Research And Implementation Of CUDA-based H.264 Video Decoding Algorithm

H.264video codec proposed in2003by the two international organizations, ITU-T and ISO/IEC, is the most optimum and mature video coding standard in the video compression field nowadays. The improvement of the performance of H.264codec algorithm is at the expense of the increasement of its complexity. How to improve the efficiency of decoding without affecting the quality of vedio is a common research direction for many people.In recent years, the rapid development of graphics processor unit(GPU) makes it progressively used for general-purpose computing.CUDA (Compute Unified Device Architecture), proposed by NVidia in2007, provides a good hardware and software development environment for general-purpose computing of GPU.The implementation of H.264video decoding algorithm based on CUDA is proposed in the thesis Division of tasks is done to H.264serial decoder. CPU is responsible for the serial implementation of stream analysing, entropy decoding, reordering, memory allocation and data transmission with GPU. And GPU is responsible for the parallel implementations of IQ, IDCT, intra prediction, inter prediction, and loop filter.In this thesis, with full analysis of IQ, IDCT, intra prediction, inter prediction and loop filter modules, efficient parallel algorithms are proposed for each module. A fully parallel algorithm for IQ is proposed. Butterfly parallel algorithm and full parallel algorithm are proposed for IDCT module. A partial parallel prediction algorithm and its optimized algorithm are applied for intra prediction module. An efficient fully parallel algorithm is proposed for inter prediction module.Fully parallel algorithms are applied respectively for the procedure of calculating loop-filter strength and executing the filter.Through adequate experiments, the implementation of H.264video decoding algorithm based on CUDA proposed in the thesis, achieves10times acceleration to tranditional FFmpeg decoder, under the same hardware and software development environment and the conditions of ensuring the quality of image restoration.