Image Retargeting by Combining Region Warping and Occlusion

Image retargeting aims to avoid visual distortion while retaining important image content in resizing. However, current image retargeting methods usually fail in difficult situations, such as complex structures and intricate arrangements of objects. In this paper, we propose a novel image retargeting approach, which provides low visual distortion retargeting results by combining region warping and occlusion. We first represent image region with occlusion probability to avoid inaccuracy in image decomposition. Then we combine region warping and occlusion in a unified framework, and transform it to a pixel-level retargeting problem. To verify the performance of our approach, we implement it with two operators, seam carving and pixel warping. Experiments demonstrate the effectiveness of the proposed approach.