Adaptive windows for image processing

Window operators for image processing are described which are trained to achieve desired transforms rather than being stated as mathematical operations. This provides the user with a great deal of freedom in choosing and optimising for specific tasks. Both the theory and some practical results are detailed in order to illustrate the relationship between the structure of alternative systems, their training and eventual performance. The creation of spatial frequency filters and edge detectors are among the examples used to illustrate the process. The nature of several implementations and their cost effectiveness in terms of memory demand is included in the discussion.