H.264 Video Decode Based On CUDA

H.264, is a high compression digital video codec standard proposed by a Joint Video Team (be made up of ITU-T Video Coding Experts Group (VCEG) and ISO/IEC Moving Pictures Experts Group (MPEG), it is also the tenth part of the MPEG-4. H.264 has the biggest advantage is its very high data compression ratio. Compared to the MPEG-2 and MPEG-4 standard under the same image quality, the compression ratio H.264 is 2 times higher than the former and 1.5 to 2 times than the latter. This compression technology not only saves much time and data traffic charges for user, but also economical since need less network bandwidth in data transmission.Due to the complexity of H.264 decoding algorithm, it demands high computation power to decode high definition video encode in the H.264 format. The decoding process requires highly parallelized computation power. Mainstream processors nowadays are weak in terms of parallel computing. Even Quad core processor with Hyper Threading enabled can offer only 8 concurrent threads parallel computing capability. General processors simply cannot cope with the compute-extensive operations, foreseeable by considering more and more complex media encode/decode algorithm.In modern computer, graphics processing has become more and more important, so it requires a dedicated graphics processor used, then the graphics processor (Graphic Processing Unit, GPU) comes into being. The development of graphics cards makes the GPU more and more powerful in computation. GPU was originally designed for 3D graphics rendering, with its strong floating-point computation power and features of highly parallelized computation power to optimize image display. Based on the purpose of using graphics cards on other aspects except image computation, graphics card manufacturers NVidia launched CUDA (Compute Unified Device Architecture). Launched by the NVIDIA, CUDA is a general-purpose parallel computing architecture that enables GPU to solve complex computational problems. Under the background of GPU computing becomes a major research focus, CUDA will also have a increasing scope of application and the technology is increasingly maturing, thus more and more popular among people.This article thoroughly takes the advantage of the 2 features of CUDA, ea. large-scale parallel computing and data processing-intensive, to study how to design and implement parallel computation between CUDA and H.264 HD. Motion compensation and IDCT takes over 35% of the whole coding processing time, so this article is devoted to how to use CUDA to speed up video decoding in motion compensation process the article also present the IDCT algorithm based on CUDA. Experimental result shows that CUDA has better decoding performance on decoding various video sequences, since there is great space for H.264 CUDA architecture to speed up video decoding. Through this study we can find that, on the basis of remaining the hardware not upgraded in the PC, using the inherent computing in large-scale parallel characteristics of CUDA architecture, in keeping with the traditional decoding considerable capabilities, can be able to further improve the decoding speed of H.246, and make it have a good applicability and practicality.