Multi-operator Image Retargeting with Preserving Aspect Ratio of Important Contents

摘要

Content-aware image retargeting can preserve the quality of visually important objects during image resizing. However, some previous approaches fail to provide desired results as the sizes of the important objects in original images are larger than the target resolution of resized images. In this paper, we propose a novel multi-operator image retargeting scheme, which combines seam carving (SC) methods with uniform scaling techniques. To guarantee the quality of visually important objects, we stretch original images in both vertical and horizontal directions and then perform similarity transformation by using the indirect SC and uniform scaling methods. Besides, the direct SC method with gradient vector flow (GVF) is also employed to shrink the image finally. Experiments demonstrated that the proposed scheme could produce more desirable resized images than the existing methods.