Image retrieval by spatial arrangement underlies a matchingalgorithm, which “interprets” entities specified in the userquery on the entities appearing in the image of the database, and whichjointly compares their features and their spatial relationships. In thispaper we provide a graph-theoretical formulation and we discuss the sizefor the problem of spatial matching as encountered in the specificapplication context of retrieval by visual content. We then expound andevaluate three state-space search algorithms which support efficientcomparison of models through a heuristic bounding approach
展开▼