Semi-Automatic 3D Reconstruction Processes for Stereoscopic Hard X-Ray Microscopy

In recent years, there has been significant progress in developing hard X-ray ptychography for high-resolution visualization of material properties. By integrating this technique with multi-slice or tomographic methods, researchers can analyze optically thick samples in 3D. However, challenges arise in scenarios where sample rotation is restricted, hindering the extraction of 3D information from 2D projections alone. Therefore, researchers have implemented additional X-ray optics to illuminate samples from different angles, creating stereo perspectives for enhanced depth perception akin to human vision. This stereo X-ray approach, aided by stereoscopic visualization techniques, facilitates the extraction of 3D depth details. While human observers can provide valuable depth insights, the quantitative evaluation of stereoscopic X-ray data remains complex due to the theoretical underdetermination of 3D reconstruction from two projections without additional constraints or prior knowledge.
The proposed thesis aims to develop a semi-automatic interactive 3D reconstruction process for materials in stereoscopic hard X-ray microscopy. As a starting point, we will focus on microchips, with the aim to differentiate between electronic layers based on distinctive shapes and material properties and extend this approach to other sample types. Integrating human expertise and deep learning methods could enhance the 3D reconstruction process and enable more efficient automated reconstruction workflows.