The Video Codec Research Based On Distritured Compressed Sensing

Computer technology and processing capabilities evolved rapidly and continuously in recent years, which blow winds of reform to the Internet, video and broadcasting field. As an important carrier of information, video technology also achieves wonderful advancements. In procedure of video processing, video coding has been a hot research topic all along. In front of the large amounts of data bring by high-resolution video and three-dimensional video, introducing more efficient video compression and coding technology is crucial significant.Compressed sensing theory as a fresh sampling theory shakes the foundations of information processing domain- Nyquist sampling theorem. As we know, obtaining information through traditional methods, if want to restore the original signal without distortion, you can’t but adopt a sampling rate which much higher than twice of the maximum bandwidth. However, the compressed sensing theory can simultaneously reconstruct the original signal without losing the necessary information, by using a sampling frequency far less than traditional methods, acquires the least number of observations to sample signals, and implementing signals’ dimension reduction. Distributed compressed sensing theory union compressed sensing theory and distributed source code theory, take advantage of inter- and intra- correlations between signals, stationing in the field of wireless sensor networks, radar and video image signal processing.This paper focus on video codec with distributed compressed sensing, the main work and innovation is following:Firstly, we studied the compressed sensing reconstruction algorithms and distributed compressed sensing theory, proposed a joint reconstruction algorithm for the first joint sparsity model JSM1, which take full advantage of the correlation between signals.This algorithm decreases the measurement number,not only accelerates the reconstructed but also ensures precise reconstruction of the signal.Then,based video features,combined with joint sparsity theory and distributed video technology, the paper researched distributed video coding technology based on compressed sensing.And, we aim at the characteristics of large amount data and time consuming collection of video encoding, take advantage of distributed compressed sensing theory and propose a simple structure, but efficient video coding framework. In this framework,adopt joint sparse model to reduce the computational complexity. Our algorithm can satisfy fast coding and more accurate image reconstruction.Finally, we studied the sparsity model of image, combined with the joint sparsity characteristics of video, introduce tree sparsity model into distributed compressed sensing,proposed distributed forest sparsity model, which fully exploit the tree sparsity structure and joint tree sparsity structure features in image frame of video,decode video image fast. Experimental results show that the model of the video reconstruction is more stable and simple.In summary, two video coding schemes proposed performance significantly better than independent use of compressed sensing video codec encoding scheme.