Evaluation of distance measures for partial image retrieval using self-organizing map

Abstract

Digital image libraries are becoming more common and widely used as more visual information is produced at a rapidly growing rate. With this immense growth, there is a need to organize and index these databases so that we can efficiently retrieve the desired images. In this paper, we evaluate the performance of the self-organising maps (SOMs) with different distance measures in retrieving similar images when a full or a partial query image is presented to the SOM. Our method makes use of RGB colour histograms. As the RGB colour space is very large, another SOM is employed to adaptively quantise the colour space prior to generating the histograms. Some promising results are reported.