Comparison of Metal Artefact Reduction Algorithms from Medicine Applied to Industrial XCT Applications

Artefacts are mainly arising due to scattered radiation and polychromatic nature of industrial X-ray sources which are typically used in X-ray computed tomography (XCT) systems. Artefacts are especially dominant in the case of multimaterial components consisting of low absorbing and metallic parts, which degrade image quality and hinder e.g. proper segmentation, failure analysis or inspection of interfaces. State-of-the-art metal artefact reduction (MAR) methods are either using iterative methods with prior knowledge or sinogram-based inpainting of the raw projection data with deletion of metal traces (from e.g. implants) and the subsequent interpolation of missing data. These approaches have already shown their usefulness in clinical routine and diagnostics by reducing image noise or resolving small structures beside metal parts. Since industrial parts are usually more complex, made of multi-material components such as plastic housings with metal pins or conducting paths made of iron or copper, an even stronger appearance of artefacts as compared to medical data can be observed. Therefore some modifications may be applied to the existing MAR methods. This contribution shows results of sinogram-based correction methods applied to real world parts as well as simulated XCT data from different industrial applications with methodical modifications such as dual energy data as input. Benefits and limitations will be discussed for each task and specimen regarding artefact reduction, gain in information, detectability and image quality.