Persistent Link:
http://hdl.handle.net/10150/288785
Title:
Assessment of reconstructed images
Author:
Abbey, Craig Kendall, 1964-
Issue Date:
1998
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 addresses the problem of assessing the image quality of reconstructed images. It adopts the framework of objective assessment of image quality set forth by Barrett (Barrett, 1990). This approach to image quality specifies image quality as the ability to perform some task of interest using the image data. The analysis of image reconstruction begins by specifying the imaging chain as a series of noisy transformations that ultimately lead to some sort of inference about the object being imaged. Each step of the imaging chain is identified with a physical process in the creation of an image or the subsequent use of the image to perform a task of interest. After development of the imaging chain, the dissertation focuses on various parts of the chain needed to analyze image reconstruction. One area necessary for investigating image reconstruction is understanding the propagation of noise in the imaging chain through to the reconstructions. This work focuses on first- and second-order statistical properties realized as a mean reconstruction and a reconstruction covariance matrix. The analysis then turns to the task of signal-known-exactly (SKE) detection where the statistical properties of the reconstructed images are used to obtain formulas for the observer SNRs for a variety of detection strategies, referred to as model observers. A number of psychophysical experiments have been conducted with the dual purpose of testing factors that influence human detection performance and providing a set of data for assessing the ability of different model observers to predict human performance. The experiments are divided into "filtered-noise" studies that simulate the entire reconstruction process, and "tomographic-reconstruction" studies that reconstruct simulated data vectors using two well-known reconstruction algorithms. The human observer performance in the psychophysical studies is then compared to that of the model observers. Some results of these comparisons are a generally high level of human performance, and a generally good agreement between human observer performance and the performance of a channelized Hotelling observer with a particular model of internal noise.
Type:
text; Dissertation-Reproduction (electronic)
Keywords:
Mathematics.; Statistics.; Psychology, Experimental.
Degree Name:
Ph.D.
Degree Level:
doctoral
Degree Program:
Graduate College; Applied Mathematics
Degree Grantor:
University of Arizona
Advisor:
Barrett, Harrison H.

Full metadata record

DC FieldValue Language
dc.language.isoen_USen_US
dc.titleAssessment of reconstructed imagesen_US
dc.creatorAbbey, Craig Kendall, 1964-en_US
dc.contributor.authorAbbey, Craig Kendall, 1964-en_US
dc.date.issued1998en_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 addresses the problem of assessing the image quality of reconstructed images. It adopts the framework of objective assessment of image quality set forth by Barrett (Barrett, 1990). This approach to image quality specifies image quality as the ability to perform some task of interest using the image data. The analysis of image reconstruction begins by specifying the imaging chain as a series of noisy transformations that ultimately lead to some sort of inference about the object being imaged. Each step of the imaging chain is identified with a physical process in the creation of an image or the subsequent use of the image to perform a task of interest. After development of the imaging chain, the dissertation focuses on various parts of the chain needed to analyze image reconstruction. One area necessary for investigating image reconstruction is understanding the propagation of noise in the imaging chain through to the reconstructions. This work focuses on first- and second-order statistical properties realized as a mean reconstruction and a reconstruction covariance matrix. The analysis then turns to the task of signal-known-exactly (SKE) detection where the statistical properties of the reconstructed images are used to obtain formulas for the observer SNRs for a variety of detection strategies, referred to as model observers. A number of psychophysical experiments have been conducted with the dual purpose of testing factors that influence human detection performance and providing a set of data for assessing the ability of different model observers to predict human performance. The experiments are divided into "filtered-noise" studies that simulate the entire reconstruction process, and "tomographic-reconstruction" studies that reconstruct simulated data vectors using two well-known reconstruction algorithms. The human observer performance in the psychophysical studies is then compared to that of the model observers. Some results of these comparisons are a generally high level of human performance, and a generally good agreement between human observer performance and the performance of a channelized Hotelling observer with a particular model of internal noise.en_US
dc.typetexten_US
dc.typeDissertation-Reproduction (electronic)en_US
dc.subjectMathematics.en_US
dc.subjectStatistics.en_US
dc.subjectPsychology, Experimental.en_US
thesis.degree.namePh.D.en_US
thesis.degree.leveldoctoralen_US
thesis.degree.disciplineGraduate Collegeen_US
thesis.degree.disciplineApplied Mathematicsen_US
thesis.degree.grantorUniversity of Arizonaen_US
dc.contributor.advisorBarrett, Harrison H.en_US
dc.identifier.proquest9829318en_US
dc.identifier.bibrecord.b38551676en_US
All Items in UA Campus Repository are protected by copyright, with all rights reserved, unless otherwise indicated.