Extracting faded mural patterns based on the combination of spatial-spectral feature of hyperspectral image

Compared with other wall paintings, tomb murals are rarely and have been most seriously damaged by natural weathering, erosion of bacteria and vandalism. Traditional identification methods are time-consuming, laborious, and confined to the visible light range, which mainly relies on expert's experience, analogy and visual color. This calls for new recording methods for mural site protection, relics restoration and repairs. Hyperspectral imaging can capture pictures of spectral range from visible to near infrared light, which render visual enhancement of faded tomb murals. This paper presents a method that can extract faded mural pattern from the hyperspectral images of mural, taking the leaf-like mural patterns in typical Tang dynasty tomb in China as an example. In the method, the faded leaf-like patterns are highlighted through normalization of sensitive spectral bands and extracted by density slice and masking on image space. The result shows that the method can extract the target pattern efficiently with the simple operations.