Neural Frank-Wolfe Policy Optimization for Region-of-Interest Intra-Frame Coding with HEVC/H.265

• Published in: Visual communications and image processing (Online) pp. 1 - 5
• Main Authors: Ho, Yung-Han, Kao, Chia-Hao, Peng, Wen-Hsiao, Hsieh, Ping-Chun
• Format: Conference Proceeding
• Language: English
• Published: IEEE 13.12.2022
• Subjects: action-constrained reinforcement learning; Bit allocation; Bit rate; Distortion; Image coding; Image quality; rate control; Rate distortion theory; region-of-interest (ROI); Reinforcement learning; Visual communication
• ISSN: 2642-9357
• DOI: 10.1109/VCIP56404.2022.10008853

This paper presents a reinforcement learning (RL) framework that utilizes Frank-Wolfe policy optimization to solve Coding- Tree-Unit (CTU) bit allocation for Region-of-Interest (ROI) intra-frame coding. Most previous RL-based methods employ the single-critic design, where the rewards for distortion minimization and rate regularization are weighted by an empirically chosen hyper-parameter. Recently, the dual-critic design is proposed to update the actor by alternating the rate and distortion critics. However, its convergence is not guaranteed. To address these issues, we introduce Neural Frank-Wolfe Policy Op-timization (NFWPO) in formulating the CTU-level bit allocation as an action-constrained RL problem. In this new framework, we exploit a rate critic to predict a feasible set of actions. With this feasible set, a distortion critic is invoked to update the actor to maximize the ROI-weighted image quality subject to a rate constraint. Experimental results produced with x265 confirm the superiority of the proposed method to the other baselines.