Local Analysis of Visual Motion

Eero P Simoncelli

Published in:
The Visual Neurosciences
Eds. L M Chalupa and J S Werner
Chapter 109, pages 1616-1623, 2003.
© MIT Press, Cambridge, MA.

We inhabit an ever-changing environment, in which sensing, processing and acting upon these changes can be essential for survival. When we move or when objects in the world move, the visual images projected onto each of our retinas change accordingly. The psychologist J. J. Gibson noted that important environmental information is embedded in this pattern of local retinal image velocities (Gibson, 1950), and thus initiated a scientific quest to understand the mechanisms that might serve to estimate and represent these velocities. Since that time, visual motion perception has been the subject of extensive research in perceptual psychology, visual neurophysiology, and computational theory. This chapter adopts the philosophy that in order to understand visual motion processing in the brain, one should understand the nature of the motion information embedded in the visual world, and the fundamental issues that arise when one attempts to extract that information (Marr, 1977). I develop the most basic computational solution to the motion estimation problem, and examine the qualitative relationship between aspects of this solution and the properties of neurons in the motion pathway.
Download: Preprint | Online Publications