ARO network: a temporal fusion based neural architecture for low level vision

Grouping of stimulus by temporal coherency is a powerful effect for figure/ground segmentation. This may relate to an important aspect of human vision whereby an object image is almost always moving over the retina, but it is seen sharp and still against varying background conditions. This paper investigates the neural basis for this perceptual ability by studying the performance of a neural based system, the adaptive response organizer network (ARO) that abstracts the permissiveness of primate early visual processing. An analysis of the system results demonstrates that the correlated activity of cells may impose temporal organization to promote a process for invariant figural discrimination at the early stages of visual pathways.