Video Skimming


Revised on May. 3, 2002


Various kinds of video contents become available via high-speed network as well as cable TV and satellite. These contents are often stored in tapes or in hard disks to watch them later. The usual way to check on video contents quickly is to fast-forward through them. However, in the current VCRs, the sound is not synchronized with the images and is either absent or partial during fast-forwarding, so it is hard to obtain and understand the important information from the video contents. We thinks that fast-forwarding makes easier to find a specific scene, because human ears are not up to the task of hearing sounds that are played back at a few times faster than normal speed. Our research objective is for users to comprehend the contents briefly by skimming videos.

There are a number of studies about video summarizing. Those are however concentrated on identifying video segmentations for constructing and representing the video structures; they do not concerned about understanding the contents within the specified total play time of the summarized video. Additinally, in most of those systems, users must specify the segmented scenes they want to look at, thus they cannot easily know the playback time of the resulting contents. In this research, we investiagte a new time reduction method for video skimming that is focused on the overall play time.

Related Papers


Toshiharu Sugawara
Toshihiro Takada
Shigemi Aoyagi
Koji Sato

Back to the top page

© Toshiharu Sugawara