Efficient content-based cropping using visual saliency and SLIC for image retargeting

Abstract

This research presents an image retargeting scheme via content-based cropping. After evaluating the visual saliency of imagery content based on color differences in local areas, SLIC (Simple Linear Iterative Clustering) is employed to identify foreground and background regions. Cropping by following some pre-defined rules is then applied to remove insignificant boundaries. The resulting image is scaled directly to the target resolution if necessary. The experiments show that promising results can be achieved by the proposed design, when compared with existing methods, which are usually computationally expensive. This scheme may thus be feasible in video retargeting or applications requiring faster processing.