Stimulus-driven segmentation by Gaussian functions
Shun Ido, S. Arai, Ryo Takamatsu, Makoto Sato · 1996
A new segmentation method called Gaussian segmentation, which can "discover" objects successively in any situation, is presented. The method extracts regions containing locally concentrated stimuli. Similarly, the visual system of humans uses this function to extract the objects from images if no prior information about the objects is available. As a mathematical model, assigning regions as the Gaussian distribution, the extraction of regions in the Gaussian segmentation can be formalized as an optimization problem. The result given by the method coincides with the fact that the extraction, of regions of interest depends naturally on the scale of observation or the visual field.