摘要
提出了一种基于三维泊松方程的数字视频修复算法.首先提取视频帧的梯度场,然后使用基于块的修复方法对梯度进行信息填充,最后在梯度场中建立视频受损区域的三维泊松方程并求解,实现对视频像素的修复.实验结果表明,基于三维泊松方程的视频修复方法具有良好的时空连续性,能保证视频画面色调、明暗的一致性,使修复痕迹更平滑,画面更自然.
A new video inpainting method based on three-dimensional Poisson equation was presented.The gradient fields of video frame were extracted first.Then,the target areas of every gradient fields were filled in the patch-by-patch inpainting methods.Afterwards,to repair the video,a three-dimensional Poisson equation was built and solved in the gradient field of the target area.Experiment results have shown that the proposed method can achieve the desired results and perform better than the existing methods in color,light and shade consistency.
出处
《湖南大学学报(自然科学版)》
EI
CAS
CSCD
北大核心
2011年第3期51-55,共5页
Journal of Hunan University:Natural Sciences
基金
国家自然科学基金资助项目(60871096
60835004)
教育部博士点基金资助项目(200805320006)
教育部科学技术研究重点项目(2009-120)
模式识别国家重点实验室开放课题基金资助项目
关键词
图像修复
视频修复
泊松方程
梯度场
image inpainting
video inpainting
Poisson equation
gradient fields