Texture Classification and Retrieval

Matlab package of tools for texture image classification and retrieval based on a local spatial frequency analysis method. This method can be catalogued among other statistical methods for texture classification and retrieval and supply tools for feature extraction (FE) and learning through a similarity measure (SM). The FE step consists in finding the<br>discrete spatial frequency histograms of multi-scale texture images, by means of a discrete 1-D pseudo-Wigner<br>distribution (PWD), directed upon the maximum angular activity of the von Mises distribution of image entropy. The learning process (LP) step compare the space-frequency histograms by a measure based on the Sørensen–Dice index. The LP refines the spatial-frequency histograms of texture models over a massive (several hundred) set of texture images. The philosophy of the retrieval operation consists in searching the best figure for the similarity measure, among a set of model-histograms vs. a given query histogram.