Using Depth Recovery in Humans
Abstract
This program investigated the relevant information content contained in a physiologically based model of the human visual system with regard to the efficient extraction of depth through stereopsis. A computational system was developed using the Gabor representational scheme to model the spatial weighting functions of simple and complex cells known to exist im primate visual cortex. The algorithm was implemented in a foveated representation produced by a complex-logrithmic, conformal mapping of each image yielding the advantage of both high resolution and a wide field-of-view. The algorithm was tested by extracting depth arrays from random-dot stereograms. The results demonstrated that the depth mapping was accurate and in excellent agreement with the percept produced when human observers fused the same stereo images. The significance of this approach is that it models not a visual system but rather an experimentally verified model of the primate visual system. This is critical to the effective application of human information processing techniques to specialized or intelligent image processing systems. Keywords: Depth perception, Bionics, Gabor functions, Disparity, Space perception.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 07, 1988
- Accession Number
- ADA201278
Entities
People
- Thomas K. Kuyk