Research in the Visual Information Engineering lab (VINE Lab) broadly concerns analysis, coding, and processing of visual information. These days, there are many sources of visual information, including digital images, video, 3D images/video, computer-generated content, and composited versions. Our research explores how these sources of information can be utilized to help society. Our overarching goal is to research and develop software and systems:
- to make fast and reliable decisions from visual sources, and/or
- to assess/ improve the appearance, security, and usefulness of the visual content.
A key theme of our research is to consider both the computational perspective and the perceptual perspective; this approach allows us to engineer models and algorithms that are aware of how the visual information is perceived by humans, and how that perception is altered based on changes to the source content, the viewing environment, and the task at hand.
Our key research topics include:
- Image/video enhancement, restoration, and compression via perceptually guided and/or machine-learning based methods
- Quality assessment of natural and synthetic images, video, 3D content
- Traditional and AI-based analysis, including detection, segmentation, and classification
- Computational modeling of the human visual system using natural-scene statistics and visual psychophysics
- Real-time analysis and processing
Some applications of our work including automatic detection and scoring of streamed visual content, perceptually lossless compression and watermarking, visual guidance for the blind, and detection, segmentation, and correction of driving video.