Calculation of Reconstruction Kernels in Computed Tomography

Filtered backprojection (FBP) is the standard reconstruction method in clinical Computed Tomography (CT). Unfortunately, due to a high-pass filter in the reconstruction process, the noise in the measured X-ray projection data is amplified. Therefore, a one-dimensional window or kernel function must be used to supress high frequencies. Depending on the design of the filter, it either provides a high image resolution or a reduced image noise. In this work, a new method to calculate unknown reconstruction kernels from clinical CT scanner is presented. The knowledge of the specific kernel shape can be used to compare new reconstruction methods with the one implemented in a clinical scanner.