Peripheral-motion-triggered attention

A model vision system that includes compensation for self-motion, in terms of an explicit conversion of retinal stimuli to an egocentric map, is considered. In particular, an extension of that model, with a hybrid scheme for object position constancy in which spatial operations are carried out in an egocentric map while object recognition is performed in allocentric space, is reported. The concept of stimulus synchronized optic flow for encoding visual motion is introduced. Its utility for preattentive direction of attention is discussed, and preliminary experimental results are presented.<>