Research On Key Techniques For Distributed Video Coding

As the rapid development of the network technology, the wireless technology and the computer technology, many multimedia applications have emerged in recent years, such as the wireless video sensor network monitoring, mobile camera-phones and portable video camera and so on. As they are severely constrained in the computing capacity, memory and power, a simpler encoder is needed to conserve resources in the aforementioned multimedia scenarios. The traditional video coding schemes can not meet the requirements because of their high coding comlexity. Distributed Video Coding (DVC) is different from the traditional video coding. The time-consuming motion estimation/motion compensation procedure is shifted from the encoder to the decoder in the DVC. DVC has the characteristic of simple encoder, complex decoder and error resilience performance, which can provide excellent solutions for the applications above. This dissertation concerns with the theory and the key technology of DVC. The specific work and innovations of this dissertation can be summarized as follows:1. The generation technology of side information, whose quality affects the rate-distortion performance of the system, is one of key techniques of DVC system. A generation method of side information based on Kalman filter is discussed in this dissertation. The Kalman filter is used to refine motion vector in the motion estimation process, and improve the accuracy of the motion vector. Then Kalman filter is combined with forward motion estimation, bi-directional motion estimation and spatial smoothing to generate the side information. The simulation results show that the performance of side information becomes better than that without using the Kalman filter. At the same time, when the adaptive Kalman filter is applied to the motion estimation, the performance of side information can be further improved. 2. Wavelet transform has the properties of local time-frequency analysis and multiresolution decomposition, and can overcome the blocking effect caused by discrete cosine transform. Therefore, the DVC scheme wavelet-domain is studied in this dissertation, and an improved DVC scheme is proposed. The Gray code is used to encode coefficient after wavelet transform, and it can reduce the correlation noise errors between the WZ frame and side information. Meanwhile, more effective virtual channel model and generation technology of side information are adopted. A decoded WZ frame contains the information of current frame, so it is used to refine initial side information. The fusion technology is adopted to generate a new side information which can assist LDPC decoder to decode re-reconstruction WZ frame. The new side information can significantly improve the quality of the video reconstruction. The simulation results show that the performance of the proposed scheme is better than that of the traditional DVC.3. Now, the main work of the study on improving the performance of DVC system is how to improve the coding efficiency of WZ frame. It can improve the performance of the system that WZ frame is encoded to use the many kinds of coding models. Firstly the feasibility of classifying WZ frame to encode independently is theoretically analyzed, and then a DVC scheme based on the chessboard pattern is proposed.At the encoder, each WZ frame is split into two sets according to chessboard pattern and then encoded separately. At the decoder side, the3D recursive search motion estimation algorithm is adopted to generate the initial side information, and then first part of WZ frame is reconstructed. Secondly, spatio-temporal boundary matching algorithm is used in motion compensation refinement for the side information corresponding to the second part of WZ frame. The simulation results show that the proposed scheme has better rate-distortion performance than that of the traditional DVC system.4. As an important tool for strengthening the public security and privacy protection, video monitoring is widely used. For the scenarios of video monitoring, combined with the ideas of distributed source coding, an effective DVC scheme with low delay for the scenarios of video monitoring is proposed in the dissertation. In fact, the video image with violent movement will not appear frequently in the video monitoring system. At the encoder, a coding mode decision method of WZ frame based on temporal correlation and spatial correlation is adopted, and SKIP mode is used for reducing the bit rate. At the decoder, Side information extrapolation technique based on Lucas-Kanade algorithm is adopted to achieve sequential encoding of system. While its computing complexity is higher, motion vector estimation is more accurate. The simulation results show that the proposed scheme can significantly improve the rate-distortion performance of the system compared with the traditional scheme based on extrapolation.