Image Warping and Homogenization to Analyse the Frequency und Amplitude Modulation of Structural-Statistical Textures. Technical Report IES-2014-10

Research is being conducted in the field of texture analysis for many years. In this process a wide variety of methods have been developed for analyzing and describing the various structural and statistical textures. For the complicated mixed case of structural-statistical textures only few methods exist which allow a general evaluation. One way to describe this texture type is to produce a deviation from a structural texture with the help of distortions and gray value differences. Inspired by the communication technology, this can be done with the help of frequency or amplitude modulations. In this report, an evaluation option for each type of modulation will be presented. Occurring geometric distortions can be described for example with the help of image warping algorithms. In this context, the retrieval of characteristic points of the texture primitive, i.e. recurrent points that are easy to detect, is an important and also challenging task. The detection of irregularities of the gray values over the entire texture can be achieved with the help of a image pre-processing method called homogenization. Thus, the combination of the two methods - image warping and homogenization - allows a complete description of the frequency and amplitude modulation of structural textures and provides a new way for the analysis of structural-statistical textures.