Figure 2

Algorithm Flowchart. Data flow throughout the algorithm key blocks. An initial pre-processing step is applied to the image in order to synthesize two images containing only the texture details corresponding to the medium and high frequency content of the original image. In order to compute the proposed texture descriptors, co-occurrence matrices are computed for each synthesized image and several features are extracted from these matrices. Then, multi-scale higher order statistical modeling is applied to extract the proposed texture descriptors. An optional dimensionality reduction of the feature vector can be applied prior to the classification stage.