A Model Parameter Estimation Scheme Based on Fast Search for CTU-Level Rate Control in HEVC

Abstract

The estimation of a model parameter is very important for coding tree unit (CTU)-level rate control as it can significantly affect bit allocation and coding performance. However, CTU-level rate control models sometimes fail as the accuracy of the model parameter is not well considered in high efficiency video coding (HEVC). Based on this observation, a model parameter estimation scheme that is based on a fast search for CTU-level rate control is proposed. First, the estimation of a model parameter is formulated as an optimization problem. Second, we transform the optimal model parameter estimation into a process to search for the most similar CTU. To improve the searching efficiency and reduce the parameter estimation error, a new spiral search mechanism is proposed to find the most similar CTU from inside to outside. According to the experimental results, the proposed method can distinctly enhance the accuracy of the CTU-level rate control and consistently outperform HM16.18 and other state-of-the-art algorithms for all testing configurations. An average reduction of 7.8% and a maximum reduction of 10.0% of the BD-Rate were achieved compared with HM16.18. An average reduction of 3.4% and a maximum reduction of 4.6% of the BD-Rate were achieved compared with other state-of-the-art algorithms. A slight computational complexity overhead is observed.