An algorithm for content-based retrieval in multimedia databases

Abstract

Proposes a linear-time algorithm, referred to as SIM/sub vs/, for processing spatial similarity queries in multimedia databases. SIM/sub vs/ computes the degree to which the spatial relationships among the objects in a database image conform to those specified in the query. SIM/sub vs/ recognizes translation, scale and rotation variants of an image, and variants generated by an arbitrary composition of translation, scale and rotation transformations. The performance of SIM/sub vs/ is compared with those of two previously-known algorithms using a testbed image collection. This comparison demonstrates that SIM/sub vs/ is not only more efficient but performs as well as or better than the other algorithms.