Exploring the Restoration Patterns of Color in Tang and Song Dynasties Paintings: Luohan Laundering as an Example
Wang Xiaosong1, Wang Jingchao1, Li Ziru2, Huang Tengyu1, Chen Xiaojiao1
1.Laboratory of Art and Archaeology Image, Zhejiang University, Hangzhou 310028, China 2.Institute for Advanced Studies of Images and History, Guangzhou Academy of Fine Arts,Guangzhou 510006, China
Abstract：The color restoration of Tang and Song Dynasties paintings holds significant importance for the research and presentation of traditional Chinese culture. However, current color restoration methods each have their limitations. Restoration methods guided by subjective knowledge tend to be time-consuming and less easily reversible, while methods guided by objective data may face challenges in result stability and in the evaluation of the finished work. Under existing technical conditions, achieving absolutely perfect restoration remains impossible. However, based on long-term research and practice, this study suggests that a pattern governing Tang and Song Dynasties painting color restoration can be systematically analyzed and traced.For this research, the primary visual material used is a high-resolution photographic image of Luohan Laundering. This serves as the unretouched reference image. The corresponding restored digital image from A Comprehensive Collection of Ancient Chinese Paintings is used as the restored version. This study proposes the following hypotheses. Hypothesis 1: The color difference between unrestored and restored paintings is a constant numerical value. Hypothesis 2: If the color difference is not a constant, then at the very least, there exists a statistical correlation between corresponding color values in unrestored and restored paintings. This correlation may vary depending on the type of color and could potentially be described using a mathematical model. Hypothesis 3: For a given color, the numerical values of corresponding color regions in unrestored and restored paintings exhibit a systematic correlation, and the restored color can be directly computed.Through empirical research, this study first refuted Hypothesis 1. The results of color difference (ΔE) calculations indicated significant fluctuations in the color difference at corresponding positions before and after restoration, rather than a constant value. Second, the results of the correlation analysis demonstrated a statistically significant relationship between the unrestored and restored colors. However, regression analysis revealed that no statistically viable mathematical model could be constructed. Consequently, it is not possible to derive the restored color values directly through computational methods based on the unrestored colors. Therefore, Hypothesis 2 was only partially supported. The research speculates that this outcome may be due to differences in the chemical properties of various pigments. Even under identical preservation conditions, the degree of fading and discoloration may vary across different pigment types, resulting in considerable fluctuations in color differences. To further explore these findings, this study proceeded to test Hypothesis 3, which posited that “the color values of the same color at corresponding positions in the unrestored and restored paintings exhibit a correlation, and the restored color can be directly obtained through computation”. The results, however, refuted Hypothesis 3. Even when analyzing a single-color category, no statistically viable regression relationship could be established between the unrestored and restored versions of the painting. This finding demonstrates that the idea of deriving a mathematical relationship for color restoration based solely on comparing pre- and post-restoration color values is currently infeasible. While the data obtained in this study may provide insights for refining existing restoration methods, practical applications will still need to rely on well-established restoration techniques.The process of color restoration is inherently complex, encompassing chemical, historical, and artistic knowledge that may not be easily expressed through a simple mathematical model. Using Luohan Laundering as a case study, this research explores the mathematical relationship between colors in unrestored and restored paintings from the Tang and Song Dynasties through an empirical approach. The findings reveal significant correlations between corresponding color values before and after restoration in both the RGB and Lab color models. However, due to the inconsistency of color shifts, constructing a fully reliable mathematical model remains a challenge. The conclusions and findings of this study could contribute to the refinement of data-driven restoration methods, aiding in the development of restoration databases and algorithmic models to enhance the precision and efficiency of color restoration.
王小松, 王鲸超, 李子儒, 黄腾宇, 陈晓皎. 唐宋绘画的色彩复原规律探索[J]. 浙江大学学报(人文社会科学版), 2025, 55(8): 63-73.
Wang Xiaosong, Wang Jingchao, Li Ziru, Huang Tengyu, Chen Xiaojiao. Exploring the Restoration Patterns of Color in Tang and Song Dynasties Paintings: Luohan Laundering as an Example. JOURNAL OF ZHEJIANG UNIVERSITY, 2025, 55(8): 63-73.
1 陈云海：《唐宋绘画的主体性及其视觉逻辑》，《艺术设计研究》2021年第3期，第11-14页。 2 刘舜强：《古书画损毁机理初探》，《文物保护与考古科学》2003年第1期，第39-42页。 3 龚建培：《中国传统矿物颜料、染色方法及应用前景初探》，《南京艺术学院学报（美术与设计版）》2003年第4期，第80-84页。 4 鲁欣月、何璐、赵军龙：《梵高画作褪色的秘密——颜料中的化学》，《大学化学》2021年第10期，第73-80页。 5 付心仪、李岩、孙志军等：《敦煌莫高窟烟熏壁画的数字化色彩复原研究》，《敦煌研究》2021年第1期，第137-147页。 6 Chen X.， Liu Q. & Chen Y. et al.， “ColorNetVis： an interactive color network analysis system for exploring the color composition of traditional chinese painting，” IEEE Transactions on Visualization and Computer Graphics，Vol. 30， No. 6 （2024）， pp. 2916-2928. 7 Wu T.， Li G. & Yang Z. et al.， “Shortwave infrared imaging spectroscopy for analysis of ancient paintings，” Applied Spectroscopy， Vol. 71， No. 5 （2017）， pp. 977-987. 8 Hu Q.， Peng X. & Li T. et al.， “ConvSRGAN： super-resolution inpainting of traditional chinese paintings，” Heritage Science， Vol. 12， No. 1 （2024）， https：//doi.org/10.21203/rs.3.rs-3940761/v1. 9 赵磊、吉柏言、邢卫等：《基于多路编码器和双重注意力的古画修复算法》，《计算机研究与发展》2023年第12期，第2814-2831页。 10 Jia M.， Hu J. & Yang Z. et al.， “Automatic restoration of Dunhuang murals and process visualization method based on deep learning，” Applied Sciences， Vol. 15， No. 3 （2025）， https：//doi.org/10.3390/app15031422. 11 刘映廷：《台湾传统建筑装饰修复研究——以壁画为例》，《南京艺术学院学报（美术与设计）》2019年第1期，第149-151页。 12 何姗、周华、吕淑玲：《浅说清三代酱釉青花碗在研究修复中的色彩复原方法》，见中国文物学会文物修复专业委员会编：《文物修复研究（2017—2018）》，北京：中国文联出版社，2021年，第246-264页。 13 魏宝刚、潘云鹤、华忠：《基于类比的壁画色彩虚拟复原》，《计算机研究与发展》1999年第11期，第1364-1368页。 14 赵男：《基于图像处理技术的中国画色彩修复系统设计》，《现代电子技术》2020年第17期，第60-68页。 15 马伟、龙晴晴、秦悦等：《基于画作线条结构分解的高清古画修复》，《计算机辅助设计与图形学学报》2018年第9期，第1652-1661页。 16 朱欣娟、雷倩、吴晓军：《基于生成式对抗网络的文物图像超分辨率重建及色彩修复》，《西安工程大学学报》2021年第3期，第86-92页。 17 梁金星、万晓霞、孙志军等：《敦煌壁画颜料颜色数据库构建方法》，《敦煌研究》2017年第1期，第132-140页。 18 赵磊、林思寰、林志洁等：《中国古画渐进式多级特征修复算法》，《计算机辅助设计与图形学学报》2023年第7期，第1040-1051页。
2025, Vol. 55 
