基于cuda的h.264视频并行编解码器研究与实现信号与信息处理专业论文.docxVIP

I II Abstract Compared with previous video coding standard, H.264 standards are more advanced and practical, but in fact it was released many years ago, still there is a great distance from the realization of truly large-scale application, the fundamental reason is that the high-performance of H.264 is at the cost of improving the computational complexity of the codec. It is the hot-pots of research in the field that how to use the current hardware and software resources to improve the efficiency of the H.264 codec on the condition of high performance. This article first analyzes the challenges that the video coding faces in the application area, also have conducted some discussions about domestic and overseas researches, and then expounded framework of the CUDA programming, and describes the architecture of the H.264 codec, the main features, as well as the key technology of the H .264, and also have analyzed the advantages and challenges of the CUDA’s applying to H.264 video codec. On this basis, this article have made parallel design of the H.264 codecs framework, and also proposed the realization according to the most time-consuming but suitable for parallel execution of the motion estimation and the loop filter module implementation of the GPU. At the same time, the algorithm was also made by the CPU-based algorithm design for the entropy decoding look-up table decoding. During the process of coding frame design, considering the time of data transfer between the GPU and CPU should be significantly less than the computing time needed by the GPU, it is proper to mapped to the GPU. In order to reduce the amount of returned data from the Device side to the Host side, at the same time to avoid data drift phenomena, adopt the motion information to replace the residual signal returns in the article. Through the analysis of the parallel model of motion estimation, not only use variable sub-block tree in three steps to complete the forecast of the entire pixel in the ar