Guiding Eye Movements For Feature Based Shape Matching

Abstract

We introduce a novel method for shape-based image database search that uses saccadic targeting for local feature choice. A simulated multiresolution sensor is directed toward salient regions of an image in a series of saccadic movements. At each fixation point 1 a region of the retinal image is stored for later matching by correlation. The utility of this approach is demonstrated on an 86 image database.