A content-based video query agent using feature-based image search engine

A novel content based video query agent (VQA) for querying distributed video databases using a multimodal (image and sound) index was proposed. As the image search engine, a simple and variable dimension feature based matching method was adopted and an optimum representative frames (r-frames) selection technique was used for the automatic indexing method. Comparative experiments with the conventional region based image search engine using static image database and visualized sound database showed that the difference of the retrieval accuracy is small and the proposed feature based search engine is advantageous for VQA application. Results of video frame retrieval experiments showed that the proposed VQA is a promising query tool for distributed video databases on the network.