Creating Digital Estampages of Sri Lankan Stone Inscriptions using Multispectral Imaging

Abstract

Abstract Stone inscriptions in Sri Lanka hold immense historical and cultural value, yet they are increasingly losing legibility due to natural weathering, erosion, and vandalism. Traditional documentation methods like estampages, while useful, are often time-consuming, dependent on favorable weather conditions, and susceptible to inaccuracies. This research presents a novel approach utilizing multispectral imaging (MSI) to improve the readability of these inscriptions and facilitate precise character extraction. A structured methodology, rooted in Design Science Research, guided the iterative design and testing of an MSI-based digital estampage pipeline. This pipeline incorporated essential steps such as image correction, preprocessing (including noise reduction and binarization), and advanced image processing techniques like edge detection. To further optimize visual clarity and text segmentation, a U-Net-based model was developed and trained. The key findings demonstrate that MSI significantly surpasses traditional estampage methods in both readability and character visibility. Integrating MSI into existing archaeological workflows proved to be feasible, e!cient, and well-received by domain experts. Furthermore, the developed pipeline e”ectively reduces fieldwork time while maintaining data integrity. Quantitative evaluations, utilizing PSNR and FID metrics, showed a clear improvement in image quality compared to traditional methods. User studies conducted with archaeologists revealed a remarkable enhancement in legibility and usability. Expert feedback further corroborated these findings, with the system receiving an average rating of 4.6 out of 5 for its usability and output clarity, thereby a!rming the significant value of MSI in heritage documentation and analysis.