A fuzzy approach to content-based image retrieval

Describes an approach to content-based image retrieval that can handle exemplar-based, graphical-sketch-based, as well as linguistic queries involving region (object) labels, attribute values, and spatial relations. Images are internally represented by fuzzy attributed relational graphs, where each node in the graph represents an image region, and each edge represents a relation between two image regions. A novel fuzzy graph matching (FGM) algorithm along with an indexing scheme based on a leader clustering algorithm (LCA) is used to facilitate fast retrieval. Several examples that illustrate FGM and LCA are shown.