Research On Embedded Real-time Video Capture System And Video Encoder Optimization

With the development of mobile computing,the application of video coding used in embedded system is increasing sharply. Among the variety of video coding technologies, H.264/AVC raised by JVT (Joint Video Team) became the standard in the industry circle for its good efficiency. But the high compress rate means more complex computation, so it's difficult to use H.264 on embedded system for its limitation of hardware computation ability. In this thesis, on the base of detailed analysis of H.264 coding theory, we study UMHexagonS algorithm, the motion estimation algorithm with a high degree of complexity, and optimize the encoder for EP9302 embedded device platform, then implement the real-time system of video capture and transmission using H.264 encoder.Firstly, under familiar the theory of video encoding premise, search algorithm of the motion estimation and their performance were introduced, the thought and steps of UMHexagonS algorithm were also described in detail,moreover, aimed at its shortages, an improved UMHexagonS algorithm was proposed.It used cross-shaped search strategy in spiral search step instead of full search strategy and add model-based threshold judgment to prediction stage, hexagon search stage and diamond search stage, and also, solve the problem that repeated calculation of macro-block 8×8 in JM 10.2 model.The test results show that improved UMHexagonS algorithm greatly improved search speed with the same performance of the original algorithm.Secondly, we built an embedded EP9302 processor platform and discussed H.264 encoder optimization. Based on the analysis of internal structure of the ARM processor and H.264 encoding module, effective optimization is made at levels of code and instructions, which improve the whole system performance.Finally, the improved and optimizational encoder was applied to embedded system platform, and the video capture and transmission system was designed and implemented. The experimental results show optimizational H.264 encoder could accomplish the basic real-time video encoding for collection and transmission video.