A concept-based approach to video semantic analysis and coding

We propose a concept-based approach to video semantic analysis and coding for video retrieval. The proposed approach is aimed at bridging of the gap between the low-level video features and high level of semantics of video and providing a convenient and efficient way to create semantic descriptions and retrieve video. In the paper, we built an ontology model of the video semantic objects and described the basic relations (part-of, kind-of, instance-of and attribute-of) among these concepts. According to the human concept response of video, we propose a scalable video encoding. Compared to feature-based indexing and retrieval, the proposed concept -based coding presents better results in several kinds of videos.