Islamic Geometrical Patterns Indexing and Classification Using Discrete Symmetry Groups

摘要

In this article, we propose a general computational model for the extraction of symmetry features of Islamic geometrical patterns' (IGP) images. We describe IGP images using the discrete symmetry groups theory. Our model contains the three following steps. (1) By noting that these patterns fall into three major categories, we begin our indexation process by classifying every pattern into one of these categories. The first pattern category describes all the patterns generated by translation along one direction. Every pattern of this category can be classified into one of the seven Frieze groups. The second type of pattern contains translational symmetries in two independent directions. Patterns of this category can be classified into one of the seventeen Wallpaper groups. The last type, called rosettes, describes patterns which begin at a central point and grow radially outward. We use rosette symmetry groups to classify patterns of this latter category. (2) For every pattern, we extract the symmetry features, namely, the symmetry group and the fundamental region, which is a representative region in the image from which the whole image can be regenerated. But for rosette groups, we can also compute the number of folds. (3) Finally, we describe the fundamental region by a simple color histogram and build the feature vector which is a combination of the symmetry feature (defined in the second step) and histogram information. Experiments show promising results for either IGP images' classification or indexing. Efforts for the subsequent task of classifying Islamic geometrical patterns' images can be significantly reduced.