Hybrid contextmodeling and wavelet transform for region of interest detection and rate control based on VP9 video codec

摘要

Region of interest detection is a useful technique in video encoding. This work presents a modelto distinguish region of interest area based on VP9 video codec. Themodel contains high motion,contrast, color sensitive areas which fits for human visual system. A hybrid detection algorithm,includingWavelet Transform and ContextModeling method, is proposed. The algorithm approximatelyreveals ROI details for the sake of pursuing computing time. In ROI-based encoding,adjusting encoding parameters for region of interest (ROI) during the encoding procedure canbring turbulence in bits output, a rate control model is established in this work. We also designa rate control mechanism for ROI encoding mode both in CBR and VBR to maintain the outputbit rate stable. The encoding parameter are set specifically in different ROI areas to improve bothvisual effects and PSNR. The proposed detection method and rate control mechanism work wellin VP9 codec and contribute saving bits according to experimental results.