Computer Models of Retinal Function.
Abstract
The vertebrate retina consists of a series of sheets of neurons each containing a patterned representation, or neural image corresponding to the incoming visual scene. Most of the operations can be described as convolutions implementing small kernels involving local neighborhoods. These patterns become more complex as the visual signal passes through the retina as each new convolution builds upon previous ones. Because we have access only to single cells, these patterns of activity have never been recorded. As a first step in visualizing these patterns we have attempted to simulate them using modern image processors operating in real time, that like the retina invoke local convolutions, in an effort to understand the mechanisms, and the physiological roles of each of these biological image processing operations. In addition, we have now developed methods for measuring these patterns in living retinas. The comparison between the modeled and measured patterns will show us what components of retinal function have not yet been incorporated into our understanding, as expressed by the models. These studies will help us to understand biological image processing by forcing us to assemble the retina using well-defined physiological functional building blocks, derived from retinal function. A long-term objective of this research is the implementation of biological image processing algorithms for man made systems such as tracking, target identification, guidance and navigation. (AN)
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 30, 1994
- Accession Number
- ADA290156
Entities
People
- Frank Werblin
Organizations
- University of California, Berkeley