基于重要度扩散和自适应采样的图像/视频缩放技术

摘要

提出了一种基于重要度扩散和自适应采样的图像和视频自适应缩放方法,它在整体概貌和重要区域保护之间进行折衷处理.重要度扩散函数将删除像素的重要性扩散至其邻域,以避免过多删除非重要区域而造成图像整体概貌失真.自适应采样函数则通过对各行和列像素的重要性进行权值的采样,以保护重要区域.此外,通过引入帧间一致性约束,该算法也适合于视频缩放.仿真实验结果表明:与剪切、Seam Carving等方法相比,本算法取得了较好的缩放效果.%An image and video resizing algorithm was proposed based on importance map diffusion and adaptive scaling. It compromised between the protections of image content and important regions. To avoid the image distortion by excessive deletion of un-important regions, the importance of current row and column was spread to their neighboring pixels. To protect importance regions, different weighting coefficients were allocated to every row and column in terms of their importance by adaptive scaling function. Moreover, it was also effective for video resizing by incorporating the constraint of temporal coherency. Experimental results show that compared with other techniques such as cropping and seam carving, satisfactory results were achieved by the proposed approach.