Representation and automatic organization of picture knowledge bases

Abstract

Methods are described for using human perceptions of pictures to represent the semantic content of these pictures in frame knowledge bases. A pictorial knowledge base management system is presented that uses a conceptual clustering algorithm for organizing a frame knowledge base into groups of semantically similar pictures. The conceptual clustering algorithm is based on the theory of gestalt. Applications of this system for content-based information picture retrieval and heuristic searching are also described.