Ink bleed-through removal of historical manuscripts based on hyperspectral imaging

Old manuscripts can be degraded by different reasons over time. The ink bleed-through from the back side of the page is a common problem which compromises the readability and the aesthetics of the document. Unsupervised bleed-through removal is a challenging task as pixel intensity of bleed-through areas can be very similar to fine-details of the writing. This paper provides a brief review about state-of-the-art methods for bleed-through removal. Moreover, we propose an algorithm for the segmentation of the bleed-through areas from a manuscript collected from the National Library of Oslo using hyperspectral imaging. This work also presents the restoration of the collected manuscript applying inpainting algorithms based on our segmentation approach