Performance enhancement of H.264 video decoder using NVIDIA CUDA

Abstract

H.264 / AVC is an industry standard for video compression, the process of converting digital video into a format that takes up less capacity when it is stored or transmitted. It provides a good compression ratio with good quality as compared to the previous video compression standards. But it all comes at a higher computational cost. So some part of this standard can be computed on GPU to free the CPU. In this report, we have discussed the H.264 video compression standard in detail and have explored the NVIDIA CUDA for using the GPU for reducing the computational requirements. NVIDIA provides Video decoding API using which we can enhance the e ciency of the decoding process. Here, we have implemented a decoder using this API's functions and compared its execution e ciency in terms of time and its Frame rate with the Joint Model(JM) Reference Soft- ware. JM Reference software is used for academic reference of H.264 and it was developed by JVT (Joint Video Team) of ISO/IEC MPEG and ITU-T VCEG (Video coding experts group).