Masters Theses


"Sparse tomography is an efficient technique which saves time as well as minimizes cost. However, due to few angular data it implies the image reconstruction problem as ill-posed. In the ill posed problem, even with exact data constraints, the inversion cannot be uniquely performed. Therefore, selection of suitable method to optimize the reconstruction problems plays an important role in sparse data CT. Use of regularization function is a well-known method to control the artifacts in limited angle data acquisition. In this work, we propose directional total variation regularized ordered subset (OS) type image reconstruction method for neutron limited data CT. Total variation (TV) regularization works as edge preserving regularization which not only preserves the sharp edge but also reduces many of the artifacts that are very common in limited data CT. However TV itself is not direction dependent. Therefore, TV is not very suitable for images with a dominant direction. The images with dominant direction it is important to know the total variation at certain direction. Hence, here a directional TV is used as prior term. TV regularization assumes the constraint of piecewise smoothness. As the original image is not piece wise constant image, sparsifying transform is used to convert the image in to sparse image or piecewise constant image. Along with this regularized function (D TV) the likelihood function which is adapted as objective function. To optimize this objective function a OS type algorithm is used. Generally there are two methods available to make OS method convergent. This work proposes OS type directional TV regularized likelihood reconstruction method which yields fast convergence as well as good quality image. Initial iteration starts with the filtered back projection (FBP) reconstructed image. The indication of convergence is determined by the convergence index between two successive reconstructed images. The quality of the image is assessed by showing the line profile of the reconstructed image. The proposed method is compared with the commonly used FBP, MLEM, and MLEM-TV algorithm. In order to verify the performance of the proposed algorithm a Shep-Logan head phantom is simulated as well as a real neutron CT image is tested to demonstrate the feasibility of the algorithm for the practical sparse CT reconstruction applications"--Abstract, page iii.


Lee, Hyoung-Koo

Committee Member(s)

Alajo, Ayodeji Babatunde
Liu, Xin (Mining & Nuclear Engr).


Nuclear Engineering and Radiation Science

Degree Name

M.S. in Nuclear Engineering


Missouri University of Science and Technology

Publication Date

Fall 2013


ix, 47 pages

Note about bibliography

Includes bibliographical references.


© 2013 Fahima Fahmida Islam, All rights reserved.

Document Type

Thesis - Open Access

File Type




Subject Headings

Tomography -- Mathematical models
Diagnostic imaging -- Digital techniques
Image processing -- Mathematics
Cross-sectional imaging -- Mathematics

Thesis Number

T 10399

Electronic OCLC #