Improved surface generation of multi-material objects in computed tomography using local histograms

During the last decade industrial computed tomography (iCT) has become one of the most important metrological procedures for internal inspection, where it sees wide-spread use in injection molding and additive manufacturing. Evaluating the CT volume data of multi-material objects represents a major technical challenge. Due to artifacts caused by beam hardening, an over-segmentation of strongly absorbing materials occurs, severely limiting the accuracy of dimensional measurements. The goal of the project presented is the development of an innovative artifact-reduced multi-material segmentation. This is applied to and tested on various complex reconstructed CT data sets. Global approaches show high signal-to-noise-ratio (SNR) but are not able to compensate for local deviations. For smaller volumes the data sets become more consistent, but the SNR decreases due to the reduced data volume. Thus, a more localized approach for the volume image data has the potential to provid e results of higher accuracy. With this newly presented algorithm it is now possible to perform segmentation of all materials, while eliminating over-segmentation errors as well as local noise artifacts almost completely for all tested datasets.