A Study Of Gpu-based Real-time Decoding Technique For High-definition Videos

In recent years, with the development of video codec technology, a new generation of video coding standards, like H.264, gradually becomes the mainstream. This video coding standard uses the latest video encoding technology, and obtains well results. However, its increasing of compression efficiency is costs of increasing the complexity of complex algorithms. It is difficult to decode in real-time simply with software, especially for high-definition video. It must be accelerated through the decoding hardware to play high-definition video in real-time.The proposed of GPGPU (general purpose computation on GPU) addresses a feasible method for solving this problem. Compared with the CPU, GPU has greater data transfer bandwidth, more computing units, and more floating-point computing power. GPU suited for those areas with high parallel features, and the high-definition video decoding just fits this characteristic. The architecture, based on GPGPU--CUDA (Compute Unified Device Architecture), is a complete GPGPU solutions. It provides access to hardware interfaces directly. It uses a new GPU computing architecture to use hardware resources in the system architecture. It uses a language , similar with C programming, to provide the capability of high-performance computing, so that the designers can establish a highly efficient data-intensive computing solutions without understanding the graphics processing unit.The author studied the H.264 video decoding standard, and realized MC(motion compensation) and CSC(color space conversion) by GPU to achieve real-time decoding of high definition video. The author simply studied the H.264 video decoding standard and its architecture, and analyzed the decoding complexity of each process. At the same time, this paper had a deep study the CUDA programming model and optimized MC & CSC by GPU . This paper proposed the method of MC by CUDA, and proposed the method of CSC by CUDA & OpenGL shader. At last, the experiment proved that it is feasible to decode high definition video in real time by GPU.