Video Retargeting Based on Spatiotemporal Saliency Model

摘要

This paper proposes a video retargeting method based on spatiotemporal saliency model. Global histograms and regional histograms are first calculated in both spatial domain and temporal domain, respectively, and then the spatiotemporal saliency model is built by evaluating the contrasts between the global histograms and each regional histogram. Based on the spatiotemporal saliency map, a salient object detection method is used to locate salient object regions in the video. Finally, cropping and uniform scaling operations are performed on the basis of salient object regions to generate the retargeted video. Experimental results on a variety of videos demonstrate that the proposed approach is simple and effective, and can preserve important objects well and achieve a better retargeting performance.