Event
Ph.D. Research Proposal Exam - Janith Bandara Senanayaka
Monday, April 20, 2026
10:00 a.m.
AVW 1146 (ISR)
ANNOUNCEMENT: Ph.D. Research Proposal Exam
Name: Janith Bandara Senanayaka
Committee:
Professor Christopher Metzler (Chair)
Professor Behtash Babadi
Professor Saikat Guha
Date/time: Monday, April 20th at 10:00 AM
Location: AVW 1146 (ISR)
Title: Fundamental Limits of Computational Imaging in Extreme Conditions
Abstract: Imaging has shaped human understanding for centuries, from the camera obscura to modern scientific instrumentation. Yet many important phenomena remain difficult—or impossible—to observe directly. Light scatters, photons are scarce, measurements are incomplete, sensors saturate, and physical systems impose constraints. In these regimes, conventional imaging fails, leaving only indirect, noisy, or highly distorted measurements.
Computational imaging addresses these challenges by integrating physical modeling, signal processing, and machine learning to reconstruct high-quality images or infer useful information from such distorted or incomplete data. By shifting complexity from hardware to computation, this approach not only expands the limits of what can be imaged but also reduces hardware costs and compensates for physical imperfections, making systems robust and adaptive.
To expand these boundaries further, this research confronts imaging challenges in two extreme physical regimes. First, in the optical domain, we explore the fundamental limits of macroscopic synthetic aperture (SA) imaging. While SA methods like Fourier ptychography excel in microscopy, extending them to long-range observation is significantly challenging due to speckle noise. We prove that sequential SA imaging of fully diffuse scenes is impossible when corrupted by per-measurement independent speckle. To overcome this theoretical limit, we introduce a snapshot SA framework and an aperture phase synchronization strategy. Remarkably, we show that speckle—traditionally treated as a nuisance—can instead be exploited to recover missing spatial frequencies in distributed, non-overlapping aperture systems, surpassing conventional resolution limits.
Second, we propose to expand the boundaries of X-ray computed tomography (CT) to photon-starved reconstruction with material characterization. We first aim to study X-ray CT under extreme physical constraints, where projections are sparse, photon-starved, and degraded by motion blur. While conventional medical X-ray CT has made significant leaps, it still relies on high photon counts; in contrast, we aim to push this to unprecedented low-photon extremes. In parallel, we integrate multi-energy spectral capabilities to enable material composition inference. By combining photon-starved scanning with material differentiation, this research provides transformative cross-disciplinary applications, ranging from long-range security screening and industrial non-destructive monitoring to ultra-low-dose medical CT that can minimize patient radiation exposure while maintaining high diagnostic specificity.
