Thesis - University Access Only
Master of Science (MS)
Department / School
Images of the Earth's surface acquired from long distances though the atmosphere are degraded due to atmospheric scattering and absorption. The image degradation can be modeled by an atmospheric point spread function (PSF}. There are many parameters that may affect the PSF. To restore noise-blurred images, one must understand which parameters influence the PSF and to what degree. This is very important for scientific applications which seek to extract information about environmental systems. One of the areas where neural networks are often applied is imaging processing. In this project, a neural network based system has been designed and implemented to study the relationship between the point spread function and physical properties of the atmosphere, and to extract the atmospheric PSF directly from the simulated imagery with minimal a-priori information. We use the divide-and- conquer technique to break a large neural network into smaller ones; each is processed by a PC or workstation at the same time. This neural network model can be mapped into a parallel system to achieve the maximum speed-up. This project is directly linked to the existing NSF-EPSCoR project "Imaging and Modeling of Coupled Environmental Processes" (IMCEP) at South Dakota State University. The results obtained indicate that the neural network approach is effective in estimating the PSF, and is an excellent alternative to the other existing methods.
Library of Congress Subject Headings
Image processing -- Digital techniques
Neutral networks (Computer science)
Number of Pages
South Dakota State University
Wu, Limin, "Estimation of the Atmospheric Point Spread Function using a Collection of Neural Networks" (1997). Electronic Theses and Dissertations. 331.