Error Resilience Of H.264 Studying And Application In NVS

With the fast development of telecommunication technology and network technology, video stream service is becoming more and more important in network communication. On the one hand, Internet and wireless network can(?) provide reliable transmission, it is quite often to congest which will lead to packet-loss and error code; on the other hand, because all the mature video compression standards use Variable Length Coding and motion estimation, the video stream becomes very sensitive, even a bit error canN make the whole picture recover. Above all, it has a practical significance on studying the error concealment in video stream.H.264 is the newest international video coding standard which has excellent compressive properties, a friendly interface of network and fault tolerance. Compared to the high coding efficiency of H.264, the algorithmic complexity of encoder has dramatic increased, so compressing and coding algorithm needs to be optimized. Considering the problem of coding efficiency, implementation complexity and current coding criterion compatibility.In this thesis, firstly, basic principles of video code, structure of H.264 codec and key technologies are introduced in this paper. Secondly fault tolerant technique of H.264 has been fully analyzed. Based on these, this thesis proposes an algorithm-sub-sampling based on Multiple Description Coding. JM86 coding and decoding model which provided from ITU-T has tested this algorithm, experimental verification shows that good reconstructed image quality can be obtained even in a sharp packet-loss environment; whatN more, peak signal noise ration of tested image has obviously improved.Finally, this thesis has studied on Network Video Surveillance model. In addition, the proposed sub-sampling technique and H.264 error concealment has been used in this model in order to solve the problems of transmission delay and picture sawtooth. Experimental verification proves that the modified method has an extensive application prospect, which possesses better error-resilient performance.