基于全景图像和状态匹配的视频修复算法  

Restoration Algorithm for Digital Video Based on Panoramic Image and State Matching

作　　者：王嵩[1] 李宏[1] 朱霞[1] 张卫[2] 

机构地区：[1]中南大学信息科学与工程学院,长沙410083 [2]湖南大学建筑学院,长沙410082

出　　处：《数据采集与处理》2009年第5期694-698,共5页Journal of Data Acquisition and Processing

基　　金：中南大学博士后基金(2008)资助项目

摘　　要：提出了一种针对数字视频中破损区域的修复算法,该算法适用于修复背景固定且前景近似周期性运动的破损视频序列。修复破损背景时,在建立的背景全景图像中直接复制信息,填充破损;修复破损前景时,采用了基于纹理合成的图像修复算法,并根据视频序列的特点将匹配块的搜索范围扩大到运动状态相似的其他帧。实验结果表明:本文算法可以较为准确地还原出破损处的真实场景,比仅使用单帧图像修复算法具有更合理、视觉上更真实的修复效果。Aimed at the damaged area in digital video, a restoration algorithm is proposed. The extensive applications of the algorithm are found in restoring damaged video sequence with fixed background and approximately periodical motion foreground. While restoring the dam- aged background, the information is directly replicated to fill in the damaged region in the constructed background panoramic images. And when it comes to restoring the damaged foreground, the image restoration algorithm based on the texture composition is used, and the matching block area is naturally extended into the other fames with similar motion status according to the features of video sequences. Experimental results show that the proposed algorithm can well revert the original scenery in the damaged area. Compared with the only used single frame restoration algorithm, the algorithm is more reasonable and achieves better restoration effects.

关 键 词：视频修复 背景全景 图像修复 状态匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]