Ing. Patrik Polatsek – Dissertation thesis
MODELLING OF HUMAN VISUAL ATTENTION
Degree Course: Applied Informatics
Author: Ing. Patrik Polatsek
Supervisor: doc. Ing. Vanda Benesova, PhD.
Slovak University of Technology Bratislava, FACULTY OF INFORMATICS AND INFORMATION TECHNOLOGIES, May 2019
In recent decades, visual attention modeling became a prominent research area. To simulate human attention, a computational model has to incorporate various attention mechanisms.
In this thesis, we explored how low- and mid-level features such as color, motion, depth and shape influence visual attention in our own eye-tracking experiments. To measure these effects, we utilized various state-of-the-art as well as novel computational models that estimate saliency of a specific feature.
To deeper understand the process of selective attention in everyday actions, we conducted several experiments in real environments recorded from the first-person perspective. Our results showed that egocentric attention is very individual and differs from 2D image viewing conditions, partially due to binocular cues that enhance viewer’s perception. We therefore suggest employing specialized saliency models for egocentric vision. Finally, we found out that high-level factors such as individual’s emotions and task-based analysis of visualizations influence human gaze behavior too.