Retrieving a video from the Web is a time consuming operation. Instead, if a user is able to retrieve the representative frames of the video, s/he can decide whether s/he really need the video.
The idea is implemented in the clustering scheme. the clustering method will cluster the frames that have similar properties into the same shot. Shot is the group of frames in the same location, same person or do the same thing. Then we can use the center frame of that shot to be the representative frame.