Texel: a methodology and an integrated environment for developing automated systems for texture classification

The techniques used when building a texture classification system are determined, in most cases, ad hoc. For developing such a system, the most commonly used sequence of steps is: acquisition, preprocessing, feature extraction and classification. For each of them, there exist many options, and the choice of one or more of them influences the rest. The expert has to choose among several kinds of preprocessing tasks, a number of texture analysis methods, and several classifiers. The process of getting a system that provides a good performance is a costly procedure due to the great variety of combinations of processing tasks. In this paper we introduce a methodology for the construction of texture classification systems, primarily focused in natural, random or pseudorandom textures, and a visual tool with a powerful interface that allows us to put it into practice.