Content-aware Stereoscopic Image Retargeting Using Non-homogeneous Warping

In this study, a content-aware stereoscopic image retargeting approach using non-homogeneous warping is proposed. The proposed approach contains three major stages: preprocessing, optimization, and transformation. In the preprocessing stage, the initial scaling factor map, the disparity map (by stereo matching), and the importance map of each input stereoscopic image pair are computed. In the optimization stage, to find the optimal scaling factor maps S*, stereoscopic image retargeting is to minimize an energy function combining image and stereo quality energies. Based on the optimal scaling factor maps, in the transformation stage, the retargeted stereoscopic image pair is obtained by pixel fusion. Based on the experimental results obtained in this study, the performance of the proposed approach is better than those of three comparison approaches.