Spatial match retrieval using signature files for iconic image databases

Abstract

In multimedia information retrieval applications, content-based image retrieval is essential for retrieving relevant multimedia documents. The purpose of our paper is to provide effective representation of images when a pixel-level original image is automatically or manually transformed into its iconic image containing meaningful graphic descriptions, called icon objects. For spatial match representations, the 2D and the 9DLT schemes were proposed to search image results efficiently, satisfying certain spatial relationships (Chang et al., 1987). However, both representation schemes have a critical problem in that they represent each icon object of an original image as a point. As a result, they are not accurate enough to express spatial relationships between objects for handling original images with a complex scene. Therefore, we propose a new spatial match representation scheme to support effective content-based image retrieval.