Research on computed tomography reconstructions from one or two radiographs: A report and the application to FXR radiography

This report documents some cooperative research into volumetric image reconstruction from single radiographs. Imaging dynamic events is the most important application for this type of work, but the techniques have possible extensions. Two general objectives guide this work. The first objective is to gain an understanding of the assumptions and limitations of single-view methods for representing internal features. Second, we endeavor to obtain and/or develop techniques for performing image reconstructions with FXR radiographs. If possible, we seek to obtain some quantitative measure of the accuracy of this class of image reconstructions in two respects: (i) in terms of the dimensional accuracy of feature boundaries, and (ii) as pertains to the accuracy of the voxel intensities. Dynamic events are not always self-calibrating, and it is important to establish the reconstruction accuracy of single-view methods for placing bounds on the kinds of conclusions which can be advanced from single-view reconstructed images. Computed tomographic image reconstructions provide dimensional detail of internal structures of objects and provide a measure of the per-voxel attenuation of material in the object. When assumptions behind a reconstruction algorithm are not satisfied, or are satisfied in a limited way, the accuracy of the reconstructed image is compromised. It is the goal of Cr analysis to discern the {open_quotes}real{close_quotes} features of the internals of an object in the midst of a certain level of artifactual content in the image. By understanding the ways in which CT reconstructions from a single radiograph can produce misleading results we hope to develop some measure of the benefits and limitations of single view techniques. 31 refs., 20 figs