This paper shows how an important phenomenon found in biological systems can be modeled by suited dynamics. This phenomenon is the adaptation of the geometrical representation of the information perceived by the visual system. The basic concept of the proposed model are generalized receptive fields. Such receptive fields are characterized by a set of parameters underlying dynamics on different time scales. The performance of the model is demonstrated by comparison to results of neurophysiological experiments. Examples are given how the proposed model can be applied to a complex real-world task in image processing.
展开▼