The Study Of The H.264 Video Encoder And Implementation On TI-DM642

Real-time is a key problem in the applications of video coding technique. As a new generation video coding standard, H.264/AVC can transport high-quality image data at different rate and situation. So that it has aroused more and more people concern in the public. But due to the itself high level of computational complexity, the H.264/AVC standard can't meet the real-time requirements without optimization, in addition also proposed a higher demand to the performance platform. DM642 is a high performance digital signal processor produced by Texas Instrument Company, which with strong parallel processing ability, two levels of buffer and abundant interface make it become an ideal platform to Image Processing research.The main task of this paper is to transplant and optimize the H.264 Encoder based on TMS320DM642. Firstly, we launched concrete discussing about the H.264/AVC video's basic framework, the key features and important technology. Through study the intra-prediction mode selection algorithm, introduced an improved intra-prediction mode selection method. The main measure is predict the macroblock information to determine the macroblock type as soon as possible, then the small probability mode will be excluded from the candidates, so that reduce redundant computation and improve the coding efficiency. Secondly, the paper describes the characteristics and function of DM642 hardware platform and CCS software. Then introduce the process of the transplanting of H.264 video encoder in detail. Finally, By means of the combination of the DSP hardware's features and CCS optimization tools to optimize the source program. It include : delete the source program, use the optimization compile options, establish software pipelining instruction, use inline function to optimize the source program, optimize the structure of cache, use the linear assembly and so on.Experimental results show that the transplant of H.264 encoder base on DM642 can distinctly reduce the computational complexity and speed up encoding rate with little quality loss of the rebuild by the proposed optimization method. The frame rate reach to 25f-30fps from the original 1-2fps, achieved the requirements of real-time encoding.