This work presents an approach for texture based image characterization through topological projections onto the Klein Bottle of small high-contrast regions (patches) extracted from the images. Several configurations of cut-off frequency were analyzed in order to reduce the vector size of features and to increase accuracy. Experiments using the proposed method for texture classification, on several established datasets, show that the proposed method not only manages to reduce feature vector size, but also improves correct classification rates when compared to other state-of-the-art methods.
展开▼
