We describe a new method to search for and localize repeating sequences in video data. Each video is divided into keyframes based on color histograms of frames in video.This is very essential to deal with videos because video data can contain very similar frames several times. Each keyframe is treated as a member of a sequence and another sequence containing similar keyframes is searched. Some features are need to find similar keyframes. Keyframes are represented by SIFT descriptors histograms and color distributions. Similar keyframes are found according to these features and sequences are detected by a tree based approacah. A tree is created based on similar keyframes of consequitive keyframes. Sequences are detected by using trees of consequitive keyframes. Experiments are carried out on two full length feature films.

DATE: 19 March, 2007, Monday@ 15:40
PLACE: EA 409