|
|
Intelligent Video Processing
Abstract
The goal of intelligent video processing is to analyze a video sequence
and extract information that benefits an application. The extracted
information might lead to higher compression ratios (by
intelligent bit allocation), data manipulation, or content analysis and
management, such as video summarization, indexing and retrieval.
It could also enhance an image by removing noise, improving resolution
or increasing the dynamic range.
The first step in the analysis of a video sequence is segmentation.
For example, foreground objects could be separated from the background.
This is important for compression applications, which might allocate
more bits to the object of interest. It is also important for the
interactive manipulation of content.
The analysis of an object's shape, size, position, color and texture
also provides critical information for content analysis
and understanding. For example, video sequences might
be classified into categories, such as "head and shoulders,"
"landscape," outdoor scene with people," etc. Such information
facilitates image compression applications as well as summarization,
indexing and retrieval.
Video segmentation is accomplished through the use of spatio-temporal
and disparity information. Object location and tracking requires
additional constraints that can be implemented by techniques such as
Kalman filtering. Finally, the estimation of object dimensions and
position requires camera calibration.
Students
- Stavros Tsavidas (M.S. - June 2001)
- Peshala Pahalawatta
- Dejan Depalov
Publications
- S. Tzavidas, A.K. Katsaggelos, "Multicamera Setup for Generating
Stereo Panoramic Video," Proc. 2002 SPIE Conference on
Visual Communications and Image Processing, San Jose, CA,
January 2002.
Theses
- S. Tzavidas, "An Immersive Virtual Reality System," M.S. Thesis,
Department of Electrical and Computer Engineering, Northwestern
University, June 2001.
More Information... |
|
|
