Compared with the method which extracts features directly in image domain, the proposed method shows better performance in detecting text in images with complex background.

• 与直接在图像空间域中提取纹理特征的方法相比,该方法对复杂背景下的文本检测更为有效,能有效地解决场景纹理特征与文本区域相近似造成的文本检测错误。