A Method of Calculating Image Saliency and of Optimizing Efficient Distribution of ImageWindows

In this paper we propose a method of calculating the saliency map of an input image. In this method the saliency gets higher in the areas where the image features are different than those in the surrounding areas and where the features are unfamiliar. Based on this method, we also propose a method to apply the idea of calculating the saliency to the problem of determining the most efficient distribution of image windows to acquire the information. This distribution might be considered to correspond to the distribution of attention. We conducted two experiments. The first is the experiment to apply the saliency model to a moving image and shows that in the area where the feature about motion is different than those in other areas the saliency gets higher. In the second experiment it is shown that a definite number of image windows to acquire the information can be efficiently distributed owing to the local image features