Research On The H.264 Rate Control Technology

Video compression technology is one of the emphases of research in multi-media information processing technology. In the past two decades, several video coding standards emerged. H.264/AVC video coding standard was developed jointly by ITU-T and ISO/IEC. Employing a number of new techniques, H.264/AVC is capable of enhancing coding efficiency greatly, and has been nominated to be the most efficient video coding standard in recent years as a result. To achieve better image quality under the channel bandwidth, time-lapse and buffer limitation, appropriate strategies must be adopted to control the rate of encoder's output into a certain range. Therefore, rate control has become one of the key techniques of video coding. However, the coding algorithm of H.264 is different from the former video coding standards, therefore its rate control is comparatively more complicated.In this dissertation, the theory of rate control is analyzed, the reasons and parameters that could cause the rate variance are discussed. The algorithms and rate-distortion (R-D) model of several video coding standards are also reviewed and summarized. On the basis of existing research achievements of H.264 rate control technology, a new H.264 rate control algorithm, prediction stimulation based and space-time related H.264 rate control algorithm (PS-STR algorithm), is proposed. The newly proposed algorithm effectively overcomes the limitation of existing algorithms in calculation cost and prediction precision. The experimental results illustrate that compared with the existing classic algorithms, PS-STR algorithm is capable of predicting the variance trend of video complexity more precisely and performing a more effective bit allocation. Consequently, the bit rate after encoding is closer to the objective one and the PSNR is also enhanced. Be dead against some problems caused by MAD prediction and bit allocation, two improved schemes, bit allocation scheme employing free-unit and fine rate control scheme employing parameter λ_mode modification, are also proposed for PS-STR algorithm. The experimental results indicate that by adopting the improved PS-STR algorithm, the actual rate is even much closer to the objective rate and the PSNR is also increased. The rate curve is much more flat with less variance.