Point of Gaze Analysis for Fixation Point Prediction

Mr. Umesh Rajashekar

Electrical and Computer Engineering
The University of Texas at Austin
Austin, TX

Monday, December 3rd, 9:00 AM, ENS 537

umesh@ece.utexas.edu

Abstract

One aspect of the Human Visual System (HVS) studied extensively for image and video processing applications, is the concept of foveation. Foveation refers to the non-uniform resolution perception of a visual scene by the human eye. The scene at the point of gaze is represented with very high resolution and the resolution falls off rapidly from the point of gaze. Foveation allows for a wide field of view without the accompanying data glut.

Concomitant with the development of a foveated vision system is the mechanism of eye movements that help the human eyes actively scan the world with the high resolution fovea. While foveation has been well modeled, eye movements are far from well understood.

In this talk, I will attempt to motivate the modeling of eye movements as an exercise in statistical analysis of image patches at points of gaze. Preliminary results using Principal Component Analysis of image patches at point of gaze have been used to generate linear kernels that may be used to predict eye fixations in novel scenarios. Also, the Discrimination Image Paradigm is presented to investigate the features used by humans in simple search tasks.

Biography

Umesh Rajashekar received the B.E. degree in July 1998 in Electronics and Communication Engineering from The Karnataka Regional Engineering College, India and the M.S. degree from The University of Texas at Austin in August 2000.

He is currently pursuing the Ph.D. degree at the Department of Electrical and Computer Engineering at UT-Austin. Since January 2000, he has been a research assistant in The Laboratory for Vision Systems at UT-Austin, where he is investigating image statistics at point of gaze to model human eye scanpaths. His interests also include developing didactic tools for education. He was awarded the 2000-2001 Texas Telecommunications Engineering Consortium Graduate Fellowship from The University of Texas.

A list of Telecommunications and Signal Processing Seminars is available at from the ECE department Web pages under "Seminars". The Web address for the Telecommunications and Signal Processing Seminars is http://signal.ece.utexas.edu/seminars