Introduction to Vision Science
The goal of Vision Science is to understand how information processing systems
acquire, represent, and process information carried by light. As a discipline,
Vision Science integrates three basic paradigms. These are a) the measurement
and modeling of visual performance in humans and other animals (Psychology),
b) the search for ways to build artificial visual systems (Engineering/Computer
Science), and c) the characterization of the neural mechanisms that implement
biological visual systems (Biological Sciences/Neuroscience). Historically,
the strongest link has been between psychologists and neuroscientists, who
share not only the goal of understanding how vision works but also the goal
of establishing direct links between behavior and neural mechanisms. For
examples see Spillman and Werner (1989), Lee and Valberg (1991) and Wandell
(1995). Over the past decade, however, there has been an increased awareness
by both psychologists and engineers that understanding the compuational
problem of vision clarifies the study of human performance and that models
of human performance provide guidance about how to construct artificial
visual systems. For examples see Landy and Movshon (1991). A similar link
exists between engineers and neuroscientists. For example, many connectionist
algorithms are based on descriptions of neural processing in the mammalian
visual pathways.
A number of factors make vision an excellent model system on which to focus
the study of information processing. First, it can be estimated that roughly
80% of all of the environmental information that humans use in the conduct
of their daily affairs is acquired visually. Correspondingly, the majority
of the neocortex is dedicated to the analysis and elaboration of visual
information. For this reason alone, if we understand vision, we will have
gone a long way towards understanding human information processing. Second,
visual processing probably taps lower level processes than those that mediate
reasoning, problem solving, or language; the study of vision may prove more
tractable than the study of higher level information processing problems.
Certainly, the hope is that experimental methods, modeling techniques, and
fundamental results worked out for the visual system may provide useful
guidance for the study of higher level processes. Third, there are a number
of immediate applications for models of human vision and for artificial
visual systems. These include quality metrics for image reproduction, robot
guidance, information display, photorealistic computer graphics, light measurement
instrumentation, automated inspection, medical image processing, biologically
motivated signal processing, and techniques for curing eye diseases and
ameliorating the consequences of low vision. Finally, Vision Science as
an enterprise is large and active: dozens of journals report the results
of vision research and a range of different organizations offer monetary
support for the study of vision. For all of these reasons, Vision Science
is well-represented in most emerging Cognitive Science programs. Indeed,
Marr's (1982) influential book on vision is often used to introduce students
to the philosophy and approach of Cognitive Science.
References
Landy, M. S. and J. A. Movshon, Eds. (1991). Computational Models of
Visual Processing. Cambridge, MA, MIT Press.
Lee, B. B. and A. Valberg, Eds. (1991). From Pigments to Perception.
New York, Plenum Press.
Marr, D. (1982). Vision: A computational investigation into the human
representation and processing of visual information. San Francisco,
W. H. Freeman.
Spillmann, L. and J. Werner, Eds. (1989). Visual Perception: The Neurophysiological
Foundations. New York, Academic Press.
Wandell, B (1995). Foundations of Vision. Sunderland, MA, Sinauer.
Author: David Brainard, brainard@psych.ucsb.edu
Last Modified: 25 November 1995
Last Modified by:David Brainard