摘要
在文档影像的自动处理中,去黑边和倾斜校正是影像预处理的首要环节。该文提出了变黑边模板的概念和基于区域填充的黑边去除算法。对于图像的倾斜校正,我们提出了基于方向投影的表格线检测方法,并由此实现图像的自动分类;对不含表格线的图像,文中将字符包围盒中心作为特征点,采用Hough变换的算法进行倾斜检测。另外,倾斜检测时还采用金字塔模型降低图像分辨率,进一步提高了算法速度。实验表明,该文的方法能够有效地去除图像黑边,快速准确地检测出图像的倾斜角,并具有很强的抗干扰性和应用适应性。
In automatic document image processing, black margin removal and skew correction are two principal steps. We introduce the concept of adaptable mask on judging a noise point and present a method based on region filling to remove the black margin effectively. As for skew correction, this paper introduces a new form - line detection method based on directional projection. If an image has no form - line, we use Hough transform to compute its slant, taking centers of the bounding boxes of some characters as the eigen - points. Additionally, we get low resolution image by pyramid processing to promote the algorithm speed of skew detection. Experiments show that our methods can reduce black margin noise well and can detect the slanting angle of an image rapidly and accurately, and also have high noise endurance and application adaptability.
出处
《计算机仿真》
CSCD
2005年第10期208-211,276,共5页
Computer Simulation
关键词
黑边
倾斜校正
方向投影
包围盒
霍夫变换
Black margin
Skew correction
Directional projection
Bounding box
Hough transform
