Multiresolution image modeling and Bayesian reconstruction algorithms with applications to emission tomography
Abstract
Presented in this thesis are multiresolution image models and Bayesian algorithms for statistical image reconstruction. The proposed methods are generally applicable to inverse problems in image processing, however, the focus of this work is the application to tomographic imaging. Chapter two presents a wavelet graph image model designed to perform space-adaptive regularization. The model is based on a novel hierarchical dependency structure in the wavelet tree. An efficient multiresolution Bayesian reconstruction algorithm is proposed that allows for enforcement of space-domain constraints such as positivity despite using a wavelet prior model. Chapter three presents a comparison of Bayesian algorithms for tomographic reconstruction with emphasis on accurately modeling the tomography scanner. Included in this work is a new approach to obtaining an empirical model of the tomography scanner's system kernel. Chapter four presents a Bayesian multiresolution framework for discrete-valued image reconstruction in transmission and emission tomography. The approach includes a new method for efficient and accurate estimation of the discrete levels in the reconstruction image. Chapter five addresses the problem of image modeling for content-based database search and presents a subject study on human perception of the similarity of images.
Degree
Ph.D.
Advisors
Bouman, Purdue University.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.