Research And Parallel Realization Of H.264Video Encoding Mode Selection Algorithm

H.264/AVC is the latest video coding standard which developed by the ITU_T Video Coding Experts Group (VCEG) and ISO/IEC Moving Picture Experts Group (MPEG). Compared to previous video standards, H.264compression ratio is bigger1.5to2times than MPEG-4at same image quality. H.264/AVC standard has better coding efficiency and is more suitable for low-bandwidth and high-quality network video applications. In the H.264/AVC, rate distortion optimization algorithm is utilized to judge mode-selection, while efficient motion estimation algorithm is used in interframe coding. It has effectively reduced the time domain and space domain redundancy. But the standard is getting more complex at the same time. In order to apply the H.264to the low bit rate real-time system, there is a demand optimizing the encoding algorithm. Therefore, how to improve the efficiency of mode selection, and make mode-selection process more efficient and more robust has become a researche focus.The basic principles and key technologies of the H.264video coding standard is discussed in this paper. Then, the article introduces the video encoding mode selection algorithm, and analyses the SKIP strategy algorithm’s disadvantage of low accuracy and poor robustness. According to the weak sensitivity of macro block histogram to the illumination changes and rate distortion algorithm inter-frame difference discriminative, a fuzzy discrimination method based on the macro block histogram and the minimum rate distortion is raised. The new method has been used to sport video sequences which have different intensity. The result shows that it improves the classification rate of the SKIP strategy effectively, thereby speeding up the video coding. Speaking with data, the time of inter-frame coding mode saved an average of18.15%, and average PSNG decreased by only0.03dB.Finally, according to the H.264encoder hierarchy, a multi-frame multi-slice parallel method is proposed, which point to frame-level parallelism and slice-level parallel data having a strong data correlation. One slice is divided into there segments, that is, before encoding-slice, encoding slice and after coding slice, which reduces the data-dependent of slices. Then the parallel code is achieved by OpenMP. The test shows that average bit rate raises4.087%while average encoding speed increases by45.43%, in almost the same image quality.