A Saliency Map Model Based on the Principle of Maximum Information Gain
Takashi Toriu, Shigeyoshi Nakajima, Hiromitsu Hama · 2010
Humans do not process the entire area of an input visual image uniformly, but usually focus their visual attention on a limited area. But where does a human focus that attention? Psychological phenomena indicate that attention is attracted to features that differ from its surroundings or the one that are unfamiliar. On the bases of this we developed a saliency map model, which predicts where attention is attracted, based on two-step principal component analysis. In this paper, we reformulate this model based by the principle of maximizing information gain. On the basis of this theoretical foundation, we further investigate how this model could be modified in a situation where human must find something. We show experimental results and discuss future problems.