Semantic Color Naturalness Breaker: Preventing Illegitimate Colorization via Content-Aware Color Priors
2026-07-20 • Computer Vision and Pattern Recognition
Computer Vision and Pattern RecognitionMachine Learning
AI summaryⓘ
The authors address the problem of unauthorized colorization of black-and-white images, like manga or old photos, using automatic colorization tools. They build on previous work that adds tiny changes to grayscale images to confuse these tools but go further by making the colorized results look unnatural without changing the original grayscale image much. They introduce a new way to measure how believable the colors are based on image content, which helps improve their method and evaluate its success. Their experiments show that their approach works well even when the image is slightly changed after protection, making it useful for real-world sharing scenarios.
image colorizationgrayscale imagesadversarial perturbationsemantic-level attackscolor plausibilitycontent-aware metricsImageNetimage protectionpost-processing robustnessuncolorable examples
Authors
Yuki Nii, Futa Waseda, Ching-Chun Chang, Isao Echizen
Abstract
Automatic image colorization enables large-scale and low-cost reuse of grayscale media (e.g., manga panels and archival photographs), facilitating unauthorized reuse and redistribution. Once released online, grayscale content can be readily turned into unauthorized colorized derivatives using off-the-shelf models, creating a practical need for proactive, content-side protection at publication time. Building on Uncolorable Examples (UE), which add imperceptible perturbations to released grayscale images to degrade unauthorized colorization, we propose Semantic Color Naturalness Breaker (SCNB) -- a semantic-level UE framework that drives colorization outputs toward content-inconsistent colors while preserving the visual fidelity of the released grayscale media. We further introduce Content-aware Color Distributional Distance (CaCDD), a ground-truth-free, content-aware measure of color plausibility derived from semantic color priors, used both as the optimization objective of SCNB and as an evaluation metric. Experiments on ImageNet show that our method remains effective under small perturbation budgets and common post-processing, supporting practical deployment in real-world content-sharing pipelines.