This thesis describes the development of a general in-scene parameter estimation method for quantitative image evaluation. The Maximum Likelihood Ratio (MLR) estimator uses samples from a selected population of known objects in the image to estimate one or more unknown parameters. The estimate is based on statistically matching the population sample residuals to their simulated distribution. The match is characterized by the likelihood ratio function. To compute the likelihood ratio, stochastic simulation is employed to estimate the density of the residuals. The likelihood ratio of the actual residuals and this simulated density is a surface that is then numerically maximized to find the parameter estimate. This in-scene method may be applied to estimating the parameters in many types of aircraft and satellite images. The MLR estimation method is applied to an aerial, thermal infrared heat-loss study to estimate the bias error in the calculation of heat flow. The estimation is shown to substantially improve the prediction of rooftop heat flow for a set of validation structures.

Library of Congress Subject Headings

Infrared imaging--Evaluation; Thermography; Image processing--Evaluation

Publication Date


Document Type


Student Type


Department, Program, or Center

Chester F. Carlson Center for Imaging Science (COS)


Schott, John

Advisor/Committee Member

Anderson, Peter

Advisor/Committee Member

Rhody, Harvey


Note: imported from RIT’s Digital Media Library running on DSpace to RIT Scholar Works. Physical copy available through RIT's The Wallace Library at: TA1570 .S5624 1994


RIT – Main Campus