Visual Perception of Structure from Motion
Abstract
The research performed in this project has examined the abilities of human observers to perceive 3D form from different types of optical structure within moving or stationary visual images. The research has been organized into four general problem areas, including the low level detection of coherent, motion, the analysis of 3D form from motion, the analysis of image shading and texture, and the identification of image contours. Our basic strategy in all of these areas has been to identify the key assumptions of current computational models; to test the psychological validity of those assumptions using appropriate psychophysical procedures; and, based on the results of those experiments, to develop alternative models that more closely match the perceptual capabilities of actual human observers. In contrast to most common methods of 3D image analysis, which are designed to compute precise metrical descriptions, our results have shown that human perception is primarily concerned with more abstract aspects of object structure, such as affine or ordinal properties, which are easier to computer and are more robust to uncontrolled changes in viewing conditions.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 30, 1992
- Accession Number
- ADA253235
Entities
People
- James T. Todd
Organizations
- Boston University