Region-of-interest segmentation based on Bayesian theorem for H.264 video transcoding
Shu‐Fen Huang, Mei‐Juan Chen, Mian‐Shiuan Li · 2011
This paper presents a region-of-interest (ROI) segmentation technique to facilitate the video transcoding. In the low bit-rate transmission, ROI video coding can maintain better visual quality based on the properties of the human visual system (HVS). In this paper, to aim this purpose, the Bayesian theorem is utilized to automatically determine the moving foreground objects by using motion intensity analysis and skin color information for inter coded frames. Experimental results show that the proposed method can extract ROI macroblocks (MBs) effectively.