Inferences from Images: Final Report 1984-1987
Abstract
How can a visual system make reliable inferences about surface properties? To answer such questions, one can use computer graphics to generate a realistic rendition of a surface, such as water, clouds or rock. Such a rendition demonstrates that all the relevent psychophysical parameters have been indentified. Given these parameters, or rather their image correlates, we then proceed to show how the surface type can be inferred from the available image information (inverse optics). The surface we have studied most extensively is water, including some of its dynamical properties. Certain aspects of the motion of planar surfaces were also investigated. Our results suggest a simplification of the Cook-Torrance model for the reflection function. Keywords: Vision; Visual perception; Motion; Material type; Water; Computer graphics; Machine vision; Short-long wave interaction; Acoustics; Sound perception; Reflectance function.
Document Details
- Document Type
- Technical Report
- Publication Date
- Dec 01, 1988
- Accession Number
- ADA210710
Entities
People
- Shimon Ullman
- Whitman Richards
Organizations
- Massachusetts Institute of Technology