On Few-View Tomography and Staircase Artifacts

This paper investigates some potential methods for few-view tomography, and investigates the cause and remedy of the staircase artifacts. This paper also discusses whether there is any benefit to use edge-preserving filters in emission tomography. We formulate a general Green's one-step-late algorithm, so that it can incorporate any linear or non-linear filters. We argue that the derivative of the penalty function can be "artificially" created, not naturally derived from a penalty function. We have gained more insight into constrained image reconstruction especially with edge-preserving constraints. Our numerical results show that the bilateral method can have better performance than the TV method, and higher order methods may not be necessary for non-piecewise-constant objects.