A NOVEL METHOD FOR WIPE TRANSITION DETECTION AND PATTERN RECOGNITION IN AUTOMATIC VIDEO PARTITION

The first step of video analysis, such as automatic video classifying, organizing and indexing, is to partition video into shots. So, various shot boundaries should be detected. Wipe is a special shot transition pattern and is very difficult to detect and recognize. Presented in this paper is a novel method for wipe transition region detection and pattern recognition based on pattern templates. In the proposed scheme, various wipe transition patterns are extracted into pattern templates. Then an algorithm is designed for matching video sequences with pattern templates. Finally, the matching result is mapped to parametric space for eliminating motion disturbance and judging easily by Hough transform. Results show that the method is capable of detecting wipe transition regions and recognizing wipe patterns accurately, including various fancy wipe patterns that are existent or inexistent.