Real-time region-of-interest video coding using content-adaptive background skipping with dynamic bit reallocation

In this paper, we propose a low-complexity content-adaptive background skipping scheme for region-of-interest (ROI) video coding in real-time wireless video telephony applications. Based on the real-time content information of the current and previous frames, such as foreground shape deformation, foreground and background motion, and background texture complexity, the proposed algorithm dynamically decides whether to skip Non-ROI macroblocks of current frame and reallocate saved bits to ROI and coded non-ROI. In this work, we proposed and compared three bit reallocation strategies, and the one predicting the future events using a Bayesian model is found the most efficient. Experimental results indicate that the proposed scheme significantly outperforms other methods by up to 2.5dB.