图像中的文本区域识别技术研究

Research on Text Region Recognition in Image

通过积分图像法快速提取图像的haar特征,利用Adaboost分类器对提取的特征进行迭代并选择,得到最优的弱分类器,然后对多个最优弱分类器进行组合,构成强分类器,解决文本区域识别的问题。实验结果表明,该方案提高了文本区域检测的查准率和查全率,同时降低了算法的复杂度。

This paper gets haar feature of image through the integral image method rapidly and iterates and selects the feature by using Adaboost classifier. Then, the best weak classifier will be got, which forms a strong classifier with the best weak classifier combination to solve the problem of identifying text regions. The experimental results show that the precision ratio and the recall ratio of text detection are improved, while the complexity of the algorithm is reduced.