Automatic visual inspection plays a growing role in today's efforts to reduce the costs of industrial production. Three to ten months of man power have to be invested for the development of a typical industrial application. In this context, a reduction of time and thus of costs will make it easier to introduce new systems. We show a solution by using several algorithms for two kinds of selection routines, feature selection and training pattern selection. This method offers a convenient way to easily adapt a given system to a special texture analysis task.