We introduce a novel region-of-interest (ROI) database for natural image content, the ROI-D database. The database consists of ROI maps created from manual selections obtained in a psychophysical experiment with 20 participants. The presented stimuli were 42 photographic images taken from 3 publicly available image quality databases. In addition to the ROI selections, dominance ratings were recorded that provide further insight into the interest of the selected ROI in relation to the background. In this paper, the experiment is described, the resulting ROI database is analysed, and possible applications of the database are discussed. The ROI-D database is made freely available to the image processing research community.
No citing papers are currently in WordNorms