Research And Design Of Video Decoder Architecture And Video Error Concealment Algorithm

Digital video has gained more and more applications with the development of multimedia communication technology. Video application platforms have been extended from desktop PCs to set-up boxes, smart-phones and other consumer electronic terminals. Due to the limitation of cost, size and power, ASICs are usually used to accelerate video codec process in these embedded systems.The current video coding standards, such as MPEG-2, MPEG-4 and H.264/AVC, are all block-based coding schemes. DCT, motion estimation and entropy coding are used to compress the video data. As a result, the compressed bit streams are easily interfered by channel noise. Entropy coding leads to the spread of random bit errors in spatial areas. Motion estimation leads to the spread of errors in time domain. So in practical applications, especially in error-prone channels such as Internet and mobile media, error resilience and error concealment have become research focus of video fields.This paper studies the architecture of video decoder, the algorithms and implementation of video error concealment, explores deblocking filter algorithms. The main work and contributions are as follows:1. A unified multi-standard video decoding SoC architecture was proposed based on the analysis of system functions, bus architecture and storage architecture. This architecture supports most of the popular video coding standards, such as MPEG-1/2, MPEG-4, H.264/AVC and AVS. A parallel decoding scheme of CABAC in H.264/AVC was proposed to speed up the decoding procedure while a high-speed multi-standard IDCT architecture was also proposed.2. Spatial error concealment algorithms based on encoding information were proposed. Existing error concealment approaches such as bilinear interpolation and directional interpolation only works for specific image features. In the video coding standards such as MPEG-2 and MPEG-4, the DCT coefficients reflect the image texture, while in H.264/AVC and AVS, the intra prediction direction and residual data reflect the image texture. The proposed algorithms fully leveraged the encoding information and classified error blocks into three categories, uniform block, edge block and high detail block. According to the block type, best algorithms were adaptively selected to recover different damaged blocks. This achieved a very good effect of error concealment. 3. An adaptive temporal error concealment algorithm was proposed to reduce the inter frame error. In this algorithm, weighted mixture boundary matching algorithm and motion searching based on adaptive block sizes were used to increase the precision of motion vector estimation, which achieved good error concealment effect. In low bit rate video communication, a channel error may cause the loss of a whole frame. An error concealment algorithm based on motion vector estimation for whole frame loss was proposed to improve this. Lost motion vectors were estimated by the hypothesis of linear movement. Then, the reliability of estimated motion vectors were evaluated by the motion vectors of reference blocks. If reliable, the lost blocks were reconstructed by the estimated motion vectors, otherwise they were reconstructed by boundary matching method. If intra macroblocks existed in reference frames, the motion vectors were estimated by neighbor motion vectors. A deblocking filter algorithm was proposed to eliminate the block artifacts caused by temporal error concealment. The algorithm eliminated the block artifacts and protected the image edges. As a result, good subjective performance and high PNSR were achieved.