Caltech Home > PMA Home > News > Seeing Farther and Deeper: An...
open search form

Seeing Farther and Deeper: An Interview with Katie Bouman

New Caltech faculty member Katie Bouman creates images from nonideal sensor data and mines for information from images using techniques that can be applied to everything from medical imaging to studying the universe.

An assistant professor of computing and mathematical sciences in the Division of Engineering and Applied Science, Bouman joined Caltech's faculty at the beginning of June. She earned her bachelor's degree at the University of Michigan in Ann Arbor, followed by a master's and PhD from MIT. After completing her graduate studies, she worked as a postdoctoral researcher at the Harvard-Smithsonian Center for Astrophysics. Bouman was one of about 200 scientists and engineers from across the globe who worked on the Event Horizon Telescope project, which made headlines in April for capturing the first-ever image of a black hole.

Recently, Bouman answered a few questions about her life and work.

How would you describe your research?

I like thinking about how we can use imaging to help push forward the boundaries of other fields. I did my PhD in a computer vision group—a group that tries to analyze images and understand images. A lot of people in the computer vision field work on object detection and action recognition. Those are really interesting problems, and researchers have been working for decades to build machines that mimic human intelligence in order to solve them. But there is another world of interesting problems that cameras and images can help us solve that humans are not even capable of doing on their own.

I like to search for information hidden in images, imperceptible to humans, that we can use to learn about the environment around us. This requires an understanding of the complete sensing system: how light interacts with the world and is then captured by our camera sensor into individual pixels. This line of research, where we work on merging sensors and algorithms to achieve something not possible with just one or the other, is often described as computational imaging or computational photography.

What kind of applications do you see for this work?

There are multiple sides to the research I enjoy: one side is coming up with new ways to reconstruct images invisible to traditional sensors, and another side is using images or videos to extract hidden information from a scene. For instance, I've used each pixel in a video like a very noisy sensor to recover the location of people moving behind a wall from imperceptible changes in shadows that appear on the ground. I've also used data to create an image. For example, in the black hole imaging work, we had really noisy, sparse data. We had to figure out how to create an image to learn something from what we were seeing.

Here at Caltech, I'm excited to start connecting with people across campus and help them use imaging to push the boundaries of their disciplines. I've already had the opportunity to speak with Zach Ross [assistant professor of geophysics] about how new techniques could help in more precisely localizing the origin of collections of earthquakes. This work, perhaps surprisingly, contains many similarities to the work I've done in black hole imaging.

I also will be having Aviad Levis join me as postdoc next year. Aviad has been working with JPL on studying cloud tomography: reconstructing the 3D structure and the particle distribution of clouds from 2D images taken by planes or satellites. Similar to imaging black holes, these clouds evolve as the measurements are being taken, so every measurement captures a different sample of the cloud structure. We are excited about exploring some ideas for solving both of these messy, time-evolving problems. By intelligently connecting the information from time-variable measurements, I'm confident we can design algorithms to solve for a more accurate cloud structure or a video of a black hole evolving over time.

Each problem, each application, has its own intricacies; understanding the structure of a problem is exciting, and by encoding that structure into our algorithms, we can learn more.

I understand that you're not the only Bouman on campus right now.

That's right. My sister, Amanda, is a graduate student in mechanical engineering, and my brother, Alexander, is an undergraduate also in mechanical engineering. It's definitely been nice having them around to show me the ropes and help me get settled on campus.

After living on the East Coast, how do you like being in California?

My husband and I are really enjoying it, but we're still getting used to it. It seems like every day I tell him that we need to eat outside because it's so beautiful, so we end up grilling outside every day. In Boston, we had to take advantage of every nice day. Eventually, we're going to have to stop eating just hamburgers and hot dogs.

Written by Robert Perkins

Robert Perkins
(626) 395-1862