Content-adaptive parameters estimation for multi-dimensional rate control

Multi-dimensional rate control schemes have been recently utilized to adapt video streams to dynamic network conditions and heterogeneous devices. However, current multi-dimensional rate control methods, which estimate the model coefficients using fixed update duration, usually yield inaccurate parameters for dynamically changing video content. To address this problem, a content-adaptive parameters estimation scheme is proposed for multi-dimensional rate control. Firstly, we propose to estimate the parameters using dynamical update duration based on video content and the update duration of the model coefficients is determined by jointly considering the varying picture complexity and feedback information from the actual encoding results, which can improve the model parameter estimation accuracy. Secondly, a coarse-to-fine initial parameter calculation method is proposed to refine the initial frame rate according to the channel condition and the video sequence characteristics. Extensive experimental results show that the proposed solutions outperform the state-of-the-art schemes, especially for video sequences with high temporal and spatial complexity. Furthermore, our algorithm also slightly reduces the computational complexity as compared to related algorithms.