Searching technique in a spectral image database

Abstract

We propose a technique for searching images from a spectral image database. In the current stage, the technique uses spectral color information. The technique is based on the Self-Organizing Map (SOM), which is used to create the Best-Matching Unit (BMU) histograms for every spectral image in a database. Searching program calculates BMUhistogram for the wanted spectral image, and then the histogram similarities between the wanted BMU-histogram and the BMU-histogram database are calculated. The search result is shown as images that are ordered according to the histogram similarity. The searching technique and the current experimental results are reported. © Springer-Verlag; 2003.