🤖 AI Summary
This work addresses the limitations of conventional image colorization methods, which operate in Lab space with a fixed luminance channel and thus struggle with non-standard grayscale inputs—such as orthochromatic photographs—and cannot adjust brightness. To overcome this, the authors propose a luminance-agnostic colorization framework that treats colorization as a full RGB image editing task. They introduce a hybrid grayscale training objective that uniformly accommodates both modern panchromatic and historical orthochromatic photographic inputs. By removing the fixed-luminance constraint, the method significantly enhances robustness to unnatural grayscale images. Experiments demonstrate superior performance on COCO, ImageNet, and multi-instance benchmarks, with markedly reduced color artifacts on orthochromatic images. Human evaluations confirm that the visual quality of the results substantially surpasses that of existing approaches.
📝 Abstract
Most image colorization systems operate in $Lab$ space by predicting chroma ($ab$) while preserving an input-derived luminance channel ($L$). While effective on standard benchmarks, this fixed-luminance design restricts brightness changes and becomes unreliable when grayscale formation deviates from natural-image luminance, as in historical orthochromatic photography. We propose a luminance-agnostic colorization framework that formulates colorization as full-RGB image editing using a foundation image-editing model. To bridge modern panchromatic and historical orthochromatic conditions, we introduce a mixed grayscale objective that trains the model under both standard luminance grayscale and a red-insensitive grayscale formation. Experiments on COCO, ImageNet, and a multi-instance benchmark show that our method is competitive on standard grayscale inputs and substantially more robust under orthochromatic inputs, with qualitative comparisons and a human study indicating fewer visible color artifacts.