Depth from Differential Defocus

Defocus reveals object location in cameras and microscopes. Inspired by the unique anatomy and behavior of the jumping spider, we explore differential defocus changes that enable efficient computation of depth, velocity, and phase. We develop a family of depth sensors for a variety of imaging settings, using standard cameras, deformable lenses, metalenses, and microscopes.

Best Student Paper ECCV 2016, Best Demo ICCP 2018

US Patent Application No. 62/928,929 - licensed to Metalenz

ICCP 2021, PNAS 2019, Dissertation 2019, ICCP Demo 2018, IJCV 2017, ICCV 2017, ECCV 2016, ICCV Workshop 2015

[TIE project page] [metalens project page] [focal track project page] [focal flow project page]


[media coverage 1 2 3 4 5 6 7]

Motion Estimation

Optic flow provides key motion cues, but the underlying brightness constancy constraint is often violated in real-world settings. We explore two kinds of mitigation strategies for brightness constancy violations: first, we show that explicit modeling of defocus-based violations can reveal depth and motion simultaneously; next, we explore spatial sampling techniques for robust self-motion estimation in generic and natural scenes, explaining biases found in larval zebrafish brains and behavior.

CB 2022, VSS 2021, ICCV 2017, ECCV 2016

[zebrafish project page, press release] [focal flow project page]

Shape & Color

The human visual system explains pixel-to-pixel changes with a combination of material, geometric, and lighting-based explanations. We show that aligning color changes with shading cues disrupts shape perception, particularly around critical contours.

Interface Focus 2018, VSS 2013, VSS 2013

[project page] [paper]