Research On Advanced Video Error Concealment Techniques

With the development of communication technologies, video communication becomes a hotspot in human life currently. This technology is widely applied in the domain of digital TV broadcasting, stream video, mobile TV, video conference, remote healthcare, on-line education, etc. Because of the huge capacity of information and real time requirement of transmission, it's difficult to transmit the uncompressed video sequence directly on a band limited channel. In this case, effective video compression method is a necessary part in the video transmission. These years, several video coding standards such as MPEG, H.26x etc. are proposed, normalized and applied. They provide the theory basis for the new service of video communication. As we all know, the compressed video is vulnerable to transmission impairment in a real channel. One bit error may results in a serious corruption in the received frame. In order to achieve the optimized visual quality, proper techniques for error resilience must be introduced.Traditional error resilience method includes: Forward Error Correction (FEC); Error Correction Code(ECC); Automatic Repeat reQuest(ARQ); Error Concealment (EC). This paper focuses on the post-processing Error Concealment method in the decoder.As a specific error pattern in video transmission, frame loss seriously depresses the visual quality in the form of temporal prediction. In this paper, we present an adaptive frame recovery algorithm based on the statistics of motion activity of previous frames. For the large motion scene, the 3 Dimension Recursive Search(3DRS) method is introduced to help recovering the lost frame; otherwise the Motion Vector Copy (MVC) method is applied. The balance between the two methods guarantees both the visual quality and the complexity of processing.Temporal error concealment is one of the most commonly used and efficient method in error resilience. After the analysis of several traditional algorithms, we propose a multi-frame based concealment method which generates the optimized motion vector for the concealment. This method concerns both the temporal correlation between the two consecutive frames and the spatial correlation inside the current frame. Finally its result has a remarkable gain from many traditional algorithms.3D-TV is a newly developed technique in order to enhance the viewing experience. During the transmission of 3D sequence, the error effect by channel disturbance still exists. Here we propose an error concealment method specifically for the 3D video system, which takes advantage of its specific stream structure. The experimental results indicate that this method outperforms other techniques in both PSNR and subject visual quality.