Video Retargeting Based on Stretchability-Aware Blocks Scaling

This paper proposes an efficient approach to retarget video based on stretch ability-aware block scaling. The static stretch ability of the frame image is first evaluated based on gradient, saliency and color features. Then the dynamic stretch ability of the frame image is determined through the mapping of blocks under the constraint of optical flow vectors. Based on the analysis of frame image dynamic stretch ability measures, the original frame image is partitioned into non-stretchable blocks and stretchable blocks, and their scaling factors are calculated based on their stretch ability measures and the stretched video size, in order to possibly preserve non-stretchable blocks and reasonably resize stretchable blocks. Finally, the stretched video is uniformly scaled to generate the target video. Experimental results on a variety of videos demonstrate that our approach achieves a reasonable retargeting performance compared to the state-of-the-art video retargeting approaches.