Color and hyperspectral image segmentation for historical documents

Several historical documents from the collection of the National Library of Oslo were acquired using a hyperspectral scanner. While each of the documents has its specific characteristics, requiring different image preprocessing steps, the common goal for all documents is to increase their legibility. The aim of this study is to show the advantage of hyperspectral imaging compared to traditional color imaging, in particular for the task of ink separation using distance-based classification method.