Method of predicting the probability of human observers recognizing targets in simulated thermal images

Abstract

A series of psychophysical trials is conducted to study the human perception of images of various shapes that are degraded by various combinations of Gaussian blurring, sampling, and fixed pattern noise. These images were created using computer programs designed to simulate the types of image degradation associated with thermal imagers that employ focal plane arrays (FPAs) of detectors. A total of 46 observers participated in these trials, during which over 130,000 observations were recorded. The responses collected during the trials are used to develop an empirical model of human vision applicable to 2-D images having a gray scale, but not including other colors. This model can be used to predict the probability of a human observer correctly distinguishing between images of two different objects.

Proposed performance measures for imaging systems with discrete detector arrays

“…The proposed performance measure arose from the work of Aulhorn and coworkers [14] and our previous work [6,9]. It is similar in some respects to the 'equivalent disc model' of van Meeteren [12] and has some connection with the 'triangle orientation discrimination' (TOD) measurement proposed by Bijl and Valeton [15].…”

“…There will be maximum and minimum values [6], but for recognition purposes it may be suae cient to obtain average values. A measurement made using a moving target will under estimate the eå ect of xed pattern noise on a static image.…”

“…This fact and the importance of detecting areas [6] suggest that a performance measure might be based on the ability of an observer using an imaging system to detect circles of varying sizes. Similar suggestions have been made previously by Thomas [11] and van Meeteren [12].…”

“…Aliasing can also be removed using anti-aliasing lters which eliminate spatial frequencies above the Nyquist frequency. This produces a cosmetically better looking image but also removes information and makes objects in the scene more diae cult to recognize [6]. So the brain is able to use aliased information in an image even though a measurement of MTF is no longer valid.…”

Proposed performance measures for imaging systems with discrete detector arrays

“…Aliasing can also be removed using anti-aliasing filters which eliminate spatial frequencies above the Nyquist frequency. This produces a cosmetically better looking image but also removes information and makes objects in the scene more difficult to recognize [6]. So the brain is able to use aliased information in an image even though a measurement of M T F is no longer valid.…”