摘要
为提高文本图像在光照不均条件下的OCR识别率,提出一种基于Retinex的光照不均校正方法。采用分块Otsu二值化减弱光照不均的影响程度,分析其得到的二值图像,判断光照情况并获得光照不均图像块,利用改进的Retinex方法对光照不均图像块进行重新处理,采用Sauvola二值化得到清晰的校正图像。实验结果表明,该算法能很好地消除光照不均的影响,取得较好的二值图像,在各式样本以及不同光照情况下都有较高的OCR识别率。
To improve the OCR recognition rate of the document image under the condition of uneven illumination,a Retinexbased illumination correction method was proposed.The influence of uneven illumination was weakened using the block Otsu binarization,and the obtained binarization images were analyzed to determine uneven illumination image blocks,the improved Retinex method was used to re-process uneven illumination image blocks,while clear correction image was obtained using the Sauvola binarization.Experimental results show that the proposed algorithm can eliminate the influence of uneven illumination,and obtain good binary image,and it has higher OCR recognition rate in various samples under different illumination conditions.
作者
曾凡锋
刘树鹏
ZENG Fan-feng LIU Shu - peng(College of Computer, North China University of Technology, Beijing 100144, China)
出处
《计算机工程与设计》
北大核心
2017年第11期3072-3079,共8页
Computer Engineering and Design
基金
国家自然科学基金项目(61371142)
北方工业大学校内专项基金项目(XN060)
