A Model for Visual Perception and Recognition of Outlines

Abstract

Many topics in vision are still not well understood. A model for visual perception and recognition of shapes is described, by combining ideas relating to receptive fields of ganglion cells in the retina, eye movements during fixation, multiple spatial scales of precision, attention, axes and coordinates. The main objective is to provide a realistic scheme for perception of outlines, with recognition that is not prevented by changes of size and viewpoint. Outlines are detected by changes with time arising from eye movements, followed by identification of vertical and horizontal axes, and use of levels of precision that depend on the size of the area of attention. The increments of position in these levels increase in size as the size of area increases, so in an index of positions the increment numbers remain constant for all sizes. After changes of viewpoint, "attentional planes" connect changed positions on outlines to positions with a standard orientation of axes, so that shapes may be matched with stored representations. Spatial coordinates are related to positions at the retina by representations based on eye movements.