Research On Based On DCAST Coding Performance Improvements

With the development of science and technology, wireless video is becomingincreasingly important, people not only want to watch the HD video, but also playall kinds of3D games. In such case, one would like to save bandwidth bybroadcasting a single video stream to multiple users, the traditional broadcasttechnology requests the source to choose one of the best bit rate to code, but in theface of multiple users with different channel quality, source how to choose the bestbit rate is a problem.Then, people propose a scalable video coding technology, this encodingtechnology will encode the video stream into the base layer and enhancement layer,according to their own channel quality,the user can choose the number of layers to decode,however, the scalable video coding technology may produce cliff effect. With theresearch of the wireless video broadcast technology, one has been proposed a newkind of multicast programs; these programs use a soft compression and softtransmission technology. SoftCast and DCAST is the kind of broadcast programs.The basic idea of SoftCast is to ensure the transmission distances betweencodewords are related to differences between pixel values, the operations ofSoftCast are linear, it can keep the pixel values digital characteristics. SoftCastproposed a new codec, it consists of the video compression and error protection andit uses DCT transform to remove redundancy, however, it does not use the typicalinter frame prediction coding technique to remove temporal redundancy, so itscompression efficiency is not high. DCAST is based on distributed source codingand applies the soft compression soft transmission technology. DCAST applies thedistributed source coding to avoid error propagation problem caused by using interframe prediction coding technology, Hence it has a better coding efficiency.However, it uses a fixed block size in the motion estimation, and the producedmotion vectors must be transmitted to the decoding side, this will increase thebandwidth consumption. In this paper, According to these two shortcomings ofDCAST, we develop a revised proposal to improve the coding performance. Themain content and novelties are listed below：1. In DCAST, the fixed block size motion estimation Modified to variable blocksize motion estimationWe introduce split mode of H.264coding standard to DCAST. According tocharacteristics of the image, our solution can choose the appropriate block size. Forthe flat area, if it chooses the big partition, it can reduce the produced motion vectornumbers, for the more detail area, if it chooses a smaller partition, it can produce a more accurate motion vector. However, in the encoding side, the choice of the moderequires a large amount of calculation, so in order not to increase too much thecomplexity, in the paper, we use a fast mode decision method. The method is basedon the motion homogeneity.2. We proposed a hybrid digital-analog coding scheme to save the transmissionbandwidth of motion vectorThis paper proposes a new coding method, which combines the features ofdigital and analog coding. Encoder has two components: the digital encoder andanalog encoder. Decoder also has two components: the digital decoder and analogdecoder. We use digital encoder to code the motion vectors which are transmittedand use the analog encoder to code the DCT transform coefficients of image, andfinally the modulated digital signal and a modulated analog signal is superimposedto transmit. This coding method can save bandwidth while has the betterperformance.