Center for Cognitive Sciences Institute for Neurophysics University of Bremen

Segmentation is one of the most basic tasks in sensory systems. Segmentation groups incoming sensory data into only a few meaningful chunks for efficient later processing.

Two basic concepts of grouping can be utilized: the detection of segments enclosed by strong borders, and the grouping of data according to the most prominent signal variations present.

In the segmentation ansatz presented here, both concepts, existing on different spatial scales, are represented within a scalespace representation and connected via synchronized activity of neural pools.

Further infos:

© 1994-2003 - all rights reserved.