ACCELERATING THE TOMOGRAPHIC RECONSTRUCTION WITH A FREQUENCY ADAPTED ALGORITHM

In several imaging modalities involving tomographic reconstruction, the model of projection should incorporate a response of the detector that takes the limitation in spatial resolution into account. Unlike analytic methods, iterative methods easily deal with changes in the model and thus perform some kind of deconvolution together with tomographic reconstruction. However, since these methods imply a slow reconstruction of the details, it is usually worth computing many iterations to reach important high frequencies. In this paper, we analyse how the frequency adapted (FA) SART algorithm proposed in solves this problem, by use of an incomplete backprojection matrix with narrower cones of backprojection. This algorithm demonstrates a significant acceleration for the reconstruction of high frequencies and also seems to give better results than those obtained with a large number of SART iterations for highly detailed structures