Based On The H.264 Bit Rate Control Algorithm Research

The main target of video compression code is to limit the distortion of code under the condition of bandwidth and restricted rate. In order to get the most optimized compression efficiency, ensure the coded data transfer in the channel successfully, and obtain the best video quality. The distorted optimization and the rate control are necessary in the encoding process. The purpose of rate control is to adjust the output code stream, ensure make use of the bandwidth effectively, and achieve the high steady image quality at the same time.The paper first discusses H.264's encoding system. Compared with the previous coding standards, H.264 adopts many new technologies, including intra-frame and inter-frame prediction, integer transform and quantization, entropy coding and deblocking filter, which make the coding efficiency increase.Secondly,discuss the principle of rate control are, and analyze the classic rate control algorithms TMN8 and TM5, VM8.Finally, study the rate control algorithm JVT-G012 that is usually sued in H.264 deeply, and find some deficiencies in JVT-G012, mainly including: Initial QP selection just depends on the bits per pixel (bpp) or accords to experience estimates, without considering the complexity of frame, is not precise enough; MAD estimation of image complexity only makes use of the temporal correlation that exists between the adjacent frames or macro blocks, namely, the estimation only considers global motion of images, while ignoring their local motion; Average distribution is used in the allocation of target bits, without considering the complexity of current frame or the macro block. This paper presents the following improvements for these deficiencies:Put forward a selection algorithm of initial quantization parameter based on bpp and the complexity of I-frame. The algorithm introduces the influence of the first I-frame's complexity of different sequences'to the best initial quantization parameter, on the basis of the initial quantization parameter computed by the original algorithm. Then get the initial quantization parameter as a function of bpp and the complexity of I-frame through a lot of experiments statistics. The improved algorithm makes the initial quantization parameter selection more reasonable, the PSNR of coded images and bit rate control are better than the original algorithm.An algorithm to estimate the image complexity is proposed, combining MAD and gradient, the image complexity can be effectively reflect its global motion and local motion. The experiments show that the improved algorithm can control rate more accurately in the same coding quality.Based on the new image complexity, allocate the number of target bits according to relative complexity of encoding frame. Experiment results show that rate control is more accurate, and the fluctuation of PSNR is small.