Role of Retinocortical Processing in Spatial Vision
Abstract
Several important image-processing functions have been proposed for the geometric distortion known as cortical (or more precisely, retinocortical) magnification. This spatial distortion can convert the radial velocities projected on the retina by egocentric motion into uniform, rectilinear motion at the cortex. It can also convert changes of size and orientation in retinal coordinates into mere translation at the cortex. (In both cases, an image-like property is converted into a map-like property.) Thus cortical magnification must play an essential role in forming our stable precepts of the world around us, even as it vexes the question of how the information from different fixations within the same scene can be arranged into a single precept. Is the cortical image subsequently undistorted, just to facilitate the superposition of multiple fixations? An understanding of the image-coding functions of the primary visual cortex (VI) should help to unravel this paradox. Using the tools of computer vision (LISP algorithms developed on Symbolics networks), we are attempting to build a working model that includes such processes as: fixational eye movements, retinal filtering and inhomogeneity, retinocortical mapping, cortical (Gabor) image-coding, and other processes involved in the coordinate shifts needed for mapping purposes.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1988
- Accession Number
- ADA200198
Entities
People
- Donald H. Kelly
Organizations
- SRI International