The first portion of this study checked water, vegetation, and urban class features of LANDSAT TM data for univariate normality using Pearson's system of frequency curves. Results indicated that of the 144 image bands tested 135 were determined to be normal in distribution. The second part of the study developed an image generator that uses the mean, covariance matrix and intraband correlation of LANDSAT TM images to create synthetic class scenes. Imagery composed of multiple synthetic class scenes, which ranged from normal to non-normal in their distributions, were classified using a maximum likelihood classifier. No significant difference in classification accuracy was found between the normally distributed data and the non-normal image data.
Library of Congress Subject Headings
Remote sensing--Data processing; Imaging systems--Image quality--Data processing; Remote sensing--Mathematics; Multivariate analysis
Department, Program, or Center
Chester F. Carlson Center for Imaging Science (COS)
Frey, Elizabeth, "An Examination of distributional assumptions in landsat tm imagery" (1995). Thesis. Rochester Institute of Technology. Accessed from
RIT – Main Campus