Computed stereo lensless X-ray imaging
Abstract
Recovering the three-dimensional (3D) properties of artificial or biological systems using low X-ray doses is challenging as most techniques are based on computing hundreds of two-dimensional (2D) projections. The requirement for a low X-ray dose also prevents single-shot 3D imaging using ultrafast X-ray sources. Here we show that computed stereo vision concepts can be applied to X-rays. Stereo vision is important in the field of machine vision and robotics. We reconstruct two X-ray stereo views from coherent diffraction patterns and compute a nanoscale 3D representation of the sample from disparity maps. Similarly to brain perception, computed stereo vision algorithms use constraints. We demonstrate that phase-contrast images relax the disparity constraints, allowing occulted features to be revealed. We also show that by using nanoparticles as labels we can extend the applicability of the technique to complex samples. Computed stereo X-ray imaging will find application at X-ray free-electron lasers, synchrotrons and laser-based sources, and in industrial and medical 3D diagnosis methods