SEMCOG: an integration of SEMantics and COGnition-based approaches for image retrieval

Image retrieval is a key issue for many image database applications. Existing approaches include browsing and keyword, semantics, and cognition-based query processing. We argue that image retrieval using either approach alone is not effective. We introduce a system, SEMCOG (SEMantics and COGnition-based image retrieval), that allows users to pose a query using both semantics expressions and image examples. In our system, a query, such as "Retrieve all images in which there is a man on the right of a building and the building looks like this sketch", can be posed. This paper presents the architecture of SEMCOG, CSQL (Cognition and Semantics-based Query Language), the query language used in SEMCOG, and query user interfaces. Example queries are used to illustrate the image retrieval process in SEMCOG.