AAV Video Encoding: Online Rate-Distortion Modeling and Multi-Timescale Rate Control

This paper is concerned with rate control (RC) for autonomous aerial vehicle (AAV) video encodingto produce a stable video bitrate and high quality videos under specific constraints. To this aim, we theoretically investigate the relationships between encoding parameters and bitrate and distortion, and design an accurate online rate-distortion (R-D) model by employing linear regression and predictive artificial intelligence (AI) methods. Considering the sensitivity of AAV video encoding to latency and power consumption, we further establish the mathematical relationships between encoding parameters and encoding time and power consumption. By integrating the R-D and encoding time and power consumption models, we formulate a multi-timescale optimization problem and propose a novel algorithm to solve it. Specifically, we decompose it into a family of single-timescale problems via an alternating direction method of multipliers (ADMM). In addition, we design an iterative optimization scheme to solve the single-timescale problems with low computational complexity. Extensive experiments are conducted to validate the designed model and algorithm. Experimental results indicate that the designed model has small estimation errors, and the variance of Y-PSNR achieved by the proposed algorithm is not greater than 26.3% of its counterpart.