Abstract
The major objective of this work has been to investigate methods of reconstructing tomographic images from incomplete projection data. Furthermore the practical application of such techniques to industrial non-destructive testing has been considered with particular regard to the nuclear industry. Two distinct situations are considered, region of interest (ROI) tomography and limited angle of view (LV) tomography. ROI tomography relates to situations where data is limited in linear extent and can be used for high spatial resolution imaging of particular areas of interest within larger structures. Data collection times are reduced by concentrating on the ROI and the imaging of structures which cannot fit in the field of view of the scanner can be made possible. It has been shown that corrected ROI images can be of equal quality to those reconstructed from complete data. The situation where data is limited in angular range is known as LV tomography. Practical applications of such situations can include in situ imaging of objects which cannot be accessed at all required angles, and the imaging of time varying objects where limitations on the data collection times restrict the angular range of measurements. The use of the Gerchberg-Papoulis algorithm has been shown to significantly reduce the resulting artifacts. The initial work involved investigation into the minimum data requirements for tomographic imaging of objects without compromising image quality. The relative performance of filtered backprojection and ART iterative reconstruction algorithms were investigated and the superiority of ART in situations of limited data was demonstrated. The most important causes of SPECT image degradation are scattering and attenuation of photons. For scatter correction the dual energy window and Wiener deconvolution correction methods have been investigated and the results compared. A number of attenuation correction algorithms have also been investigated and their comparative performance evaluated.