2009 Activities and Events

February 11th, 2009

Talk 2, 3: 10:00-12:00

Talk 2: A Signal-Processing Approach to Modeling Vision, and Applications

Talk 3: Incorporating Cognition into Image and Video Coding: A Task-based Approach

Abstract 2: Current state-of-the-art algorithms that process visual information for end use by humans treat images and video as traditional signals and employ sophisticated signal processing strategies to achieve their excellent performance. These algorithms also incorporate characteristics of the human visual system (HVS), but typically in a relatively simplistic manner, and achievable performance is reaching an asymptote. However, large gains are still realizable with current techniques by aggressively incorporating HVS characteristics to a much greater extent than is presently done, combined with a good dose of clever signal processing. Achieving these gains requires HVS characterizations which better model natural image perception ranging from sub-threshold perception (where distortions are not visible) to supra-threshold perception (where distortions are clearly visible).

In this talk, I will review results from our lab characterizing the responses of the HVS to natural images, and contrast these results with 'classical' psychophysical results. I will also present several examples of signal processing algorithms which have been designed to fully exploit these results.

Abstract 3: Many digital images are not acquired for documentary evidence or for archival representations, but instead serve as an intermediate piece of information to be ultimately used by a human to assess a situation, make a decision, or reach a conclusion. Often, the image need not provide a visually lossless representation of the original in order for a human to successfully perform a task, and many tasks can be easily performed with images that would be considered to be severely degraded in an aesthetic sense. To date, however, image assessment algorithms for natural images have been primarily focused on evaluating aesthetic quality. Such algorithms are designed using the implicit or explicit assumption that the target is the distortion and the image simply serves as a mask; the algorithms attempt to quantify the visibility of the distortion. When humans use degraded images, however, the target is the image content itself, and the distortions can be considered to be a mask, which may or may not impact the human¹s ability to perform the task.

Psychovisual experimentation in our laboratory has suggested that performance in low-level vision tasks such as detection and discrimination can yield results pertaining to higher-level vision. Motivated by these results, we have developed a experimental methodology to characterize suitability of an image for the recognition task. In this talk, I will describe our results from these experiments and our work toward developing recognition and utility measures for natural images. I will also describe our development of an intelligibility measure for American Sign Language video.

She is currently the Editor-in-Chief for the IEEE Transactions on Multimedia and a member of the IEEE Signal Processing Society's Board of Governors. She has chaired the IEEE Image and Multidimensional Signal Processing Technical Committee and has served on various technical and program committees. Prof. Hemami received a National Science Foundation Early Career Development Award in 1997 and has received numerous college and national teaching awards. She held the Kodak Term Professorship of Electrical Engineering at Cornell University from 1996-1999, and was a Fulbright Distinguished Lecturer in 2001 at the Faculte de Sciences in Rabat, Morocco. She has held visiting positions at Princeton University, Rice University, and EPFL.

She is a Fellow of the IEEE and member of Tau Beta Pi and Eta Kappa Nu.

https://foulard.ece.cornell.edu/SHemami.html