Dongming LU, China | ISBN: | Submitted: Jun 24, 1998

Color Restoration Techniques for Faded Murals of Mogao Grotto

*) This work is supported by National Natural Science Foundation Key Project of China.

The Chinese civilization with thousands of years has accumulated many heritage sites. Mogao Grotto, located in Dunhuang Gangsu province northwest of China, is the typical one of them, which is especially famous for its murals with thousand-year's history. However, they are being seriously damaged by many causes. Scientists and artists have been worked for a long time on protection and restoration. The traditional methods are time-consuming and ineffective. Our work is just to employ computer-aided methods to facilitate research, i.e., aiding artificial mural imitating, checking pigment color fading laws under various environments as well as showing virtual changing process of mural over different ages.

In order to realize above goals, we are undertaking several technical approaches: (1) Domain knowledge based method, to formalize the pigments color fading laws under different environments; (2) Rule based method, to setup artists/experts experience models about mural color fading;; (3) Analogy based reason, to create analogical relationships among murals based on their age, topic and cave. In this paper, we will focus on the first approach. The main idea for restoring a color faded mural is as follows:

1. Acquire the color structure of a mural by analyzing the color histogram; then, separate the mural image into several level sothat each of them represents the color of one pure pigment.

2. Determine the original pigment used for each color leval with pigment domain knowledge.

3. Restore every color levels of the mural image respectively according to the color fading laws over time of pigments under different light, tempreture as well as moisture.

4. Combine the restored color levels again into one image to so as to obtain the color restored mural image.

In the above restoring process, we have dealed with some key issues:

1. Mural representation. Besides text description, i.e. cave no, dynasty and topics, we propose a image model with multiple levels, each of which represents the color of one pure pigment. 2. Domain knowledge formalization. For commonsense of pigments, we maintain a standard color table for each pigments and create the relationships among pigments, cave, dynasty and topics; For pigments fading laws, we design a series of precedures based the changing law.

3. Technical implementation. Color histograms are used to separate different image levels, while commonsense is exploited to determine pigments and choose the proper procedures for restoration.

Presnetly, we have implemented a mural color restoration system based on pigment knowledge. By experimenting with typical faded murals, it is proved to be quite effective. The included figure is an example of experiments. It presents the color fading process of the background color according to the color fading rule. where E is curent apperance of the mural, A is its virtual apperance 1000 years ago in Shong Dynasty, B is its virtual apperance 600 years ago, C is its virtual apperance 400 years ago,D is its virtual apperance 200 years ago, and F will be the future appreance after about 200 years. Now, besides improving this method, we will begin to investigate the second and third approach, and finally to implement a more effective system by integrating three methods.

[more information]

[other authors]
Xiangyang LI,Baogang Wei,Yunhe PAN

[keywords]
Heritage Restoration, Virtual Reality, Knowledge Processing, Mogao Grotto