PIXEL-BASED RECONSTRUCTION (PBR) PROMISING SIMULTANEOUS TECHNIQUES FOR CT RECONSTRUCTIONS

We present some new algorithms belonging to a class of algorithms, pix el-based reconstruction (PBR), similar to ''SIRT'' (simultaneous itera tive reconstruction techniques) methods for reconstruction of objects from their fan beam projections in X-ray transmission tomography. The general logic of these algorithms is discussed, and. as a corollary, t wo new ideas are presented, which gave promising results in the simula tion studies. It was found in our simulation studies, contrary to prev ious results with parallel beam projections, that these iterative alge braic algorithms don't diverge when a more logical technique obtaining the pseudo-projections is used. These simulations were carried out un der conditions where the number of object pixels exceeded (double) the number of detector pixel readings, i.e., the equations were highly un der-determined, however, the reconstructions were quite satisfactory. The effect of the number of projections on the reconstruction and the convergence (empirical) to the exact solution is shown. For comparison , the reconstructions obtained by convolution back-projection are also given.