In Computed Tomography (CT), two-dimensional (2-D) slices or three-imensional (3-D) volumes of an object are reconstructed from many projected line-integrals (usually x-ray transmission data) around the object. As the data collection capabilities and reconstruction algorithms for CT have become more sophisticated over the years, the demands on computer systems have become correspondingly greater. For example, cone-beam data acquisition of a single 2-D projection containing 1024 by 1024 resolution is now easily achievable in much less than 1 second. Accepting and processing a volume of data at those rates is impossible for most conventional computers. Also, recent limited-data reconstruction algorithms using iterative schemes between image and projection domains [1] require large amounts of very time-consuming calculations. In this case, repeated use of a constrained projection model (or the Radon transform, named after mathematician Johann Radon [2]) followed by a reconstruction algorithm (or inverse Radon transform) is used to converge on the correct answer.