Content based sub-image retrieval via hierarchical tree matching

Abstract

This paper deals with the problem of finding images that contain a given query image, the so-called content-based sub-image retrieval. We propose an approach based on a hierarchical tree that encodes the color feature of image tiles which are in turn stored as an index sequence. The index sequences of both candidate images and the query sub-image are then compared in order to rank the database images suitability with respect to the query. In our experiments, using 10,000 images and disk-resident metadata, for 60Σ (80Σ) of the queries the relevant image, i.e., the one where the query sub-image was extracted from, was found among the first 10 (50) retrieved images in about 0.15 sec.