Implementation of three block mat0063hing search algorithms in h. 264 on CUDA
Over the past two decades, improvisations in computing field indicated by fast networks, distributed systems and
parallel computer architectures delineates that parallelism is the future of computing. H.264 is a video coding standard which
is a bitstream structure and decoding method for video compression. In this paper, we combine the power of two i.e.
GPGPU and H.264. Here, we use NVIDIA CUDA GPGPU to further increase the efficiency of H.264. We use three block
matching search algorithms Full search, Diamond search and Hexagon based search. We implement all of them
independently for the extensive analysis of Motion Estimation (ME). We discuss the components like inter and intra-
prediction models, frame analysis and CUDA implementation which are required to make the system parallel on GPGPU.
Motion estimation is computationally expensive process in video coding. Thus, this paper overcomes the limitation of time
required for motion estimation in CPU by using GPGPU.
Keywords- Compute Unified Device Architecture (CUDA), General Purpose Graphics Processing Unit (GPGPU), H.264
Video coding Standard, Motion Estimation, Block Matching Search Algorithms