Low-level grouping mechanisms for contour completion

The human visual system achieves effective segmentation in situations where many computational methods fail. This study seeks to demonstrate how methods based on low-level models of early visual processing can be adapted to perform this task in the difficult domain of muscle histology. Through analy...

Full description

Saved in:
Bibliographic Details
Published in:Information sciences 2000, Vol.125 (1), p.19-35
Main Authors: Todman, Alison G, Claridge, Ela
Format: Article
Language:English
Subjects:
Amodal completion
Cells
Closure
Contrast polarity
Image analysis
Mathematical models
Muscle
Muscle cells
Neural mechanisms
Neurology
Perceptual grouping
Psychophysiology
Segmentation
Vision
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The human visual system achieves effective segmentation in situations where many computational methods fail. This study seeks to demonstrate how methods based on low-level models of early visual processing can be adapted to perform this task in the difficult domain of muscle histology. Through analysis of typical images of muscle tissue we have identified basic features likely to be of importance in human perception. Considering these in the light of current neurological and psychophysical research, we have concluded that the manner of boundary completion is, in this case, akin to that underlying amodal completion of occluded surfaces in normal situations. The issue of contrast polarity is discussed, and a model for the completion of fragmented line figures, incorporating both polarity–sensitivity and the concept of partial closure, is presented. This has been adapted into a computational system, aimed at the segmentation of muscle cell images, that attempts to exploit useful features of low-level processing in combination with knowledge of the domain.
ISSN:0020-0255
1872-6291
DOI:10.1016/S0020-0255(99)00147-4