PRevivor: Reviving Ancient Chinese Paintings using Prior-Guided Color Transformers

2025-11-03

AI Summary

Ancient Chinese paintings suffer severe color fading due to pigment degradation, and the scarcity of high-quality annotated datasets hinders end-to-end digital color restoration. To address this, we propose PRevivor, a prior-guided Color Transformer model that—uniquely—leverages Ming–Qing painting styles as stylistic priors for reconstructing the original colors of Tang–Song masterpieces. Our method decouples restoration into two stages: luminance enhancement and chromaticity correction. A dual-branch query module, guided by local chromatic priors, balances fine-grained detail refinement with global semantic reasoning. We integrate dual variational U-Nets, multi-scale feature mapping, and a Transformer backbone. Quantitative and qualitative evaluations demonstrate that PRevivor significantly outperforms existing colorization methods, achieving superior reconstruction of both color distributions and brushstroke textures.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Cognitive Modeling & Cognitive Systems: Computational Creativity
• Machine Learning: Transparent, Interpretable, Explainable ML

Application Category

• Web Mining and Content Analysis: Large pretrained models with web data
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Security and Privacy: Data transparency and provenance

Abstract

Ancient Chinese paintings are a valuable cultural heritage that is damaged by irreversible color degradation. Reviving color-degraded paintings is extraordinarily difficult due to the complex chemistry mechanism. Progress is further slowed by the lack of comprehensive, high-quality datasets, which hampers the creation of end-to-end digital restoration tools. To revive colors, we propose PRevivor, a prior-guided color transformer that learns from recent paintings (e.g., Ming and Qing Dynasty) to restore ancient ones (e.g., Tang and Song Dynasty). To develop PRevivor, we decompose color restoration into two sequential sub-tasks: luminance enhancement and hue correction. For luminance enhancement, we employ two variational U-Nets and a multi-scale mapping module to translate faded luminance into restored counterparts. For hue correction, we design a dual-branch color query module guided by localized hue priors extracted from faded paintings. Specifically, one branch focuses attention on regions guided by masked priors, enforcing localized hue correction, whereas the other branch remains unconstrained to maintain a global reasoning capability. To evaluate PRevivor, we conduct extensive experiments against state-of-the-art colorization methods. The results demonstrate superior performance both quantitatively and qualitatively.

Problem

• Addressing irreversible color degradation in ancient Chinese paintings
• Overcoming dataset limitations for digital restoration tools
• Developing prior-guided transformers for luminance and hue correction

Innovation

• Prior-guided color transformer for ancient painting restoration
• Variational U-Nets with multi-scale mapping for luminance enhancement
• Dual-branch color query module guided by hue priors