Leveraging Motion in Video Analysis

Dr. Richard Souvenir

Computer Science Dept., UNCC

Friday, October 19, 2007

3:00 p.m. - 4:00 p.m.

Woodward Hall, Room 106

Complete Description:

In automated video analysis, the motion of interesting objects is often treated as a distracter which, ideally, would be factored out. These approaches are quite common (e.g., image stabilization, motion normalization). In this talk, I discuss how motion cues can be useful, specifically in the cases of medical image segmentation and human action recognition. This work uses ideas from the field of manifold learning, which provides powerful tools for parameterizing high-dimensional data points (in this case, images) using a few parameters when this data lies on or near some manifold. I demonstrate results on an image segmentation problem of cardiopulmonary MR data. Bio: Richard Souvenir is an assistant professor in his 2nd year in the Department of Computer Science at UNC-Charlotte. His research work is in the fields of computer vision and machine learning. He completed his doctoral work at Washington University in St. Louis where he was a National Science Foundation Graduate Student Fellow.