Michael Kellman

kellman at eecs dot berkeley dot edu

I am a PhD graduate student at UC Berkeley in the EECS department, where I work on problems in the fields signal processing, computational sensing, compressive sensing, and inverse problems. I am advised by Prof. Laura Waller and Prof. Michael Lustig and am funded by NSF GRFP. I am also affiliated with the Berkeley Artificial Intelligence Research Laboratory and the Berkeley Center for Computational Imaging. I am an enthusiast of engineering, science, food, and art.

I did my undergrad at Carnegie Mellon University in the ECE Department, where I was advised by Prof. Richard Stern. I have also spent time on Fitbit's R&D team and at the Laboratory of Biomedical Stochastic Physics at NIH working with Prof. Jana Kainerstorfer.

Curriculum Vitae  /  Github  /  Google Scholar

Data-Driven Design for Fourier Ptychographic Microscopy
Michael Kellman, Emrah Bostan, Michael Chen, Laura Waller
IEEE International Conference for Computational Photography, 2019.
Fourier Ptychographic Microscopy (FPM) is a computational imaging method that is able to super-resolve features beyond the diffraction-limit set by the objective lens of a traditional microscope. This is accomplished by using synthetic aperture and phase retrieval algorithms to combine many measurements captured by an LED array microscope with programmable source patterns. FPM provides simultaneous large field-of-view and high resolution imaging, but at the cost of reduced temporal resolution, thereby limiting live cell applications. We learn LED source pattern designs that compress the many required measurements into only a few, with negligible loss in reconstruction quality or resolution. This is accomplished by recasting the Fourier Ptychographic (super-resolution) reconstruction as a Physics-based Neural Network and learning the experimental design to optimize the network's overall performance. We learn context-specific designs for both histological and quantitative phase imaging applications.

Featured Research

I'm interested in the areas signal processing, computational imaging, optimization, inverse problems, statistics, and learning. Much of my work is in the areas of optics, microscopy, medical imaging, and image reconstruction. Specifically, I focus on joint design of hardware systems and signal processing algorithms to beat traditional imaging limits.

Physics-based Learned Design: Optimized Coded-Illumination for Quantitative Phase Imaging
Michael Kellman, Emrah Bostan, Nicole Repina, Laura Waller
IEEE Transactions on Computational Imaging, 2019.
blog / poster / bibtex

We present a novel data-driven design method to optimize the end-to-end performance of the LED array microscope and a non-linear quantitative phase reconstruction. By unrolling the iterations of a phase reconstruction we build a network that incorporates both the system physics of measurement formation and the non-linearities of iterative reconstruction. Our Unrolled Network is effiently parameterized by only a few design variables which can be learned from just a few simulated training examples and still generalize well in the experimental setting.

Motion-resolved Quantitative Phase Imaging
Michael Kellman, Michael Chen, Zachary Phillips, Michael Lustig, Laura Waller
Biomedical Optics Express (BOEx), 2018.
talk / poster / video / bibtex

The temporal resolution of quantitative phase imaging with Differential Phase Contrast (DPC) is limited by the requirement for multiple illumination-encoded measurements. This inhibits imaging of fast-moving samples. We present a computational approach to model and correct for non-rigid sample motion during the DPC acquisition in order to improve temporal resolution to that of a single-shot method and enable imaging of motion dynamics at the framerate of the sensor.

Node-Pore Coded Coincidence Correction: Coulter Counters, Code Design, and Sparse Deconvolution
Michael Kellman, Francios Rivest, Alina Pechacek, Lydia Sohn, Michael Lustig
IEEE Sensors Journal, 2018.
poster / code / data / bibtex

Using communication theory and detection theory, we create a Barker-coded high-throughput microfluidic channel. We jointly design the channel's system response and cell detection heuristic to efficiently resolve coincidence events by way of an inverse problem formulation.

This work is done in collaboration with the Sohn Lab.

Physics-Based Learned Design: Teaching a Microscope How to Image
Michael Kellman, Emrah Bostan, Laura Waller
Nov. 26, 2018

Computational imaging systems marry the design of hardware and image reconstruction. For example, in optical microscopy, tomographic, super-resolution, and phase imaging systems can be constructed from simple hardware modifications to a commercial microscope and computational reconstruction. Traditionally, we require a large number of measurements to recover the above quantities; however, for live cell imaging applications, we are limited in the number of measurements we can acquire due to motion. Naturally, we want to know what are the best measurements to acquire. In this post, we highlight our latest work that learns the experiment design for a non-linear computational imaging system.


EE16A - Fall 2018 (discussion and content GSI)

Copyright 2018 Michael Kellman. All rights reserved.
Template borrowed from Jon Barron.

URL Counter