Electronic Theses and Dissertations

Identifier

1051

Date

2014

Document Type

Thesis

Degree Name

Master of Science

Major

Electrical and Computer Engr

Concentration

Electrical Engineering

Committee Chair

Chrysanthe Preza

Committee Member

Russell Jerry Deaton

Committee Member

Eddie L. Jacobs

Abstract

The presence of spherical aberration (SA) due to the refractive index mismatch between the objective lens immersion medium and the specimen embedding medium makes three-dimensional (3D) imaging with a light microscope depth variant. Therefore, depth variant (DV) image restoration algorithms are required to restore an object from its 3D microscope image. In this thesis, the performance of an expectation maximization (EM) algorithm that solves the maximum likelihood microscope imaging problem is investigated. This algorithm is based on a principal component analysis (PCA) for the representation of the DV imaging model (PCA-EM). Simulated noisy and noiseless images are restored and compared to the true object using qualitative and quantitative performance analysis metrics. A comparative study is also conducted between the PCA-EM algorithm and a strata-based DV-EM algorithm, which shows that the PCA-EM is more efficient in terms of both the restoration accuracy and execution time. In addition, in order to show the performance of the PCA-EM algorithm in a real application, an experimentally acquired image of a test sample is restored. An interpolated coefficient based PCA-EM (IC-PCA) algorithm is proposed and tested using both experimental and simulated image which reduces the computational cost in computing principal components and thus, the required computational resource. It is observed from the comparative study between PCA-EM and IC-PCA that interpolating the coefficient does not affect the image restoration accuracy.

Comments

Data is provided by the student.

Library Comment

Dissertation or thesis originally submitted to the local University of Memphis Electronic Theses & dissertation (ETD) Repository.

Share

COinS