Detecting Visual Attention and Its Application

Abstract

The visual focus of attention is an important clue for inferring our internal state and, therefore, can be used effectively for developing human-centric media such as interactive advertising, intelligent transportation systems, and attentive user interfaces. Since our visual focus of attention is closely related to our gaze, many gaze sensing techniques have been developed; however, it is still difficult to measure our gaze in casual and unconstrained settings. An alternative way of estimating the visual focus of attention is to use a visual saliency map model. The saliency map identifies regions that draw human visual attention by hierarchically computing differences of features from the surroundings for an image or a video in multiple resolutions and then fusing them in a fully bottom-up manner. In this talk, I will present our recent work on extending the existing saliency computation model: adaptation to egocentric videos, incorporation of visual field characteristics, and usage of auditory cues for effective computation of the saliency map. I will also present some potential applications of the saliency map.

Bio

Akihiro Sugimoto received his B.E, M.E., and D. Eng. degrees in mathematical engineering from the University of Tokyo. After working at the Hitachi Advanced Research Laboratory, ATR, and Kyoto University, he joined the National Institute of Informatics, Japan, where he is currently a Professor. From 2006 to 2007, he was a visiting professor at Pari-Est University, France. He is interested in mathematical methods in engineering. In particular, his current main research interests include discrete mathematics, approximation algorithm, vision geometry, and modeling of human vision.