Temporally Adaptive Quantization Algorithm in Hybrid Video Encoder

Abstract

In video coder, inter-frame prediction results in distortion propagation among adjacent frames, and this distortion dependency is a crucial factor for rate control and video coding algorithm optimization. The macroblock tree (MBTree) is a typical temporal quantization control algorithm, in which a quantization offset δ is employed for adjustment according to the amount of distortion propagation measured by the relative propagation cost ρ. Appropriate δ-ρ model is the key to the MBTree-like adaptive quantization algorithm. The default δ-ρ model in MBTree algorithm is designed in an empirical way with rough model accuracy and insufficient universality to different input source. This paper focuses on this problem and apply the competitive decision mechanism in exploring optimal δ-ρ model, and then proposes an improved δ-ρ model with rate distortion optimization. The simulation results verify that the improved MBTree algorithm with the proposed model achieves up to 0.14 dB BD-PSNR improvement, and 0.29 dB BD-SSIM improvement. The proposed algorithm achieves better temporal bit allocation and reduces the distortion fluctuation in temporal domain, achieving in adaptive quantization control.