Computational attention model for children, adults and the elderly

• Published in: Multimedia tools and applications Vol. 80; no. 1; pp. 1055 - 1074
• Main Authors: Krishna, Onkar, Aizawa, Kiyoharu, Irie, Go
• Format: Journal Article
• Language: English
• Published: New York Springer US 01.01.2021; Springer Nature B.V
• Subjects: Adults; Age; Computer Communication Networks; Computer Science; Computer vision; Data Structures and Information Theory; Eye movements; Image processing; Intersections; Model accuracy; Multimedia; Multimedia Information Systems; Observers; Older people; Salience; Special Purpose and Application-Based Systems; Viewing; Visual fields; Visual perception; Saliency; Human visual system; Fixation dispersion; Eye-tracking; Depth bias
• ISSN: 1380-7501; 1573-7721
• DOI: 10.1007/s11042-020-09474-5

Computational models of saliency estimation have been studied in a wide range of research fields, including visual perception, image processing, computer vision, multimedia, and their intersections. However, most of them seek to simulate scene viewing by adults only, and the impact of observer’s age has rarely been considered. In this paper, we quantitatively analyze age-related differences in gaze landing positions during scene viewing. From the results, we draw the following three conclusions: child observers focus more on the foreground in a scene, i.e., locations that are near, while elderly observers tend to explore the background, i.e., locations farther in the scene; adult observers are more explorative than child and elder ones; and adult observers have significantly lower center bias compared to child and elderly observers. Based on these observations, we developed a novel computational model for age-dependent saliency estimation. The prediction accuracy suggests that our model better fits collected eye-gaze data of observers belonging to different age groups than several existing models do.