Improved Computed Tomography by Variable Desmearing Model Reconstructions by Iterative DIRECTT Algorithm

In order to improve the spatial resolution of computed tomography reconstructions which suffer from un-sharpness due to an extended source size a new procedure of de-smearing has been developed. As the width of the source related smearing function of the projections may even double between the position near and far from the detector the desmearing procedure has to be variable. In contrast to the essentially wrong approach of de-convolution by a constant smear kernel for all voxel positions an implicit desmearing is established. The pre-known position dependent smearing is applied to the iteratively repeated projections of partial reconstructions obtained from the DIRECTT reconstruction algorithm. This approach of variable desmearing from sinograms of a pixel model with smearing from one to five pixel FWHM achieves a spatial resolution near one pixel.