一种身份证图像字符分割的改进方法

An Improved Method of Character Segmentation for ID Images

为了提高二代身份证图像中文字信息的正确识别率,研究了一种改进的身份证字符分割方法。该方法首先利用Radon变换对不规则的身份证照片进行水平矫正,然后根据身份证图像的颜色特点,提出了动态选取红绿蓝三通道阈值的理论,并通过阈值比较对身份证图片进行颜色筛选,最后分别用水平投影和垂直投影的方法对图像中的汉字信息进行块分割和字符分割,在切割过程中,加入了防错机制,增加切割鲁棒性。实验表明,与传统方法相比,身份证图像字符分割的改进方法降低了对照片质量的要求,具有较高的切割成功率,是一种较理想的身份证切割方法。

In order to improve the recognition rate for the information on the second generation ID card,the paper presents a method for character segmentation. Firstly,the horizontal correction is conducted for the irregular photo ID card by Radon transform,then according to the color characteristics of the ID card,the theory of the three RGB dynamic threshold is creatively put forward,and colors of ID pictures is chosen by comparing thresholds. Finally,the horizontal and vertical projection are respectively used to split the information blocks and characters,in the cutting process,error- proofing mechanism is added to increase the robustness of cutting. The experiment results show that,comparing with the traditional method,the improved method reduces the quality requirements for the ID image,has a high cut power and is an ideal identification method of cutting.