- New method enables fast 3D reconstructions of nanoscale objects.
- It works by taking two images of an object from two different locations and combining them to form a spatial image.
There exist countless object in nature that has complex 3D dynamical structures. Understanding such objects in details could be beneficial in multiple fields of study.
Recent advances in lensless imaging techniques have enabled 2D single-shot imaging on femtosecond or nanometer scale, using ultrafast coherent X-ray sources. However, achieving ultrafast 3D imaging, from a dynamic perspective, is a difficult task.
Now a team of international researchers has developed a new method to accelerate the process of 3D imaging considerably. It enables 3D reconstructions on a nanoscale through a single image that contains two images from two different perspectives.
How Did They Do It?
In the human brain, 3D perception is created from a stereopsis reconstruction. Whereas, in computer stereo vision, depth information is extracted after 3 steps: image rectification, disparity map estimation, and 3D point cloud projection.
In this study, researchers applied the computer stereo vision concept to X-rays and carried out 3D reconstruction at the nanometer scale. Since machine stereo vision relies on computational constraints that are often based on uniqueness and similarity, it is challenging to obtain precise and realistic disparity maps.
The team found that X-ray phase-contrast images fulfill such computational constraints and occulted sites restored through phase-contrast views as stereo pairs. However, for ambiguous transmission geometry, such as smooth composition gradients and spherical topographies, this method doesn’t precisely capture the depth of structures.
The object (cut out cross in the center) is illuminated from two different directions, and star-shaped diffraction pictures are reconstructed into a three-dimensional object (in the bottom). | Courtesy of researchers
Nonetheless, one can use labels to retrieve more complex 3D structures without preprocessing images. Researchers used nanoparticles of gold to sample the disparity and obtain a precise 3D representation of a nanopyramid. Such nanoparticles can also be used to image biological systems.
Overall, this stereo technique requires only two views (captured by a single laser pulse) to obtain ultrafast 3D reconstruction on a nanometric scale.
By using adaptive compressed stereo images, the image processing can be further optimized for real-time 3D vision at high X-ray sources, such as X-ray free-electron lasers.
The technique developed in this work could have a substantial impact on three-dimensional structural imaging of individual molecules. Moreover, it could be used in medicine and biology field. For instance, the protein structure of a virus, which plays a decisive role in medical diagnoses, can be examined quickly with very little effort.