A ROI-based bit allocation scheme for HEVC towards perceptual conversational video coding

In this paper, we put forward a simple yet efficient scheme, based on the HEVC platform, to improve the subjective quality of conversational videos. Our scheme, in fact, is a perceptual bit allocation scheme (PBAS) for HEVC, aiming at increasing the amount of bits assigned in ROI regions. First, PBAS scheme employs the robust SURF cascade face detector to extract the regions of interest (ROI) in a conversational video. Then, based on the extracted ROI regions, a perceptual rate-distortion model is presented in PBAS scheme towards subjective quality improvement. By optimizing the rate-distortion model, we can have the optimal settings of bit rates for the ROI and non-ROI regions using PBAS scheme, respectively. Finally, the experimental results show that the PBAS scheme can improve significantly the subjective quality of conversational videos compressed by HEVC.