Fast AVS to HEVC transcoding based on ROI detection using visual characteristics

Video transcoding technique is essentially needed to meet the demands of different multimedia applications. The latest video coding standard HEVC has been finalized. It can achieve much higher coding efficiency than previous coding standards. AVS, Chinese advanced video coding standard, has been widely used in all kinds of applications. To convert AVS encoded video contents into HEVC coding format, a region-based transcoding algorithm is proposed in this paper, which exploits visual characteristics to divide video images into regions of different interest levels. According to the sensitivity of visual system to video contents, the proposed algorithm first utilizes the decoded information, such as macroblock (MB) partition mode, motion vector (MV) and transform coefficients, to segment a frame into three regions, namely: region of the most interest (mostROI), region of interest (ROI) and region of less interest (lessROI). Different transcoding strategies are then applied to those three regions, especially, search depth ranges for CTU partition are restricted correspondingly and prediction modes of coding units (CUs) are differently selected on the basis of the detected regions. Experimental results with different video contents are presented to show the performance of the proposed AVS-to-HEVC transcoding. It concluded that the proposed algorithm can save more than 50% of encoding time for HEVC while keeping a good transcoding performance.