Research on Digital Restoration of Mural Paintings from Late Tang Tomb M1373 in Xi'an Based on Hyperspectral Analysis and Image Interaction Processing

Tomb murals are an important type of painting culture heritage. The mural in late Tang Dynasty tomb M1373 unearthed in November 2022 in the eastern suburbs of Xi'an City, Shaanxi Province is a well preserved Late Tang Dynasty painting culture heritage. Its colorful murals are of great significance for studying the worship, etiquette, music, clothing, art, and other aspects of the late Tang Dynasty. But due to its old age, environmental impact, and human factor, the pigment layer of this mural still exhibits many typical mural diseases. Especially the discoloration and detachment of mural pigment layer have seriously affected the historical, cultural, and artistic value of the tomb murals. It is urgent to take necessary protection and restoration measures for them. Therefore, before the excavation and interception to the mural, a digital acquisition was adopted for saving the mural information, meanwhile physical reinforcement and cleaning are the main repair methods at present. However, above common physical protection and restoration method is not sufficient for studying and restoring the integrity and authenticity of murals. With the development and application of new digital information technology in various fields, digital recording and analysis method based on hyperspectral imaging can also provide important support for the more diverse protection and more reasonable restoration of mural pigment layers. Based on this, in this study, we takes the mural of late Tang Dynasty tomb M1373 in Xi'an, Shaanxi Province as the research object, and used hyperspectral imaging technology, hyperspectral analysis technology, image processing technology, pseudocolor display technology, and color space theory, man-machine interactive, etc, non-contact non-destructive recognition of mural pigments, virtual restoration of mural pigment colors, and integrity restoration of mural painting patterns are achieved. In particular, a spectral recognition method of mural pigment based on fusion spectral analysis and mixed spectral modeling was used in the real scene, a virtual restoration method for mural pigment colors based on bandpass energy integration, pseudocolor display and color space correction was proposed, meanwhile, a integrity restoration method of painting patterns including draft enhancement and human-computer interaction image processing was used in the real mural color restoration. The experimental results show that the mural uses traditional mineral pigments such as ochre, earth yellow, mineral green and carbon black for its red, yellow, green and black pigments, whose accuracy rate is higher than 97% in simulation experiments with noise of %4 spectral amplitude. The mural restoration method proposed in this article is feasible and effective, and the restoration results significantly improve the richness and completeness of mural painting compared to the original color photos. Especially, the method with color space correction is significantly better than the method without color space correction in terms of texture and color changes. Unlike existing digital restoration methods based on visual images, this method is based on the material properties of hyperspectral images and takes into account visual information, resulting in better authenticity of color and texture restoration, while also considering image integrity. For the physical restoration and activation display utilization of murals, the research results of this article provides more complete and real data reference and support.