Comparison of Two Methods for Tomographic Imaging from Severely Incomplete Data

E. P. A. Constantino, J. L. Davidson and K. B. Ozanyan

School of Electrical and Electronic Engineering, The University of Manchester, UK Email: e.pereira-araujo-c@student.manchester.ac.uk

ABSTRACT

The problem of limited data in industrial tomography calls for ever-innovative data processing strategy and reconstruction algorithm design. A novel sinogram restoration method has been suggested which not only provides the mean for missing data estimation for image reconstruction, but also is capable of computing the location and extent of object constituents present within the image space without the need for inverse transformation. In this study, the new method is compared with an adapted version of the Landweber reconstruction method. Tests have been carried out to determine the performance and reconstruction characteristics of the two, using both simulated and experimental data, with numerical error analysis. Results have shown that the sinogram restoration method is particularly suitable for regular imaging geometry and capable of identifying the number and location of objects with good fidelity.

Keywords Hard-Field Tomography, Hough Transform, Landweber method, sinogram restoration