PixelView
What if a graphics card could simultaneously generate all views of a scene, giving a person in a room full of viewers a unique perspective? The PixelView project presents a forward looking architecture for a point-based hardware design for sampling, rendering, and displaying polygonal, procedural, and point cloud primitives within a 4D frame buffer.

Video Enhancement Using Per-Pixel Virtual Exposures
How many poorly exposed home videos have been shot that capture events we would like to watch, but cannot due to overwhelming noise and dark details? By simulating an optimal virtual exposure time for each individual pixel, it becomes possible to bring out details that were barely visible in the original source. By accumulating multiple temporally and spatially aligned pixels, we can simultaneously enhance and tonemap video without intermediate HDR representations.

Video and Image Bayesian Demosaicing
The demosaicing process converts single-CCD color representations of one color channel per pixel into full per-pixel RGB. We introduce a Bayesian statistical technique for demosaicing Bayer color filter array patterns that is based on a statistically-obtained two color per-pixel image prior. By modeling all local color behavior as a linear combination of two fully specified RGB triples, we avoid color fringing artifacts while preserving sharp edges.

HDR Tone Mapping For Fine Feature Preservation
This work presents an HDR tone-mapping technique that preserves "fine" image features (sharp specular highlights, thin structures, and high contrast textures) by using techniques originally developed for shot-noise reduction.

Multispectral Video Fusion
To create high-quality night-vision video, this work combines dark and noisy visible spectrum video (RGB) with well-exposed infrared (IR) video to enhance the quality of the RGB input. This differs from traditional video fusion where IR heat data is overlaid on the RGB, both of which are assumed to contain little noise. With our method, correct daylight chrominances and natural luminance ratios are maintained while reducing noise levels.

Computational Time-Lapse Video
Time-lapse videos can be created by starting with video-rate footage and temporally downsampling the video. In this manner, non-uniform samplings can be made to optimize for motion seen throughout the video and to create synthetic exposures of many video-rate frames in a post-process. A non-uniform sampler which finds the optimal time-lapse video and a virtual shutter for exposure generation are both presented.

Education
Ph.D. in Computer Science from The University of North Carolina at Chapel Hill, Chapel Hill, NC 2002-2007
M.S. in Computer Science from The University of North Carolina at Chapel Hill, Chapel Hill, NC 2002-2004
B.S. in Computer Engineering from Case Western Reserve University, Cleveland, OH, 1998-2002
Minor in Drama Performance from Case Western Reserve University, Cleveland, OH 1998-2002
Graduate of St. John's Jesuit High School, Toledo, OH 1994-1998