Generation of an importance map for visualized images

Abstract

Visualized images have always been a preferred method of communication of information contained in complex data sets. However, information contained in the image is not always efficiently communicated to others due to personal differences in the way subjects interpret image content. One of the approaches to solving this issue is to determine high-saliency or eye-catching regions/objects of the image and to share information about the regions of interest (ROI) in the image among researchers. In the present paper, we propose a new method by which an importance map for a visualized image can be constructed. The image is first divided into segments based on a saliency map model, and eye movement data is then acquired and mapped into the segments. The importance score can be calculated by the PageRank algorithm for the network generated by regarding the segments as nodes, and thus an importance map of the image can be constructed. The usefulness of the proposed method is investigated through several experiments. © 2009 Springer-Verlag.