Refined segmentation of synchrotron XCT-based characterization of metal powders under different noise conditions

Abstract

The characterization of metal powders is crucial for assessing their suitability in additive manufacturing applications. Synchrotron X-ray computed tomography provides high-resolution, three-dimensional data for evaluating particle morphology and internal porosity. However, accurate interpretation of such datasets, particularly for metallic powders, remains challenging due to reconstruction artifacts and segmentation sensitivity. This study presents an optimized post-reconstruction segmentation framework using marker-controlled watershed implemented using Fiji, aimed at improving robustness and reproducibility in synchrotron-based powder analysis. Aluminum alloy powders sieved to less than 150 μm and less than 100 μm were analyzed to investigate the influence of image quality, filtering, and marker definition on segmentation outcomes. The finer powder sample, affected by noise voxels, required a two-step adaptive filtering strategy and careful tuning of the contrast threshold based on the distance map, using the h-parameter in the extended maxima function. In contrast, the coarser sample, exhibiting minimal noise, showed stable segmentation performance across a broader range of threshold values. Morphological metrics such as Feret diameter and Equivalent Spherical Diameter were employed to evaluate the accuracy of segmentation and particle shape. Results indicate that dataset-specific preprocessing and distance-driven marker definition are essential for minimizing under- and over-segmentation in noise-affected data. The proposed framework enhances the reliability of synchrotron X-ray computed tomography-based powder analysis, supporting its use in quality control of feedstock materials for additive manufacturing.