A Saliency Detection Technique Considering Self- and Mutual-Information
Ahmed Boudissa, Joo Kooi Tan, Hyoung Seop Kim, Takashi Shinomiya, Seiji Ishikawa · Institutional Repositories DataBase (IRDB) · 2014
In this paper, we present a novel approach to saliency detection. We define a visually salient region in an image with following two properties; global spatial redundancy, i.e., mutual-information, and local saliency, i.e., self-information or simply the region complexity. The former is its probability of occurrence within the image, whereas the latter defines how much information is contained within a region, and it is quantified by the entropy. By combining the global spatial redundancy measure and local entropy, we can achieve a simple, yet robust saliency detector. We evaluate it quantitatively and qualitatively. The comparison to Itti et al. [6], the spectral residual approach by Hou and Zhang [5], Achanta et al. [13] as well as to Zhai and Shah [14], on publicly available data shows a significant improvement.