The Research On Fast Motion Estimation Algorithm Based On H.264/AVC

In the era of information technology, multimedia technology is also developing rapidly. The researchers found that about 70% of people’s daily access to information was received through the visual. With the progress of science and technology, our demand for video technology will be more and more. For example: watching online movies or TV shows, live video broadcast, through the video sharing sites to upload their own works, video conference, online education, remote medical service and so on. However, the volume of the original video is very large. If it is stored, it requires a lot of storage space; if it is transmitted, it requires a lot of bandwidth. Therefore, it is necessary to compress the original video. H.264/AVC is a video coding standard developed by VCEG of ITU-T and MPEG of ISO/IEC, is now the most mature, the most widely used video compression technology. Motion estimation is the core technology of video coding, which consumes 70%(1 reference frame) to 90%(5 reference frames) of the whole coding process. If we can accelerate the speed of motion estimation, it is necessary to improve the real-time coding performance of the encoder, which is the significance of the study of fast motion estimation algorithm.Firstly, the codec principle of H.264 is described, and some key techniques of H.264 are introduced. Then, this paper expounds the principle of motion estimation, and introduces several classic early motion estimation algorithms. At the same time, make a concrete analysis of the advantages and disadvantages of these early algorithms. Then deeply study of the UMHexagonS algorithm with excellent performance, and analysis of its advantages and disadvantages. After that based on UMHexagonS algorithms from three different aspects make the optimization:(1) based on the direction of the motion vector search;(2) based on the motion vector distribution optimization;(3) search the adjustment of the dynamic range.Experimental results show that the above three optimization points will be combined to achieve the best optimization results. Compared to the UMHexagonS algorithm, the proposed algorithm can effectively save the motion estimation time 11.490% to 42.892% while the rate-distortion performance is not compromised, and improves real time coding performance of the encoder.