Data Acquisition Errors and the Choice of Filter in Filtered Backprojection Algorithms.

摘要

In this paper, we examine issues specific to Filtered Backprojection (FBP) algorithms. While FBP techniques can perform quality image reconstructions, in an economical way, FBP algorithms can produce troublesome artifacts in image reconstructions. At LLNL we have investigated artifacts which result from errors in the measurement of the center of rotation, bad pixels in the projection data (the sinogram), the choice of filter used in the reconstruction code, and difficulties resulting from combinations of these effects. This paper is divided into five sections: a presentation of some fundamentals of computed tomography (CT), and FBP algorithms, a characterization and discussion of centering errors, the effects of singularly-bad pixels, a section on the consequences of choosing different filters in the FBP algorithms in the DONNER library of CT reconstruction routines, and a summary discussion on CT data analysis in industrial inspections. 7 refs., 2 figs.