 |
The STIR group currently has three major research themes.
Photon-Efficient Imaging and Signal Processing for Single-Photon Detectors
Conventional photography collects tens of thousands of photons at each camera pixel. However, that number of photons is not required to form accurate images.
Using time-correlated single-photon counting, our group and collaborators at MIT demonstrated that forming accurate depth and reflectivity images is possible
using as little as one photon detection per scene pixel, increasing the photon efficiency of image formation by several orders of magnitude.
Such methods can enable long-distance range measurements, fast acquisition for real-time systems like autonomous vehicles,
or even low-dose measurements for biological applications.
We have continued to expand the capabilities of imaging with very few photons, including exploring scenes with multiple depths,
increasing robustness to high ambient light levels, improving depth resolution for systems with coarse time quantization,
and compensating for dead times in single-photon detectors.
IEEE SPS: First-Photon Imaging and Other Imaging with Few Photons from Radha Giduthuri on Vimeo.
Key papers:
- Photon-Efficient Imaging with a Single-Photon Camera
D. Shin, F. Xu, D. Venkatraman, R. Lussana, F. Villa, F. Zappa, V. K. Goyal, F. N. C. Wong, and J. H. Shapiro,
Nature Communications, vol. 7, art. no. 12046, 24 Jun 2016.
- Photon-Efficient Computational 3D and Reflectivity Imaging with Single-Photon Detectors
D. Shin, A. Kirmani, V. K. Goyal, and J. H. Shapiro,
IEEE Trans. Computational Imaging, vol. 1, no. 2, pp. 112-125, June 2015.
- Winner of a 2019 IEEE Signal Processing Society Best Paper Award
- First-Photon Imaging
A. Kirmani, D. Venkatraman, D. Shin, A. Colaço, F. N. C. Wong, J. H. Shapiro, V. K. Goyal,
Science, vol. 343, no. 6166, pp. 58-61, 3 Jan 2014.
Non-Line-of-Sight Imaging
When scenery is hidden from view, light from the hidden region may be cast onto visible diffusely reflecting surfaces forming partial shadows called ‘penumbrae.’
If one is able to determine the proportions of penumbra light arriving from different directions, it is possible to form estimates of hidden scenery.
This is an approach to passive non-line-of-sight (NLOS) imaging.
NLOS imaging provides value in a variety of circumstances such as monitoring of hazardous environments, navigation, detecting hidden adversaries,
and in collision avoidance for automated vehicles.
Many NLOS imaging systems rely on the fact that photon travel time is proportional to distance, and thus require expensive, ultra-fast optical systems.
Our passive approaches use inexpensive, ubiquitous sensors, such as typical digital cameras,
to make inferences without requiring time-varying or controlled illumination of the hidden scene.
Key papers:
Focused Particle Beam Microscopy
There are several processes that limit the image quality of focused ion beam microscopy. In a fixed dwell time, randomness in the source ion beam
results in a random number of ions incident on the sample; we call this source shot noise. Each incident ion causes a random number of secondary electrons
to be dislodged from the sample. In addition, the detection of the dislodged secondary electrons is also a further source of noise.
Our work focuses on developing an understanding of (and models for) these phenomena, along with estimation algorithms, that push the limits beyond
what is presently achievable.
Conventionally, increasing ion dose will improve the image quality but also causes more damage to the sample.
We have shown that time-resolved sensing can enable dose reduction for any desired image quality, in return minimizing damage caused to sample of interest.
Key papers:
A few topics have information pages. This does not represent an exhaustive list of the work we are most proud of.
|
|
