An Adaptive Model Parameters Prediction Mechanism for LCU-Level Rate Control

摘要

In this paper, an adaptive model parameters prediction mechanism is proposed to take the place of parameter updating method based on the experience value in HEVC. And, normalized mutual information is exploited to guide the model parameters of α and β prediction. Experimental results show that the proposed algorithm controls the rate error within 0.1%. Compared with HM16.9, it further improves average 0.03% bit rate accuracy. Meanwhile, the proposed algorithm yields average 1.10% BDBR reduction and 0.05 dB BDPSNR enhancement without introducing additional computation. And it demonstrates less bit rate fluctuation, which achieves better adaptability for HEVC in real-time transmission.