As discussed above we wish to maximize the distance of two textures described by non parametric models such as histograms or transition matrices. There are many ways of defining the distance between such representation. In this section we introduce some of these measures and select ones that are adapted for the problem of texture boundary detection for tracking which requires decision making based on few number of observations. A survey of dissimilarity distances and their characteristics can be found in [3] and has inspired the following brief summary.