Abstract:

Conventional cameras capture only a small fraction of the information carried by light. Rich properties such as spectrum, polarization, depth, coherence, and fast temporal dynamics are typically discarded, or require bulky instruments and sequential measurements to recover.

In this talk, I will describe a different approach to imaging in which photonics, sensors, and computational algorithms are designed together to capture much more of the information available from a scene. Inverse-designed and unconventional optical elements can encode otherwise inaccessible physical information directly into sensor measurements, while computational methods, including machine learning, decode these measurements into high-dimensional images. When combined with emerging sensor technologies such as event-based cameras, this approach can also capture dynamics at timescales far beyond those accessible to conventional frame-based imaging.

I will present examples spanning spectral and polarization imaging, ultrafast sensing, computational microscopy, and compact flat-optics systems, with applications ranging from biomedical imaging and industrial inspection to remote sensing and astronomy. More broadly, these results point toward a transition from cameras designed primarily to form images to information-aware imaging systems that jointly optimize the physics and computation for the information (or task) that ultimately matters. Such systems offer new opportunities to dramatically reduce size, weight, power, data volume, and latency while expanding what cameras can measure.