The Study Of Encoding Complexity Prediction And Rate Control In H.264 Encoder

H.264/AVC video coding standard achieves much higher coding efficiency than the previous standards such as H.261 H.263 and MPEG-2,4. In order to produce high quality decoded frame at a given target bit rate, rate control as an effective means of regulation varying bit rate characteristics of the coded bit stream is thus a necessary part of an encoder.In this paper, we aim at improving perceptual video quality and accurately estimating target bit for H.264 rate control.Firstly, the development of video compression technology and the structure and characteristics of H.264 video standards is introduced; the difference between H.264 and MPEG-4 video standard and a criterion for evaluation image quality are also presented.Secondly, we reviewed MPEG-4 rate control for multiple video objects and adaptive basic unit layer rate control for JVT, then the core problem which need to be addressed for designing H.264 bit rate control framework is presented, at the same time, the deficiency of the adaptive rate control for H.264 framework is analyzed.Thirdly, an improved rate control scheme is described. In the existing expression method of encoding complexity for H.264, MAD PSNR and MSE are widely adopted, with linear prediction as their mainly prediction methods. In the condition of high motion scenes or scene changes, traditional linear model for complexity prediction will be unable to gain accurate predicted complexity of current frame. MAD is used to stand for encoding complexity in this paper. By analyzing the relationship between average MAD of some previously encoded P frames in the GOP and the actual MAD of the last previously encoded P frame, a more accurate MAD prediction model is proposed to substitute linear model for MAD prediction. The experiment result shows proposed model performed better than linear model in test sequences. At the same time, in JVT-G102rl, the target bit estimation for a frame, as an important factor in determining the quantization parameter, is not allocated based on frame complexity. So, a method that bit budget is allocated to frames according to their MAD ratio, combined with the bits computed based on their buffer status is proposed; in order to improve overall visual quality, an adaptive QP adjustment method is also presented. With the above improvements, a new rate control scheme for H.264 rate control in this paper.Lastly, experimental results show that proposed MAD prediction model can accurately predict the MAD of current frame as compared with the linear model and the H.264 coder, using our proposed algorithm effectively alleviates PSNR surges and sharp drops for frames caused by high motions or scene changes. Moreover our proposal scheme improves the average PSNR.