H.264 Coding Algorithm Study

With the prevalence of communication and Internet, the requirement for video information is getting more and more, and the contradiction between the requirements and the present capability-limited video encoding technology is getting acute. To solve that, stepping after each has drew up the H.26x and MPEG-x video compression standard separately, the VCEG and ISO/IEC join together and make up the new generation standard-H.264, the newest video compression standard. The standard has the unique advantage that supporting the transmitting of high quantity video under variety net environments with very low bit rate. But because it adapts some new advanced technology, the complexity of the H.264 encoder is very high, which makes its application limited. For that, to reduce the encoder complexity with good encoding performance is one of the goals in the field, which is the goal of this paper too.Following simple analysis about video compression principle and current video encode standards, H.264 en-decoder is introduced, including encoding framework, character, some main modules such as intra prediction, inter prediction, translation, quantity and entropy. And the encoder was implemented on the visual C++. This paper made a more research on the inter prediction module:based on the character of the current algorithms and the reason for their weakness, this paper proposed some new algorithm as follows: firstly, when block matching, It was found that some algorithms can not be accord with the distribution of motion vector very much because they adopt only one search mask or the shape of the masks, to make the search process more object-oriented and adaptive, this paper identified the block as certain type based on the motion area, and then apply relevant search mask to search fine. Secondly, in the processing of the selection of macro block partition modes, JM reference software check all the modes and select the best one, so make time consumption too much. While some other algorithms make the selection processing faster by only check some "sample" modes, and because it is coarse, the encoder performance is worse. After discussing the reason for not making a good balance between the encoding performances and encoding complexity, based on the character of video sequences and the relationship between the partition modes, this paper proposed a fast algorithm which based on modes group and early termination. By the upper two optimizations, the new processing gets purposive and simpler.The experiment results show that, compared with the present standard algorithms, the new algorithm proposed in this paper can save more than 55% encode time with almost the same PSNR and bit-rate. So the new algorithm is good for practicality of H.264 standard.