Title

A Neural Network Representation of a Decentralized Model of the Human Retina

Abstract

The human vision process is presented as a linear, decentralized, continuous time system. This model is then transformed into a representation using neural networks and incorporating non-linearities. First, each cell in the human retina is analyzed to find its specific function. Second, the individual cells are grouped into interconnected systems. Third, a neural network is implemented to model each subsystem. The neural network architecture selected is the Three-Neuron Controller (TNC). Finally, the interconnections are also represented as a neural network, with the nodes being composed of the subsystems. Two major results are presented. First, the overall image quality is improved with the incorporation of neural networks. Second, better edge enhancement is achieved. The edge enhancement is a product of the interconnections between the subsystems.

Department(s)

Electrical and Computer Engineering

Keywords and Phrases

Artificial Intelligence; Neural Networks

Document Type

Article - Conference proceedings

Document Version

Citation

File Type

text

Language(s)

English

Rights

© 2000 American Society of Mechanical Engineers (ASME), All rights reserved.


Share

 
COinS