My research interests include computational imaging, computer vision, computer graphics and color science.
Due to the per-pixel “overhead” time consumption, modern sensors face a fundamental space-time resolution trade-off. The more pixels to sample, the more time to take. This issue becomes significant if high performance (high spatial and temporal resolution) video sensing is the goal. Compressive video is one of the research topics to overcome this issue. The basic assumption is that data is redundant and “compressible”. Thus, it is possible to sample data at a much lower ratio while maintaining a reasonable fidelity.
- compressive holography
This project seeks to answer the question: how do we encode a 4-dimensional (3D position and time) scene into a single image and how to recover it?
- lens-free on-chip video
One of the applications of compressive video is to monitor in vivo biological samples. This task requires an imaging platform with large field of view and high resolution. The lens-free on-chip setup stands out due to its compactness and cost-effectiveness. In biomedical imaging, people are more interested in phase information of the object as it potentially reveals more information. Thus and furthermore, how do we recover phase when a compressive sampling scheme is applied?
- image stitching
Photos taken at low light and high light conditions sometimes have different color tones, although they were part of the same scene. This tone mismatch will become an issue when we attempt to stitch the photos together. This project looks into this problem.
Human visual system is a distinctive “imaging system” that we humans have been taking efforts to understand. An approach to analyze our visual system is to quantitatively “measure” human visual response, i.e., establishing a link between physical measurements and visual outputs. Appearance is one of the visual attributes challenging to quantify. Specifically, appearance can be divided into object appearance and scene appearance.
Recent advances in Bidirectional Reflectance Distribution Functions (BRDF) acquisition have provided a novel approach for appearance measurement and analysis. In particular, since gloss appearance is dependent on the directional reflective properties of surfaces, it is reasonable to leverage BRDF for gloss evaluation. This project investigates gloss appearance from both soft metrology and hard metrology.
- glint impression and diffuse coarseness
Glint impression (sparkling) and diffuse coarseness (graininess) are two special surface effects widely discussed in car finishes industry. We looked into these two effects by conducting 3 psychophysical experiments.
- rendering color correlated temperature (CCT)
One aspect of visual appearance is scene appearance. This project seeks to render indoor scenes at different color temperatures.