What is Computer Vision?
According to the International Data Corporation, in 1999 alone, consumers in the United States purchased about 10 million still and video cameras. Images and video clips are abundant on the Internet and the entertainment industry is becoming more sophisticated in digital manipulation of images. However, virtually all visual information today is communicated to the user in the from of raw (captured) images and videos. One of the goals of computer vision is to analyze these images to help the user (or an automated system) perceive and interact with the physical world.
Extracting the visual content of images either to assist perception or to facilitate transmission (decompose an image down to its building blocks, transmit each component separately, and re-compose the image) is a complex task. The visual appearance of objects changes as the viewing conditions (weather, light, pose) change. The three-dimensional world is projected down to a two-dimensional image and most of the times there is no knowledge about the projection transformation.
Research Interests
The majority of the computer vision algorithms are based on interpreting only high-level features (edges, corners, contours etc). My research exploits the interaction of light and matter at a per pixel basis (photometry) in order to extract from every single pixel as much information as possible. My previous work on binocular stereo, on object recognition and on segmentation has shown that visual information can be reliably extracted by looking at the individual pixel values. Current research efforts are concentrating on understanding the light reflectance of human skin, on multispectral imaging and on shape extraction. I am also exploring the development of new imaging sensors that will help the user extract visual information that is not captured by today's sensors and is imperceptible with the naked eye.
Research Grants
NSF-CAREER: Exploring the Multispectral Frontier in Computer Vision. PIs: Elli Angelopoulou. NSF: $393,920. 2002-2007.
Reflectance and Geometric Properties of the Human Face. PIs: Elli Angelopoulou. NEC Research Institute: $16,500. 2001.
NSF-ITR: Interacting with the Visual World: Capturing, Understanding and Predicting Appearance. PIs: Shree Nayar, Jitendra Malik, Peter Belheumer, Pat Hanrahan, Elli Angelopoulou, Kristin Dana. NSF: $3,499,997. 2000-2005.
Seminars
If you want to learn more about the field, attend one of our Multimedia, Vision and Visualization seminars.