Colour map classification for archive documents

Abstract

This paper presents our work on color map classification for archive documents. An approach is proposed which is very similar to the way that humans perceive document colors, by using the diversity quantization algorithm applied within the HSV colour space. Template color maps are manually registered for sample document images; images from batches to be classified are then mapped onto the closest template color map using a fuzzy color classification algorithm. The results of testing this approach on batches of archive index cards from the UK Natural History Museum were very encouraging. We tested over 400 biological specimen index cards, and achieved more than 98% correct color classification. © Springer-Verlag 2004.