Year: 1998
Type: conference-paper
Abstract: We present an approach to the perceptual organization of closed contours into parts that integrates both bottom-up and top-down information into the grouping process. The top-down information is given in the form of a "target", a contour with a pre-computed part structure. Our model gives us the ability to deal with image contour distortions that would render purely bottom-up procedures unsuccessful. It also allows different hierarchical levels in the description of an object to influence perceptual grouping. As output we find not only the part structure of the image contour, but also a correspondence to the target. Moreover, a naturally asymmetric measure of similarity is obtained, as suggested by psychophysical data. We represent the part structure of a contour as a self-matching, a pairing of the contour with itself. It incorporates region and symmetry information into a tree structure that describes the contour parts and their relations. To compute the self-matching for the...
Authors: Ian H. Jermyn, David M. Jacobs, D. Geiger
Language: English
Subfield: Cognitive Neuroscience
Field: Neuroscience
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W2606738379
