Gradient-based global features and its application to image retargeting

We propose gradient-based global features and its application to image retargeting. The proposed features are used for an importance map for image retargeting, which represents rough location of salient objects in an image. We focus on areas rather than points and lines to be assigned as an important part. The information about areas in multiple layers provides global features. Experimental results compared to the state-of-the-art salient features for image retargeting demonstrate the effectiveness of the proposed features.