Persistent Link:
http://hdl.handle.net/10150/188006
Title:
VISUAL PERCEPTION IN CORRELATED NOISE (MODELS).
Author:
MYERS, KYLE JEAN.
Issue Date:
1985
Publisher:
The University of Arizona.
Rights:
Copyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.
Abstract:
This dissertation concerns the ability of human observers to perform detection tasks in medical images that contain structured noise. We shall show that physical measures of image quality, such as signal-to-noise ratio (SNR), resolution, modulation transfer function (MTF), and contrast, do not accurately predict how well an observer can detect lesions in an image. We have found that for images with equal pixel SNR, humans can detect a low contrast object more readily in images that have a low-pass noise structure, as opposed to a high-pass noise structure. This finding is important in the comparison of images generated by a classical pinhole imaging system with images generated by a computed tomography imager. We would like to have a figure of merit that accurately predicts a physician's ability to perform perceptual tasks. That is, we want a figure of merit for imaging systems that is more than an evaluation of the physician's performance, measured using human observers and an accepted method such as receiver operating characteristic (ROC) techniques. We want a figure of merit that we can calculate without requiring lengthy observer studies. To perform this calculation, we need a model of the imaging system hardware in cascade with a verified model of the human observer. We have chosen to approach this problem by modelling the human observer as an ideal observer. Our hypothesis is that the human observer acts approximately as an ideal-observer who does not have the ability to prewhiten the noise in an image. Without this ability, the ideal observer's detection performance for even a simple task is degraded substantially in correlated noise. This is just the effect that we have found for human observers. In search of a physiological explanation for a human observer's inability to do prewhitening, we shall investigate the detection capability of the ideal observer when a frequency-selective mechanism is invoked. This mechanism corresponds to the frequency channels known to exist in the human visual system. We shall show that the presence of such a mechanism can explain the degradation of human observer performance in correlated noise.
Type:
text; Dissertation-Reproduction (electronic)
Keywords:
Diagnostic imaging.
Degree Name:
Ph.D.
Degree Level:
doctoral
Degree Program:
Optical Sciences; Graduate College
Degree Grantor:
University of Arizona
Advisor:
Barrett, D. H.

Full metadata record

DC FieldValue Language
dc.language.isoenen_US
dc.titleVISUAL PERCEPTION IN CORRELATED NOISE (MODELS).en_US
dc.creatorMYERS, KYLE JEAN.en_US
dc.contributor.authorMYERS, KYLE JEAN.en_US
dc.date.issued1985en_US
dc.publisherThe University of Arizona.en_US
dc.rightsCopyright © is held by the author. Digital access to this material is made possible by the University Libraries, University of Arizona. Further transmission, reproduction or presentation (such as public display or performance) of protected items is prohibited except with permission of the author.en_US
dc.description.abstractThis dissertation concerns the ability of human observers to perform detection tasks in medical images that contain structured noise. We shall show that physical measures of image quality, such as signal-to-noise ratio (SNR), resolution, modulation transfer function (MTF), and contrast, do not accurately predict how well an observer can detect lesions in an image. We have found that for images with equal pixel SNR, humans can detect a low contrast object more readily in images that have a low-pass noise structure, as opposed to a high-pass noise structure. This finding is important in the comparison of images generated by a classical pinhole imaging system with images generated by a computed tomography imager. We would like to have a figure of merit that accurately predicts a physician's ability to perform perceptual tasks. That is, we want a figure of merit for imaging systems that is more than an evaluation of the physician's performance, measured using human observers and an accepted method such as receiver operating characteristic (ROC) techniques. We want a figure of merit that we can calculate without requiring lengthy observer studies. To perform this calculation, we need a model of the imaging system hardware in cascade with a verified model of the human observer. We have chosen to approach this problem by modelling the human observer as an ideal observer. Our hypothesis is that the human observer acts approximately as an ideal-observer who does not have the ability to prewhiten the noise in an image. Without this ability, the ideal observer's detection performance for even a simple task is degraded substantially in correlated noise. This is just the effect that we have found for human observers. In search of a physiological explanation for a human observer's inability to do prewhitening, we shall investigate the detection capability of the ideal observer when a frequency-selective mechanism is invoked. This mechanism corresponds to the frequency channels known to exist in the human visual system. We shall show that the presence of such a mechanism can explain the degradation of human observer performance in correlated noise.en_US
dc.typetexten_US
dc.typeDissertation-Reproduction (electronic)en_US
dc.subjectDiagnostic imaging.en_US
thesis.degree.namePh.D.en_US
thesis.degree.leveldoctoralen_US
thesis.degree.disciplineOptical Sciencesen_US
thesis.degree.disciplineGraduate Collegeen_US
thesis.degree.grantorUniversity of Arizonaen_US
dc.contributor.advisorBarrett, D. H.en_US
dc.identifier.proquest8522817en_US
dc.identifier.oclc696622287en_US
All Items in UA Campus Repository are protected by copyright, with all rights reserved, unless otherwise indicated.