Yuanrui Chen

I'm a 3rd year PhD student at ECE department of Purdue University in West Lafayette, where I work on computational imaging using metasurfaces. I am advised by Prof. Qi Guo.

Email  /  CV  /  Github

Research

I'm interested in optical and photonic design, computational imaging and deep learning. I also work on nanophotonics fabrication.

High dynamic range (HDR) imaging enables resolving the full brightness profile of scenes where critical information spans a wide range of intensity values. Traditional solutions based on sequentially capturing frames with different exposure times as well as sensor pixel mosaicking techniques often suffer from limitations in terms of light collection, motion blur, and sensor readout complexity. In this article, a system comprising a single meta-optic coupled with a post-processing routine enables single-shot HDR imaging with up to a 50 dB improvement in dynamic range. The core innovation relies on a meta-optic in which a randomized interleaving of scatterers enables forming an array of focal spots with different relative intensities. When imaging a scene, this design produced an array of sub-images with different brightnesses, which when coupled with a novel, gradient-based processing method reconstructed scenes with minimal artifacts. The system was used to estimate the surface curvature of specular objects with a single shot, which is a challenging problem typically entailing stereo capture and extensive calibration. While traditional HDR approaches support broadband imaging, whereas the method here is limited to narrowband operation, it’s expected this approach could be applied in a variety of sensing, security, and industrial applications.