skip to main content

PhD Thesis Defense

Thursday, December 21, 2023
10:00am to 11:00am
Add to Cal
Moore 239
Computation and optimization in imaging: non-line-of-sight imaging and label-free, large field-of-view imaging.
Ruizhi Cao, Graduate Student, Electrical Engineering, California Institute of Technology,

Computation and optimization methods empower modern imaging techniques in various aspects, such as high-resolution imaging, deep-tissue imaging, and aberration correction. In my thesis defense, I will discuss two main imaging techniques we have developed, one focuses on solving the scattering related non-line-of-sight imaging problem and the other focuses on developing high throughput, label-free imaging technique.

In the first part, I will discuss non-line-of-sight imaging where the target is obscured and cannot be directly imaged. The time-of-flight method is a dominant approach for solving this challenge, but its complex image reconstruction and hardware requirements limit the resolution and fidelity of its results in certain scenarios. We have developed an active imaging technique that utilizes wavefront shaping to generate a physical focus on the hidden target. By optimizing the wavefront disturbance introduced by the scattering surface of a wall, we can actively compensate for such wavefront imperfections and generate a focus onto the target. With this physically generated focus, we can directly employ the confocal-type raster scanning strategy to form an image.

In the second part, I will introduce a new high-resolution, large field-of-view (FOV) label-free imaging technique. One of the most successful methods for achieving this is Fourier ptychographic microscopy (FPM), which uses low-magnification objectives and a series of tilted illuminations to generate high-resolution, large FOV images. However, its non-convex iterative reconstruction algorithm can be sensitive to convergence criteria and makes FPM prone to parameter selections. I will introduce a new analytical phase retrieval framework that we have developed. This framework simultaneously reconstructs the complex sample field and corrects for system aberrations by using tilted illuminations. We have demonstrated that our new method outperforms FPM in terms of both reconstruction fidelity and aberration robustness – it corrects aberrations of up to a maximal phase difference of 3.8π in our experiments.

For more information, please contact Tanya Owen by email at [email protected].