垃圾邮件图像中的文字角点检测

Detecting corners of text in spam images

为提取垃圾邮件图像中文字的角点信息,提出一种新的基于图像边缘和圆形模板的角点检测算法。算法首先利用彩色边缘检测算子和阈值分割方法获取文字图像的边缘,然后采用圆形模板提取文字的角点信息。边缘检测和阈值分割降低了干扰背景和噪声对角点检测的影响,圆形模板使得角点检测对文字方向变化不敏感。实验表明,在真实的垃圾邮件图像中文字角点定位精度略高于SUSAN算法,并能同时获取角点角度的大小。

Combining edge extraction methods and a circular template,this paper proposes a new method for detecting corners of text in spam images.Firstly,the edges of text in spam images are extracted by color-roberts algorithm and threshold segmentation. Then,the corners is detected by using a circular template.Most noise in spam image is eliminated by edge extraction and threshold segmentation.Besides,the novel circular template allows the corner detection is insensitive to the orientation of text.A comparative experiment engages SUSAN method and the proposed method for the corners detection of text in images from a real spam image archive.The results indicate that the method has a better performance and can obtain the angle magnitude of corners contemporarily.