In the research and investigation on the historical materials and cultural properties, not only conventional RGB digital images but also spectral reflectance images serve valuable information. However, some suitable techniques to extract effective information corresponding to the purpose of the use of the spectral reflectance images are desired because the spectral reflectance images have so huge of information.
In this article, a method for investigating the historical materials and cultural properties will be described. This method can extract useful features relating to color compositions in the cultural properties. In this method, the measured spectral reflectance images of the icons, which are examples of the cultural properties, are clustered and the principal component analysis is applied in each cluster. The first principal component in each cluster is used for approximating the original spectral reflectance. The color difference caused of this approximation is calculated and used to control features to be extracted from the icons. Based on this number of clusters, results of the clustering are compared and possibilities of this method for further investigation are discussed.
View full abstract