The Optimization By Using CUDA And Its Application For Video Encoding Algorithm Of HEVC

With the advancement of technology and development of network, People's demand for high-definition video is increasing, H.264/AVC video encoding standard has not been well positioned to meet the needs of the network as well as people demand. The international organization ITU and MPEGE established JCT-VC(Joint Collaborative Team on Video Coding) specializing in the development of a generation of video coding HEVC(High Efficiency Video Coding), who is also known as H.265, its goal is to improve the efficiency of video encoding, and its compression ratio double improves than the video coding standard H.264/AVC High Profile in the same image quality premise. The performance of the HEVC is greatly improved, however, the performance improvements is increased by the complexity of the H.264/AVC. Due to the high degree of complexity, it brings a lot of encoding time delay and its application has been limited. So it is necessary to make further optimization, improve the encoding speed and meet the needs of the application.In order to effectively reduce processing time delay, we analysis and research the key technology of HEVC video encoding framework, especially the large calculation modules. Finally, we chose the rate distortion function and DCT(Cosine Transform Discrete) transformation to optimize the system. The optimization method is based on the CUDA(Unified Device Architecture Compute) environment, using GPU(Processing Unit Graphic) parallel computing architecture to carry out parallel optimization. CUDA is a development environment using GPU computing, and GPU is a parallel data computing device, on which the resource allocations of the kernel functions are managed. In this paper, we use CUDA to optimize the rate distortion function and DCT transform in HEVC.Experimental results show that the encoding time of rate distortion function and DCT transform can be reduced by using CUDA. The encoding time of the rate distortion optimization can be decreased by 81%, and the encoding time of the DCT transform can be decreased by 78%. The overall encoding time to accelerate the ratio can be reached 1.10(that is, encoding time decreased by 10%) and 1.13(that is, encoding time decreased by 13%) respectively. In the experimental results, it can be seen that the optimized result is better for the high resolution video. For the video sequences with a resolution of 1280 *720, the optimization of the video sequence is better in a degree of 1% than the sequences with a resolution of 416 * 240, because the high resolution video sequence compression encoding will use more large size encoding unit, and for the large size of the structural unit, its CUDA optimization results better.In the last chapter, this paper introduces the application of HEVC, and selects the monitoring system to realize the application of the scene. Due to the large delay of the current HEVC video encoding frame, and video encoding standard interface and framework are very similar, to verify the performance of the system, this thesis use the engineering version of the HEVC, which is x265. In the design of the monitoring system, in order to be able to real-time encoding, we use double buffer structure design and multi threads to complete the system.