Skip navigation

Optimal Denoising of Image Contrast and IMU-Based Image Deblurring

Event Date: November 1, 2018
Speaker: Keigo Hirakawa
Associate Professor
Speaker Affiliation: Electrical &
Computer Engineering
University of Dayton
Time: 11:00 am
Location: MSEE 239
Priority: No
School or Program: Electrical and Computer Engineering
College Calendar: Show

Keigo Hirakwaw
Associate Professor
Electrical and Computer Engineering 
University of Dayton

In the first part of the talk, we address the denoising problem of Poisson imaging. Most conventional imaging modalities detect light indirectly by observing high-energy photons. The random nature of photon emission and detection are often the dominant source of noise in imaging. Such case is referred to as photon-limited imaging, and the noise distribution is well modeled as Poisson. Multiplicative multiscale innovation (MMI) presents a natural model for Poisson count measurement, where the interscale relation is represented as random partitioning (binomial distribution) or local image contrast. In this work, we propose a nonparametric empirical Bayes estimator that minimizes the mean square error of MMI coefficients. The proposed method achieves better performance compared with state-of-art methods in both synthetic and real sensor image experiments under low illumination.

In the second part of the talk, we develop a technique to deblur images corrupted by camera shake. The proposed deblurring method learns the blur kernel by combining the inertial measurement unit (IMU) data that track camera motion with techniques that seek blur cues from the image sensor data. Specifically, we introduce the notion of IMU fidelity cost designed to penalize blur kernels that are unlikely to have yielded the observed IMU measurements. When combined with the image data-based fidelity and regularization terms used by the conventional blind image deblurring techniques, the overall energy function is nonconvex. We solved this non-convex energy minimization problem by a novel use of distance transform, recovering a blur kernel and sharp image that are consistent with the IMU and image sensor measurements.

Keigo Hirakawa is an Associate Professor and the Program Director for the Computer Engineering Program at the University of Dayton.  Prior to UD, he was with Harvard University as a Research Associate of Department of Statistics. He simultaneously earned Ph.D in Electrical and Computer Engineering from Cornell University and M.M. in Jazz Performance from New England Conservatory of Music after receiving M.S. in Electrical and Computer Engineering from Cornell University and B.S. in Electrical Engineering from Princeton University. He is an associate editor for IEEE Transactions on Image Processing and SPIE/IS&T Journal of Electronic Imaging, and served on the technical committee of IEEE SPS IVMSP as well as the organization committees of IEEE ICIP 2012 and IEEE ICASSP 2017. He has received a number of recognitions, including a paper award at IEEE ICIP 2007 and a number of keynote speeches. 

Stanley H. Chan,, 49-60230
Patrick Wolfe,, 49-41730