Extended Symbolic Mining of Textures with Association Rules

The association rules algorithms can be used for describing textures if an appropriate texture representation formalism is used. This representation has several good properties like invariance to global lightness and invariance to rotation. Association rules capture structural and statistical information and are very convenient to identify the structures that occur most frequently and have the most discriminative power. This paper presents the extended textural features which are based on association rules. We extend the basic textural features of our algorithm ArTex in three ways, using (a) various interestingness measures, (b) the multi-resolution approach, and (c) the meta-parameters. Results of our experiments show that the extended representation improves the utility of basic textural features and often gives better results, comparable to standard texture descriptions.