The Research Of Fast Algorithm Based On H.264 Video Coding

H.264 is the new generation of video coding standard proposed by the Joint Video Team of ITU-T VCEG and ISO/IEC MPEG. Compared with the former video coding standard, it adopts a series of new technologies, its Hierarchical design greatly improves the affinity of the coder to various networks. By using the abovementioned technologies, H.264 satisfies the needs of the real applications which desire high quality of video transferring through different networks meanwhile using bandwidth as low as possible. At the same time, the complexities of coding and decoding the H.264 standard is greatly increased, thus it can not meet the needs of some real-time applications. In order to get a suitable compromise between quality and speed, it is important and meaningful to study the fast algorithms of video coding.The thesis first introduces the basic theories of video coding and the development of the video coding standard studies and analyzes the key technologies of the H.264 standard. Considering the literature review of the current researches, the thesis aims at designing fast algorithms for motion estimation module and inter-frame mode selection module which have the highest coding complexity.Generally, there exist high temporal correlations among video sequences, especially in low/medium rate video coding. During the coding, most of the macro block are relatively static or only moves among a small range. Upon this observation, the thesis first proposes an improved all-zero blocks detection algorithm. By introducing it to UMHexagonS algorithm, the thesis proposes an Improved fast motion estimation algorithm. By setting a threshold, the all-zero blocks could be pre-detected, thus, the motion estimation cost could be greatly saved.Focusing on the mode selection module, the thesis proposes a fast inter-frame mode selection algorithm based on the adaptive threshold. On one hand, the algorithm chooses the best predicted mode for early exit based on the statistical information and the adaptive threshold. On the other hand, according to the temporal and spatial correlation among the macro blocks, the algorithm narrows the prediction range. By setting an adaptive threshold, the algorithm predicts the mode which has the lowest possibility. Thus, the coding speed is greatly enhanced.Finally, the thesis gives out the simulation and experiment results which verify that at the prerequisite of not affecting the bit rate and distortion, the proposed fast algorithms enhance the coding rate and achieve the aim of optimization.