基于运动历史图的智能裁剪视频重定向

Video Retargeting Based on Motion History Image and Cropping

保持视频帧的时间连续性和重要对象的整体结构在视频重定向中至关重要。针对这一问题,提出一种基于运动历史图的智能裁剪方法进行视频大小的调整。首先基于时间轴将视频帧通过估计帧间摄像机运动对齐,从而构造全局运动模型;然后在全局运动模型的基础上,利用重要度图获得全局运动历史图;最后,基于运动历史图检测重要区域、裁剪重要区域并缩放至目标尺寸。运动历史图的应用不仅保证视频帧时间连续性,而且能处理包含复杂运动场景的视频。实验结果表明,提出方法针对复杂的摄像机运动和物体运动情况均能获得很好的视频重定向结果。

It is very important to keep time consistency of video frames and the overall structure of important objects in the video retargeting.To solve this problem,this paper proposes a cropping method based on motion history image to adjust the video size.Firstly,the video frames are aligned by estimating the inter-frame camera motion based on the time axis to construct the global motion model.Then,based on the glob⁃al motion model,the global motion history images are obtained by using the importance map.Finally,the important region is detected,cropped and scaled to the target size based on the motion history image.The application of motion history image not only ensures the tem⁃poral continuity of the video frames,but also can process video containing complex moving scenes.The experimental results show that the proposed method can obtain better results for complex camera motion and object motion.