Hidden Markov Model for Content-Based Video Retrieval

Content-based video retrieval system is fairly recent and it is currently necessary to examine where it would just replace existing systems, where it can really bring some improvement and where it will open new possibilities. The users want to query the content instead of the raw video data. In this paper, we surveyed the art of video retrieval and proposed a basic framework for video retrieval based on an iterated sequence of navigating, searching, browsing, and viewing. We presented a framework of structural video analysis that focuses on the processing of high-level as well as low-level visual cues. We used HMM as the main content-based retrieval processes. Video and audio extracted features are used in the framework proposed. By using semantic features, it is possible to recognize high-level semantic actions and to encode more semantic details, which will enable users to find answers on questions easier and faster.