Quantitative and Qualitative Evaluation of Geometric Deconvolution of Distortion in Limited-View Computed Tomography

Images reconstructed using a set of a few projections spanning a narrow angular range suffer from a systematic geometric distortion due to the point spread function of the reconstruction process. This distortion can be removed by deconvolving a complementary set of projections calculated from the initial reconstruction. Homomorphic deconvolution and inverse filtering techniques were used for this purpose. A second reconstruction is computed from the union of both sets of projections. Although the distortion is removed, the results are noisy due to problems associated with inverse filtering. Two-dimensional (2-D) filtering of the second reconstruction is performed to reduce the noise, while preserving the reduction in geometric distortion. Results obtained using several 2-D filters were compared visually and by point noise content, distortion, and projection error measures. Quantitative measures of geometric distortion removal for test images are also described.