A semantic aware image retargeting scheme combining seam carving and non-homogeneous warping
Yu Gong, Lifang Wu, Xiuzhen Zhang · 2010
In this paper we propose a novel approach to semantic aware image retargeting combining seam carving and non-homogeneous warping. We extract the playing ground of basketball matches as the semantically important region and compute the Semantic Weight Function (SWF). We improve seam carving and warping using SWF. Furthermore, we combine seam carving and warping coherently. Our scheme starts from the seam carving operation on the original image. After a seam is removed, we compute Deformation of Semantic Edges (DSE). Seam carving is continued until DSE is larger than the assigned threshold. If the size of a current image is not the target size, it is further adapted to the target size using non-homogeneous warping. We compare our approach with scaling, seam carving and semantic aware seam carving. Experimental results show that our approach preserves more semantically important regions with less deformation.