H.26x Series Algorithm

With the development of computer and communication technology, network video conference, video telephone and network monitor etc. have got more and more applications. In all of these applications, the most widely used video compression standards are the ITU-T H.26x standards. One vital requirement of video conference system and monitor system is real-time. But, because of the high computation complexity of H.26x, especially H.264, we must improve algorithms' processing speed in order to improve system's efficiency. This thesis analysis the computation complexity through three aspects of rate control, integer-pixel motion estimation and quarter-pixel motion estimation, then it compares the existing algorithms and brings about new algorithms.Rate control algorithms can keep video stream to transform smoothly and to utilize the bandwidth efficiently. The traditional rate control algorithm is the R-D model. These algorithms either have high complexity or lack precision. This thesis proposes a new rate control algorithm based on p-domain linear source model. The new algorithm reduces the computation complexity without loss of precision.In the area of integer-pixel motion estimation, the gross searching algorithm is too complicated to put into practical use. A series of fast algorithms have been developed such as the classical 3-step searching, 4-step searching, gradient decent searching, diamond searching and hexagon searching. This thesis proposed a multiple level diamond searching algorithm based on the multi-resolution model and diamond searing model. The computation complexity of the new searching algorithm is much less than the previous algorithms.The traditional quarter-pixel motion estimation searching needs to search 16 pixels. But in practice, the half-pixel searching has great correlation with the integer-pixel searching, and the quarter-pixel searching has great correlation with the half-pixel searching. This thesis proposes a new searching method based on these correlations to enhance the search efficiency.