Semantic Color Naturalness Breaker: Preventing Illegitimate Colorization via Content-Aware Color Priors
TL;DR
- Grayscale images disrupt unauthorized colorization: Adds small perturbations while preserving the released image’s visual quality.
- CaCDD measures content-aware color plausibility: Uses semantic color priors to assess whether generated colors are natural.
- Protection withstands common post-processing: Evaluates modest perturbation constraints and image transformations.
Type
Publication
2026 IEEE International Conference on Systems, Man, and Cybernetics (SMC 2026)
Presentation Video
Presentation Slides
