Bi-directional Carving Based on Saliency Map via Absorbing Markov Chain

Traditional methods for image retargeting such as scaling and cropping cannot generate satisfactory results because of their neglect of the content of the image. In order to consider the content of the image, saliency detection for finding the regions of interest (ROIs) plays an important role in the process of retargeting an image. In this paper, we present a novel image retargeting method using bi-directional seam carving technique combining scaling. Guided by an important map based on a saliency map via absorbing Markov chain combining with gradient map, the proposed method can effectively protect the ROIs in the image when downsizing or enlarging an image. Extensive experiments demonstrate the efficiency of the proposed method against the traditional methods and several other methods.