Global Journal of Computer Science and Technology

Rapid growth in multimedia technologies facilitates the acquisition and storage of videos in a cost effective manner; leads to the processing of ginormous videos. However, for effective processing, suitable search methodologies are essential pre-requisite in any video processing system. In this paper, we propose a proficient content-based video retrieval system with the aid of extensive features and LSI. The system is comprised of three stages, shot segmentation, feature extraction and retrieval of similar video clips. Initially, the system performs the task of shot segmentation and feature extraction for every database video. In the feature extraction, we extract the most dominating features that include motion object, color histogram, texture and contour features from every shots. All the extracted features that correspond to a video are stored in a feature database. In the retrieval of similar video clips, the system retrieves a required number of video clips that are similar to a given query clip. The retrieval can be accomplished by performing the shot segmentation and feature extraction for the query clip. Then, the LSI-based similarity is measured between the feature set of the query clips and the feature database. Hence, the database videos that are relevant to the query clip can be retrieved more proficiently using the proposed system. Eventually, the system is evaluated by determining the precision and recall for different query clips. KeywordsContent-based Video Retrieval (CBVR), Shot segmentation, Motion Estimation, Gaussian Mixture Model, Color Histogram, Query Clip, Latent Semantic Indexing (LSI)I.