Design And Implementation Of Parallel Algorithms For Some H.264Encoding Modules Based On OpenCL

Video, which is the most abundant information carrier, has become an indispensable part of the information age. Because video usually has huge amount of information, how to compress video data effectively has been a hot research project all the time. As a popular video coding standard, H.264has a good compression performance, but its computational complexity is also very high.Due to its strong floating computational capability, GPU has been gradually applied to the general computing in recent years. OpenCL, which has been put forward in2008, has made it possible for researchers to develop portable parallel programs on GPU and other heterogeneous hardware platforms. Therefore, there is a wide prospect in applying GPU on the video compression field.In this paper, we focus on some important H.264encoding modules, which have high computation complexity. And in view of the parallel characteristics of OpenCL and GPU, we design some efficient parallel algorithms. First, a full parallel algorithm for interframe prediction is designed, including the sub-pixel interpolation parallel algorithm, motion compensation algorithm, integer pixel motion search parallel algorithm and sub-pixel motion search parallel algorithm. Then, we further put forward an improved step-like parallel motion search algorithm. This algorithm, which has a substantial increase in the speed of motion search, can effectively control the bit rate. Next, we realize a parallel algorithm for DCT and quantization. Lastly, a parallel algorithm for the deblock filter module is developed.The experimental result shows that, compared with the traditional CPU serial algorithm, the parallel algorithms designed in this paper can speedup the computation notably, and ensure the image quality and compression efficiency at the same time.