Research Scientist, Columbia University
Interests: Computer Vision, Computational Imaging
Research Statement: My objective is to build computer vision systems that solve real-world problems and work robustly "in the wild". I believe that vision systems perform the best when their two components - image acquisition and image interpretation (inference) - work together. Based on the two components, my research has two themes:
- Building high-capability computational cameras that capture maximal amount of scene information in images. To design such cameras, I leverage advances in optics, electronics, sensing and programmable illumination hardware.
- Developing algorithms for scene inference using the physics and statistics of how light interacts with the world.
The goal is to design each component while being aware of the other, so that they function synergistically. My work has applications in autonomous transportation, 3D imaging for factory automation, consumer imaging, digital cultural heritage, robotic surgery, performance capture and human-computer interfaces.
Research Highlights (click on thumbnails for details)
Summary: Outdoor 3D scanning in strong ambient illumination with a limited power-budget. Applications in autonomous transportation, outdoor robotics and urban mapping.
Summary: 3D scanning of optically challenging scenes, such as with translucent materials (e.g., marble, human skin) and concave objects. Applications in digital cultural heritage and performance capture, factory automation and indoor mapping.
Summary: Structured light 3D scanning for scenes with concave geometries (e.g., indoor spaces, concave machine parts) and nearly transparent objects (e.g., thin plastic films).
Summary: High resolution 3D scanning for specular materials (e.g., metallic machine parts), intricate geometries (e.g., finger-prints) and objects with sharp depth discontinuities.
Summary: Computational model for light propagation in scenes with complex light transport effects such as defocus, scattering and interreflections. Applications in scene understanding (estimating depth and material properties of scenes).
Summary: High dynamic range imaging for moving scenes and cameras by capturing images with Fibonacci sequence exposures. Applications in consumer imaging.
Summary: High speed compressive imaging (up to 1000 fps) using a conventional 30 fps camera, an LCoS (liquid crystal on silicon) light modulator and large scale dictionary learning. Applications in consumer and scientific imaging.
Summary: Post capture control of spatial resolution and frame-rate of videos, depending on the scene content. Applications in consumer imaging, scientific imaging and microscopy.
Summary: Performance analysis of computational imaging techniques such as defocus and motion deblurring and light field capture. Practical guidelines for imaging system design.
Summary: Seeing clearer and farther in scattering media using active illumination and polarization. Applications in underwater imaging, autonomous transportation and medical imaging.
Summary: Reduced space framework for fast simulation and rendering of scattering media such as smoke and fog. Applications in gaming, special effects and virtual reality.
Summary: Imaging and illumination setup for measuring scattering properties of liquids. Applications in computer graphics rendering and ocean monitoring.