H.264 Rate Control Algorithm And Optimization

Video coding is an important technology in video processing. It aims at compressing video data effectively by lossy or lossless coding method. H.264/AVC is the lastest video coding standard. It has been co-developed by the joint team of ITU-T and ISO. Compared with the former video coding standards, H.264/AVC achieves clearly higher compression efficiency, better image quanlity and network compatibility.Rate control is a very important part of the video compressing encode. It has a significant impact on the coded video quality. The main task of rate control is to ensure that on the average, coding bit-rate does not exceed the target bit-rate. This has to be done while minimizing distortion in the decoded sequence. Optimising the tradeoff between bit-rate and image quality is a challenging task of rate control.The dissertation mainly researches on the rate control algorithm of H.264/AVC. The research work in this dissertation could be summarized as below:1,By more accurately predicting frame complexity in terms of using the statistics of previously encoded frames of video sequences, we use MAD_ratio (ratio of mean absolute difference) as a measure for global frame encoding complexity. Target bits are allocated to frames according to their MAD_ratio, combined with the bits computed based on their buffer status. Based on the time-space coherence, we present a fixed-coefficient weighted model to substitute the linear model to predict the MAD of basic unit, and adjust the predictive value of MAD by the change trend of MAD. The experimental results show that improved algorithm effectively elevates PSNR and closes target bit-rate.2,In the application of low bit-rate video coding, a ROI-codec-supported rate control algorithm gives priority to ensuring the image quality of ROI (Region Of Interest). It has dramatically significance of application. This dissertation proposes a ROI-codec-supported H.264/AVC rate control algorithm. The proposed rate control algorithm is suitable for the applications of low-bit-rate video conferencing. In this algorithm, a simple and efficient approach of extracting ROI is proposed. According to the image complexity and interested extent, it distributes bits to ROI and non-ROI respectively. Meanwhile, we can get the target bits of one macroblock by the encoding complexity and the target bits that allocated to ROI or non-ROI. The experimental results show that improved algorithm significantly enhances the PSNR value of ROI and gets better subjective and objective image quality significantly from decoded frame.