A Statistical Method for Content-Aware Image Shrinking

This paper proposes a statistical method for content-aware image shrinking. This method divides image into equal quads by an initial uniform grid and computes an important factor for each of quads guided by importance map. The importance map is generated automatically using a novel combination of image edge density and Harel saliency measurement. The method then computes scaling factor for each of quads by using of statistical measures such as the mean and the variance of quads importance factors. To preserve the important areas as far as possible and to destroy the unimportant areas more, this method uses statistical functions to give the larger value to important quads. To prevent bending of straight horizontal lines, in the vertical direction, the algorithm equally resizes quads that are placed in the same row. Also, to prevent bending of straight vertical lines, the algorithm equally squeezes and stretches quads that are placed in the same column. Experimental results of different types of images show that our method can achieve better objective and subjective quality with preserving prominent objects.