摘要
在研究Hu矩和Zernike矩的基础之上,提出一种基于Hu矩和Zernike矩的文字识别方法,在采集的文字图像中提取Hu矩和Zernike矩特征,利用K近邻的分类方法进行分类,理论和实验表明,该识别方法具有很强的抗图像平移、拉伸和旋转识别能力,其中Zernike正交矩由于其正交性在具有较高的识别能力的同时,还具有很强的冗噪能力。
In this paper, a character recognition method was proposed based on the Hu moments and Zemike moment. The proposed method extract the moments after collection the characters image, and then used K-neighbor classifier to classify it. Theoretical and experimental result shows the high classification accuracy of this approach in images＇ translation, scaling and rotation. The Zernike moments has high noise robust as well as high recognition accuracy.
出处
《科技信息》
2009年第17期62-63,共2页
Science & Technology Information
