On Retrieval of Nearly Identical Video Clips with Query Frame

Abstract

Video clips retrieval from a video database is a challenging problem for researchers because of voluminous data. Video Clips retrieval with query frame finds the video clips that have similar frames to query frame. Applying matching to all frames of any video clips is time consuming and computationally complex for database containing of several large size video clips. In the present article, we present a three level near identical video clips retrieval algorithm. In the first step, we identify keyframes of video clips in database through establishment of shot transition boundaries and then on keyframes of video clips we apply Integration of Curvelet transform and Simple Linear Iterative Clustering algorithm (SLIC) to get segmented query frames. In the second step, query frame will undergo Integration of Curvelet transform and Simple Linear Iterative Clustering algorithm to get segmented query frames. Now in the third step, the segmentation results of query frame are searched to obtain matched result in database. The method is experimented on 25 different video clips belonging to the categories- animations, serials, personal interviews, news and movies. Performance of the method is assessed with the parameters Precision, Recall, F-measure, Accuracy, Detection Percentage, Jaccard Criteria, and Missing Factor. The furnished performance results have shown sound performance of the proposed method over the other state-of-art methods.