Research On Image Coding And Reconstruction Algorithms Based On Compressed Sensing

Compressed Sensing (namely CS) is a newly signal collection and processing method inthe field of signal processing. It can bring us excellent reconstruction qualities with thenumber of signal processing and the Sampling rate is lower than the Nyquist theorem ask.Since the high needing on mass data processing, how to make a break through base on theexisting theory is a difficult problem in front of IT workers. Fortunately, Compressed Sensingput forward by Candes and Tao T has stirred quite an amount of excitement in signalprocessing community. Based on the structure of the signal and the content, The CS theorycompress the signal and then reconstruct the signal from the received side with highprobability, no longer depengds on the bandwidth of the signal.Many applications based on CS about image processing and computer vision have beingexplored. Besides this paper, we have extensively read associative paper at home and abroad.According to the superfluous information between frames, we proposed a kind of adaptivesegment coding algorithms about sequences images based on CS.As for arbitrary image signal, different parts not have the same sparse degress, usinghigh speed sampling rate, the volume of samping data is big and will lead to the waste ofresources. On the other hand, low sampling rate can reduce the number of data acquisition,but the quality of the reconstructed image is very poor, for the reason that the distortion islarge. This paper proposed adaptive segment coding algorithms based on CS theory, themethod first determines the image sparse degress of each block in DWT domain, Accordingto the sparsity of each block to determine image adaptive sampling rate, so as to obtain higherquality of the reconstructed image with lower sampling rate.The simulation rusults show thatthe method can improved PSNR up to33%.When comforted with sequence images, as for the superfluous information between theadjacent frames, this article put forward a kind of adaptive coding algorithm form theperspective of residuals between adjacent frames, the method is especially suitable for slowmovement image and other image sequences with little scene changed. Experiments show thatthe newly method can reduce the amount of sampling data to a great extent and ensureexcellent reconstruction quality at the same time.