Multispectral imaging for analyzing ancient manuscripts

In this paper we propose a method for segmenting characters in multispectral images of ancient documents. Due to the low quality of the document images the main idea of our study is to combine the multispectral behavior and contextual spatial information. Therefore we utilize a Markov Random Field model using the spectral information of the images and stroke properties to include spatial dependencies of the characters. The whole process is parameter free since we calculate the stroke properties and the Gaussian parameters for the imaging model automatically. The study shows the effectiveness of using multispectral data for a computer aided analysis of ancient text documents. We compare the results of the proposed segmentation method to traditional methods based on color images as well as gray level images.