This PhD specification talk first reviews the work done so far in document image analysis, along them especially the enhancement of the readability in multispectral images. The next contribution and main part of the work covers the segmentation of characters in multispectral images. Since the application is based on ancient manuscripts there are challenges which are not encountered in “modern” document images, above all the fading out of ink, injuries through humidity and so on. These vitiations are caused through the aging process. Therefore multispectral imaging has been proven to be of value for the cquisition of such degraded documents. The segmentation of the characters is then based on the combination of spectral and spatial information. The main contribution is to model the character behaviour in the multispectral image data. A solution includes the usage of Markov Random Fields (MRF) and Conditional Random Fields (CRF) where local and global features are taken into account. The approach will combine local classification, local features and global features.