The Art of the Invisible: Super-resolution, NIR and UV Imaging Basics

Ultraviolet (UV) and Near-Infrared (NIR) imaging allow us to see far beyond the limits of human vision, revealing hidden details in art, artifacts, and forensic evidence. UV techniques, including both fluorescence (UVF) and reflected (UVR), expose surface-level phenomena such as varnish layers, retouching, and certain pigments. NIR imaging, particularly infrared reflectography, penetrates through paint layers to uncover underdrawings, pentimenti, and hidden signatures. The reconstruction process combines these spectral images with advanced processing like 3D photogrammetry and principal component analysis (PCA) to create detailed digital twins, enabling conservators and researchers to study an object’s history, condition, and composition in unprecedented depth. Ultimately, computational image processing holds tremendous potential to reveal hidden histories and preserve our cultural heritage for generations to come. I ran super resolution on an image with a familiar face yesterday using a transformer based model instead of a GAN. GANs can introduce hallucinations, meaning they invent plausible looking details that are not truly there. Transformers avoid this problem by reconstructing high frequency details through long range pixel relationships without fabricating false features. The result was a clean, faithful upscale with no hallucination, just sharper resolution from the original data.