I. Much worse than a camera A. The eye has a ``blind spot'' (which corresponds to the optic nerve); the corresponding portion of the visual field is blank spot, into which objects can be made to disappear. B. The eye has only limited temporal and spatial resolution. The image reaching the retina goes through a lens (made of protoplasm and so not rigid) and the actual transduction occurs via irregularly distributed cellular processes. C. Images formed by the eye are constantly subjected to movement, including, among others, 1. induced by head and body movements 2. saccades (small ``flicks'' in the direction of gaze), around 3/sec. 3. ocular microtremor (OMT), around 90/sec II Much better than a camera A. Overlaps different detection strategies: 1. foveal for fine detail and color (the fovea is a region of the retina in which the color-sensitive cells are concentrated) 2. peripheral for movement and light detection (falling meteors and dim stars are more easily seen ``from the corner of the eye'') 3. color vs. form brain region activity show multiple representations 4. includes output formating for memory and symbolic routines B. Object-oriented (blind-spot is filled in through object dynamics) C. Can achieve hyperacuity (resolution exceeds that possible with the passive use of sensors, i.e., below the cell-spacing) D. Can determine color even under sharply varying illumination III How to test and explore? A. What to study? E.g., are hyperacuity and texture detection related to the high-speed, low amplitude eye movement OMT? Why is color stable? B. How to observe patterns in CNS? C. What can we learn from clinical studies? D. Psychophysical experiments - e.g., the Butterfield effect (color on a black-and-white-only TV) E. Noninvasive neuroimaging F. Mathematical simulations IV Relation between vision and cognition (Pylyshyn, Brain and Behavioral Sci) V AI, selectionist, connectionist and neural network models VI Comparison of various depth perception channels VII Design of ``terminal of the future''
pck; 1-16-01