Luminance-Agnostic Colorization Framework for Historical Orthochromatic Photography

A recent paper on arXiv (2608.10798v2) presents a colorization framework that is independent of luminance, conceptualizing colorization as a complete RGB image editing process utilizing a foundational image-editing model. In contrast to conventional approaches that function within Lab space and maintain the input luminance channel, this innovative technique accommodates brightness variations and shows greater resilience when grayscale departs from typical natural-image luminance, as seen in historical orthochromatic photography. The researchers propose a mixed grayscale objective that trains the model using both standard luminance grayscale and a red-insensitive grayscale format, effectively linking modern panchromatic and historical orthochromatic contexts. Tests conducted on COCO, ImageNet, and a multi-instance benchmark indicate that this approach performs competitively with standard grayscale inputs and demonstrates significantly enhanced robustness with orthochromatic inputs. The paper is classified as a cross-type announcement and can be accessed at https://arxiv.org/abs/2608.10798.