Style Recognition of Door God Paintings by Hypothesis Testing for Texture Features of Painting Patterns

Abstract

Many studies in the literature have presented multiple remote sensing techniques for defect inspection of paintings. At present, however, papers on defect inspection and restoration of oriental architectural arts—such as door god paintings—are still rare. If an aged and damaged door god painting needs a restoration, then following the style and treatment skill of the original artist as much as possible is important for the restoration. Unfortunately, it is usually difficult to access the original artists for some of the aged door god paintings. This paper considers the texture features of auspicious patterns of armors on warrior door gods as useful information to recognize styles of door god paintings by unknown artists. First, a two-level two-dimensional discrete wavelet transform coupled with co-occurrence matrix calculation was adopted to analyze the texture features, based on the descriptors of angular second moment (ASM), entropy (ENT), contrast (CON), homogeneity (HOM), dissimilarity (DIS), correlation (COR), and cluster tendency (CLU), in the four orientations of 0° (horizontal), 45° (vertical), and 90° and 135° (double diagonal). Second, a two-tailed t-test based on the analyzed texture features was introduced into the hypothesis testing for demonstrating the master and apprentice relationships between the surveyed artists, and for recognizing the door god painting styles of unknown artists as well. The experimental results show that the proposed method effectively describes the texture features of the auspicious patterns of the surveyed door god paintings, and is able to determine the useful co-occurrence features for recognizing unknown artists’ painting styles.