A perceptual approach for visible and NIR render in ancient documents visualization

Imaging techniques are commonly used to improve the legibility of text in ancient or degraded documents. Infrared reflectography is one of the best methods for examining documents written in a carbon-based ink since the contrast between the ink and the support is usually much greater in that band. In these cases, for archival and study purposes, the visible and the corresponding infrared images are usually shown next to one another. In other cases, like e.g. papyri, this separation between text and background is not easily achievable, and therefore, a more sophisticated elaboration is needed. Moreover, in these cases, the background is a relevant source of information on the original document, and thus its integration with the extracted text could represent an effective solution for scholars. In this paper, we present a new method to improve the legibility of the text in visible light reflectance images without altering the appearance of the support. The method is based on the fusion of the text extracted from the infrared image with the visible image. The contrast of the text in the infrared image is first enhanced with the Automatic Color Equalization algorithm, a perceptual-based enhancement technique, and then extracted with a thresholding process. The proposed method allows to visualize different information (text, background, alterations, etc.) on a single image, and can be also used for an efficient archival approach.