For H.264 Video Coding Rate Control Strategy

Rate control has a direct impact on the bandwidth and picture quality of the coded video data, the features of H.264 video bit stream and the problems associated with rate control are intensively analyzed and studied. Three improved rate control algorithms are proposed for different applications: a single-pass constant bit rate (CBR) rate control algorithm, a single-pass variable bit rate (VBR) rate control algorithm and a two-pass variable bit rate (VBR) rate control algorithm.Firstly, the features of single-pass CBR rate control are detailedly analyzed, and two points of H.264 — RDO technology, integer transform and quantization which highly relate with rate control are discussed. Then we proposed an improved single-pass CBR rate control algorithm according to the features of H.264 video bit stream. In this algorithm, we estimate the complexity of current picture in DCT domain based on the information of coded pictures, then control the bit rate in GOP and picture layers. Scene cuts are detected and processed. This algorithm not only has an accurate control effect, but also has low complexity and makes it applicable to real-time video coding.Secondly, we discuss the advantages, characteristics and key points of single-pass VBR rate control, and then proposed an improved algorithm. Based on R-D model, the algorithm allocates target bit and determine quantization parameter according to the predicted complexity of current picture in DCT domain. Scene cuts are detected and processed. The proposed algorithm can well adapt to scene cuts and implement a smooth quantitative video quality. Additionally, its low computational complexity makes it applicable to real-time video coding.Thirdly, we analyzed the theory, characteristics and applications of two-pass VBR rate control and proposed an algorithm for H.264. In the first pass, CBR coding is employed to gather the encoding statistics of the input sequences. On the basis of the obtained statistics information, such as encoding complexity, PSNR, macroblock distributions of each coding type and etc, an offline processing is applied on these data to detect scene cuts, accurately allocate target bits and determine each frame's quantization parameter based on rate-distortion model, then the second pass coding is performed. Simulation results show that the proposed algorithm can not only meet the global bit budget and buffer conformance, but also result in a constant picture quality.This VBR algorithm is capable of working well in digital video storage applications.