视频图像内文字的自动提取新方法

提出了一种基于彩色边缘检测、形态学和逻辑与运算的视频文字提取方法,首先,分别运用垂直、水平、对角三个方向的彩色边缘检测算子提取出原图的边缘图像,然后分别对上述三幅边缘图像进行数学形态学运算,得到三幅不同方向的连通域图。最后,将这三幅连通域图相与,消除无用的信息,得到最终的文字区域。实验证明,该方法具有高的文字提取率和好的鲁棒性。

A novel approach for automatic text extraction from video images

A novel text entraction method is proposed. This method is based on color edge extraction, morphology and a logic operator "AND". First, Three color edge extraction operators of vertical, horizontal and diagonal directions are used separately to get three edge images from the original image, then, further morphological operations are used to get three images of connective regions of defferent directions from the above three edge images. Finally, useless information is greatly decreased by the integration of the connective images and then the last object text region is acquired. Experiments show that the proposed method doesn't only have high text extraction ration , also has good robust.