Computed tomography (CT) is a three-dimensional imaging method, widely used in medicine and non-destructive testing (NDT). The mathematical model of CT leads to the Radon transform, whose inversion constitutes a classic ill-posed inverse problem. Apart from the reconstruction algorithms of filtered backprojection type, there also exist iterative methods, which allow to exploit a priori knowledge about the object. Especially if only few or limited data are available, as in computed laminography

more »

... CL), using this a priori information leads to much better results than those of standard reconstruction algorithms. The aim of the present work lies in optimizing iterative reconstruction techniques to compensate for artifacts caused by limited data thus rendering the resulting recontructions amenable to NDT defect analysis. To that end, sampling strategies are developed, reducing aliasing artifacts. Three-dimensional volume filters are integrated into the reconstruction process to correct noisy and artifact-ridden reconstruction due to few projections. Morphologic operators are used to compute effective a priori weighting volumes. Previous, faulty a priori methods are rectified by adapted ray length corrects thereby achieving significant increases in image contrast. Region-of-interest artifacts caused by truncated projections, which are common in computed laminography, are compensated for by a geometric ray weighting algorithm. Simulated and real measured data sets are used to demonstrate the achieved algorithmic improvements.