基于图像分析的数字视频字幕检测及恢复

摘要

Video has been widely applied in various fields, therefore to process video captions has its great significance. This paper proposes a new algorithm for the detection and extraction of digital video caption. The first step is to conduct edge detection, through analysing the edge density projection of the image in both horizontal and vertical directions, the captions sketchy region is obtained. Then the level-set method is used to further determine the specific location of the caption. Finally the detected caption is restored with Criminisi algorithm.Through the experimental verification on different original video materials with the algorithms presented in the paper, all the results indicate that this method can achieve ideal effect.%视频已广泛应用于各个领域,所以对视频字幕的处理具有重要的意义.提出一种新的数字视频字幕检测及恢复算法.首先进行边缘检测,通过分析图像的边缘密度在水平和垂直方向的投影,得到字幕的粗略区域;再利用水平集方法进一步确定字幕的具体位置;最后用Criminisi算法对检测到的字幕进行恢复.通过对不同类型的视频素材进行实验验证,结果表明该方法能够达到理想效果.