The proposed project involves psychophysical investigations of the mechanisms underlying the recovery of surface structure in stereopsis and motion perception. Virtually all of the work in stereopsis and motion in the past century has focused on the matching or correspondence problem. The tacit assumption has been that an understanding of matching is both necessary and sufficient for understanding the recovery of 3D structure from 2D images. The proposed work will pursue experiments into two forms of image segmentation that reveal the insufficiency of the view. Specifically, we propose experiments to evaluate how multiple views are segmented into distinct objects along occluding boundaries, and into distinct layers when computing surface lightness and/or opacity. The broad goal of the proposed project is to fill the current gap in our understanding of the mechanisms underlying how the human visual system uses multiple views of images to decompose 2 D images into a 3D representation of surfaces in depth.
The specific aims of the project are: 1) To explore the psychophysical properties that cause binocular and moving displays to be decomposed into matchable and unmatchable features, critical for understanding how surface boundaries are recovered in these domains. 2) To understand how image regions are decomposed into a multi-layered, 3D representation of surface lightness and transparency. 3) To develop a theoretical framework that integrates these two forms of image segmentation. The experiments entail a number of different methods that are tailored to the specific question under study. In general, two types of methods will be used. One method will measure perceptual sensitivity by requiring observers to discriminate (typically two) displays. Sensitivity to the property on interest (such as surface boundaries) is captured in an observer's ability to perform the discrimination. A second method requires observers to manipulate a graphics display to match the perceptual quality of interest (such as depth, orientation, or three-dimensional shape of a pattern). These measurements will be analyzed to determine how surface boundaries are recovered and surface properties such as lightness and transparency is attributed to moving and/or stereoscopic displays.

Agency
National Institute of Health (NIH)
Institute
National Eye Institute (NEI)
Type
First Independent Research Support & Transition (FIRST) Awards (R29)
Project #
5R29EY011487-02
Application #
2888515
Study Section
Visual Sciences B Study Section (VISB)
Project Start
1998-09-30
Project End
2003-09-29
Budget Start
1999-09-30
Budget End
2000-09-29
Support Year
2
Fiscal Year
1999
Total Cost
Indirect Cost
Name
Massachusetts Institute of Technology
Department
Miscellaneous
Type
Other Domestic Higher Education
DUNS #
City
Cambridge
State
MA
Country
United States
Zip Code
02139
Anderson, B L (2001) Contrasting theories of White's illusion. Perception 30:1499-501
van Ee, R; Anderson, B L; Farid, H (2001) Occlusion junctions do not improve stereoacuity. Spat Vis 15:45-59