Similarity Retrieval of Vector Images with Indirect Matching

Abstract

Listen

Translate article

Increase use of vector images have generated interests on effective ways of vector image retrieval. This paper presents a fast similarity retrieval of vector images. For reducing the computational cost in similarity evaluation, we proposed an indirect matching framework. This framework uses a small number of key images, which are previously randomly selected from a database. The basic idea of the framework is offline calculation of creating a similarity table with pre-selected key images. In the retrieval phase, by referring to the similarity table, the similarity of each database image to the given query is quickly estimated. Experimental results have shown that the retrieval time is greatly reduced by the proposed method without much deterioration of retrieval accuracy.