Computational Framework for Symmetry Classification of Repetitive Patterns

Abstract

The interest shown recently in the algorithmic treatment of symmetries, also known as Computational Symmetry, covers several application areas among which textile and tiles design are some of the most significant. Designers make new creations based on the symmetric repetition of motifs that they find in databases of wallpaper images. The existing methods for dealing with these images have several drawbacks because the use of heuristics achieves low recovery rates when images exhibit imperfections due to the fabrication or the handmade process.

To solve this problem we propose a novel computational framework based on obtaining a continuous-value symmetry feature vector and classifying it using an NN classifier and a set of prototype classes. The prototype parameters are automatically adjusted in order to adapt them to the image variability. Moreover, a goodness-of-fit classification can be applied in the content-based image retrieval context. Our experimental results improve the state of the art in wallpaper classification methods.