Event Location: NSH 1507
Bio: John Oliensis received his PhD in theoretical particles physics from the University of Chicago and carried out research in physics at Princeton University, the Fermi National Accelerator Laboratory, and the Argonne National Laboratory. He began research in computer vision in 1988, joining the University of Massachusetts at Amherst as a member of the research faculty.  From 1994–2003 he was a research scientist at the NEC Research Institute, where he organized three workshops bringing together researchers in computer vision, human vision, neuroscience, and learning. Since 2003 he has been an associate professor in the computer science department at Stevens Institute of Technology. His interests include the estimation of object shape from images, perceptual organization, the recognition of objects, and human vision. He is a senior member of the IEEE and an Associate Editor of PAMI.

Abstract: People understand an unexpected new image amazingly fast. How can an artiʂ57;cial recognition system achieve this? Given an unknown image, the system must start by ʂ57;nding image structures distinctive enough to focus in on plausible interpretations. Researchers agree that this requires perceptual organization (PO), the identiʂ57;cation of mid-size image structures such as salient lines, regions, or object shapes. Yet the intrinsic ambiguities of images make PO unreliable, and systems based on PO have not worked well.

I present a recognition method that exploits PO without succumbing to its unreliability. It recognizes objects by shape, comparing shapes between images by comparing segmentations. To deal with unreliability, it computes the average similarity between all possible segmentations of the two images weighted by probability. Other applications include tracking moving objects in video, image segmentation, and edge-preserving smoothing.