Rotation and gray-scale invariant classification of textures improved by spatial distribution of features

Abstract

In this paper, we present a framework for texture descriptors based on spatial distribution of textural features. Our approach is based on the observation that regional properties of textures are well captured by correlations among local texture patterns. The proposed method has been evaluated through experiments using real textures, and has shown significant improvements in recognition rates. © Springer-Verlag Berlin Heidelberg 2005.