Research On Video Compression Algorithm

The video compression techniques have been widely applied in many fields, such as video monitor, aviation, video conference and so on. H.264 and JPEG as the mainstream algorithms need to sample before the compression, which waste a lot of time. The technique of compressed sensing is emerging recently, it can complete the compression while sampling, which saves the time by avoiding the whole sampling before compression.In this theory, the sampling frequency is only associated with the structure of signal, not be constrained by the frequency of signal, so the signal can be recovered better under the condition of existence of sparse, the compressed sensing is applied to the video compression, which will have a great value of application. The matching pursuit algorithm of compressed sensing in video compression was studied mainly, two improvements were made in working time and the precision of recovery:Firstly, the limitation of the traditional compressive sampling was analyzed, the theory of compressive sensing and algorithms of video compression were researched deeply. Based on various reconstructed algorithms had been deeply investigated, the shortcomings of the traditional algorithm of compressed sensing video compression were analyzed mainly. The original orthogonal matching pursuit consumed plenty of time to generate the index set with only once selection in the atom. In order to improve the efficiency, an algorithm of duplicated selection was proposed with the application of frames’ difference. It reduced the calculated time considerably by sacrificing slight peak signal to noise ratio after the simulating in MATLAB.Secondly, for the time-consuming of the whole frame image, the block algorithm was studied deeply, the time and peak signal to noise ratio of the whole image was researched.Therefore the algorithm of block threshold selection was proposed based on the block algorithm. This algorithm divided frame image into multiple blocks, then respectively process each block with compressive sensing after threshold selection. The simulated results proved that the quality of images and the time were improved by using the algorithm of block threshold.