An Improved Algorithm for Reprojecting Rays through Pixel Images

Abstract

It is often desired to calculate line integrals through a field of reconstructed CT density pixels for the purpose of improving CT image quality. Two algorithms widely published and discussed in the past are known to either degrade spatial resolution or generate errors in the results due to the discontinuous "square pixel" modeling of the reconstructed image. An algorithm is described, based on linear interpolation between pixels, which provides superior accuracy without unnecessary loss of resolution. It was tested on simulated data for a head section and on a narrow Gaussian density distribution. The experimental results demonstrated improved performance. The method is expected to prove useful for many types of post-reconstruction processing, including beam hardening, missing data, and noise supression algorithms.