Artifact reduction filtering method for CT images

Artifacts often observed in computed tomography (CT) images are required to be suppressed. We propose an artifact reduction filtering method. This method is an iteration procedures consisting of series of simple linear and nonlinear operations. The filtering operation is applied to sinogram having a two-dimensional filter kernel. Only the size of kernel is set ahead of time but its shape is determined in the least square sense. Although the performance of each series of operations are simple and restricted as it is, that of their iteration are expected to produce more complicated performance. One time use of the method is considered as a two-dimensional filtered backprojection. We applied this method to a micro X-ray CT image. The original image suffered from much ring artifact. The results indicated that the ring artifact was reduced but additional artifacts were produced, with the number of iterations increased. The filter kernel performed as lowpass along the detector direction as well as highpass along the projection angle direction. Although decision of the number of iteration should be paid attention, this method will be beneficial to reducing artifacts for CT images.