字符图像识别的分布加权模板匹配方法研究

Research on Distribution-weighted Template Matching Algorithm of Character Image Recognition

使用标准模板匹配识别算法对图像中的字符进行识别时,图像中的背景噪声会导致识别准确率较低。为了提高识别准确率,提出一种基于分布加权的模板匹配识别算法,在获取图像与模板在对应的各个像素点上的匹配情况后,根据点间距分别对匹配点和不匹配点进行聚类,将聚集在一起的点划分到同一个分组中,根据聚类结果对不同组内的点设置不同的权重值,再计算图像与各个模板的匹配度,将匹配度最高的模板代表的字符作为识别的结果。实验结果证明,此算法提高了对有背景噪声的字符图像进行识别的准确率。

The low recognition accuracy is reduced by the background noise of image in process of recognizing characters in images with standard template matching algorithm. Distributionweighted template matching recognition algorithm was introduced to achieve higher recognition accuracy. After obtaining the matching results between images and templates, all the matching points and mismatching points were clustered according to the distance between points. Then mismatching points and matching points would be divided into different groups. Points in different groups were set the different weights. The collective points would be appended with high weights. The dispersive points would be appended with low weights. Matching degree between images and templates was computed at last. The template which has the highest matching degree would be chosen as the recognition result. Theexperiments show that this algorithm has a higher recognition accuracy in recognizing the character images with noise background.