Hierarchic texture classification using statistical steganography techniques

A novel method for adaptively selecting texture features is presented. We apply statistical steganography techniques with searching for an optimal set of binary masks to extract texture features and provide the best discrimination of texture images. The extracted texture features are robust to noise attacks. Moreover, a tree structure containing the selected set of masks has been set up for classification. Experiments show that the proposed method can achieve high classification rate and also work well in a noise environment.