperformance of different sparse transform based compression ratio to restore theoriginal image in different image.At the same time, propose a new imagecompressed sensing reconstruction algorithm, and will be used in the MR imageprocessing. This paper is divided into three parts.First,make a comprehensive exposition of compressed sensing theory. Focuseson compressed sensing theory the basic framework of the three core content-thesparse representation of the signal, the measurement matrix and signalreconstruction, pointed out in this three-part design difficulties. And will make acomparison of image compression algorithm which based on compressed sensingand the traditional image, point out its advantages.Then, introduce multi-scale geometric analysis tool-Contourlet transform.Study compressed sensing algorithm based on Contourlet transform,compare itwith the algorithm based on wavelet transform.The experimental result show thatContourlet transform can better show the edge profile geometric characteristics ofof the image. Original OMP reconstruction algorithm has been improved, so that the image compression recovery achieved better results, various indicators haveimproved.Finally, the perception based on Contourlet Transform compressed sensingalgorithm is applied to the MR image processing.Combined with the characteristicsof the MR images, the traditional Contourlet transform is proposed to improve theSFLCT(Sharp Frequency Localization Contourlet Transform). It is relative to thetraditional Contourlet transform makes the image more sparse, and based on softthreshold iterative method,propose a new iterative thresholding algorithm, it canquickly and accurately restore the original MR images.