Image sharpness evaluation based on visual importance

Abstract

This paper describes a method of image sharpness evaluation while taking into account the photographer's aesthetic intention. The main idea is utilizing a visual importance map that estimates the weight of each pixel to guild evaluating image sharpness. The visual importance map is computed automatically with a saliency detection algorithm based on global color contrast. Our technique allows to treat pixels in an image differently based on their content, such that the perceptually important features and photograph's subjective intention can be reflected in the result. The proposed method is validated by experiment on public data set.