Visual and audio segmentation for video streams
Tetsuya MURAMOTO, Masashi Sugiyama · 2002
In order to achieve object detection/location and tracking in video streams, this paper describes the video scene segmentation using visual and audio cues. For visual segmentation, frame distance and two techniques for speed-up are introduced. For audio segmentation, Cepstrum Flux and Block Cepstrum Flux parameters are introduced. Furthermore, experimental results of segmentation in both cases are described.