Automatic scene change detection for h.264 video coding

Abstract The new generation video coding standard H.264/AVC became hot spot of research and application in multimedia field. Contributing higher coding efficiency and web applicability, H.264 video encoders usually assigned frame-type as I, P, or B with fixed rate for codec, but couldn't use various characters of videos or shots to alter encoding parameters flexibly. In this paper, mutual information and entropy theory were used to measure information difference between frames, and statistic based automatic threshold was used to cut video into shots. Then starting frame of each shot was assigned as I frame, while frames within one shot were assigned as P or B frame. We discussed way to avoid flashlight interrupt as complement. Simulation shows the method cuts shot edge exactly and robust to photoflash noise, camera or staffs moving; the encoder algorithm combining shot detection yields better compressing rate and enhances quality of re-build pictures than conventional approaches. Shot edges marked with H.264's own syntax could be easily extracted on decoding end, which provides an instead way of compressed field videos shot detection and good support for video severs basing on exact shot cutting.