Error Concealment For H.264 In Wireless Channels

With the development of wireless communication technologies, wireless multi-media communication has been widely used. The transmission of video signal in wireless channels requires not only high compression efficiency, but also robust error-resilience features. As a new generation video coding standard, H.264 provides better compression ability than earlier version of MPEG-4 and H.263. The adoption of "network-friendly" syntax gives H.264 a good applicability to different network, the coded video data can be transmitted over any existing and future networks. This makes H.264 an attractive candidate for wireless multi-media application.This thesis is organized as follows. Firstly, the key technologies of H.264 are introduced, and a detailed analysis of recently proposed error concealment technologies is given. After that, the simulation of H.264 bitstream transmission in wireless channels is described. The wireless channel is characterized by harsh transmission conditions in terms of signal fading, multipath effect, high-speed terminal, and inherent noise, which result in time- and location-varying channel conditions. Simulation results show that the hash channel conditions result in burst packet loss and the corrupted coding bitstream will impede correct decoding or produce a visible degradation in decoding video. So, the introduction of new technologies that can improve the error resilience of compression video is urgent. Error concealment in decoder is playing an important role in combating this problem. A refined error concealment strategy in decoder is implemented on the test platform JM (Joint Model), aim at error concealment in wireless channels. The concealment of erroneous video is achieved by utilizing the spatial and temporal correlation of reliable neighbor block to interpolate the pixels of drop erroneous block and estimate the motion vectors. This refined strategy shows two important features: first, it provides more reliable temporal and spatial information for error concealment by utilizing the box-in searching method in interlaced video; second, it adjusts the weight of edge pixels adaptively based on the difference of neighboring reliable macro block, in order to protect the edge integrity and eliminate trailing. Performance evaluations show that the proposed approach achieves significant improvement over traditional methods in both PSNR and subjective picture quality. Besides, it also makes a good performance in protecting the edge integrity of object and restraining error propagation. The additional computational costs due to the refined method are small.