Video shot segmentation and key frame extraction based on SIFT feature

Abstract

Content complexity can be defined by SIFT feature of video content. According to the differences in the content complexity and the content change ratio calculated from video content, a video shot boundary can be detected. In general, for programs with many storylines, every story takes place by turns so that the number of key frames in a story can be estimated with the method as stated above. At last, based on the video content complexity, as well as the difference between frames and the number of key frames, every key frame can be extracted from a video. The experimental results show that the proposed method can improve both the ratio of accuracy and recall in the stages of shot segmentation and key frame extraction, by using both the features of the content complexity and the content change ratio based on SIFT feature.