Research On Distributed Video Coding Based On Compressed Sensing

Video is a kind of signal which transmits continuously and has huge amount of data. There are a lot of needs of real-time video-transmission in a wireless security monitoring, aerospace image transmission and some other areas. However, the huge amount of video data brings enormous pressure to Storage, computation and the channel. Traditional technology of video compression and transmission is difficult to break through the constraints of distal transfer load devices or the resource-constrained. High efficiency compression, decompression of high-quality, low coding complexity, are the trends of technological development of real-time video transmission.Compressed sensing theory can fully utilize sparse signal characteristics. In the condition of far less than the Nyquist sampling rate, it could use a random sample signal sampled perfectly reconstructing the signal through a nonlinear reconstruction algorithm. This method is effective in saving the cost of coding samples, making it ideal for such large data and high redundancy communication scenarios. Distributed video coding is a new kind of video coding, the inter-frame correlation of the video sequence will use for decoding. At the decoding side, we achieve to using the difference between the compensation or the side information, and the implementation of distributed encoding side only by intra-frame coding, so that greatly reducing the complexity of the encoding side. We used continuous real-time video transmission as scenarios, proposed an decoding algorithm based on double sparse dictionary: Side information(SI) is generated by the adjacent key-frame; Separate sub-bands with different scales of the sample images under wavelet domain, then use K-SVD algorithm for getting different redundant dictionaries with multi-scale properties; At last we combine these dictionaries with Gradient Pursuit for Sparse Reconstruction(GPSR) algorithm to reconstruct the non-key frames. Simulation results show that at the same compression ratio, the PSNR of proposed method obtained more than 0.5 ~ 1.5d B gain than traditional methods.What is more, for the effects of different motion intensity of scenes on the video decoding performance, in this paper we presents a method for assess of motion intensity, and based on the rating results we built video codec system with feedback mechanism and adaptive compression ratio. Without increasing the complexity of encoding conditions, it could effectively solve the problem of low quality in fast motion scenes decoding.