基于区域的光照不均文本图像校正方法  

Uneven illumination of text image correction method based on area

作　　者：曾凡锋[1] 付亚南[1] 李良旭[1] 

机构地区：[1]北方工业大学信息工程学院,北京100144

出　　处：《计算机工程与设计》2014年第12期4233-4237,4242,共6页Computer Engineering and Design

基　　金：国家自然科学基金项目(61371142);十一五国家科技支撑平台重点基金项目(2009BA171B02);北京市自然科学基金项目(4132026);北京市属高等学校人才强教计划基金项目(PHR201007121)

摘　　要：针对光照不均文本图像识别率低下的问题,提出一种基于区域分类的自适应校正算法。通过八邻域三维增强,拉大背景点与目标点差异;通过对该类文本图像的特征分析,利用投影算法实现图像的区域分类;采用全局校正处理区域分类后的页白区域和均匀区域,结合评定图像质量的特征参数,对阴影区进行局部校正处理,实现背景与目标的彻底分离;分析噪音分布特点,采用邻域去噪法纠正误判像素点。实验结果表明,该算法能有效分离光照不均文本图像的目标点和背景,处理后的效果图可用于阅读以及提高OCR系统识别率。To deal with the low recognition rate of the uneven illumination of text image,an adaptive correction algorithm based on regions was put forward.Firstly,the difference between the background points and destination points was pulled through the eight fields of three-dimensional enhancement.Then,through the analysis of characteristics of the text image,the image’s region was classified by using the projection technique.The global correction was used to deal with the non-text area and the uniform illumination area,and combining with characteristics parameters for evaluating image quality assessment,the shaded area was corrected locally to achieve complete separation of backgrounds and objectives.Finally,the characteristic of image noise distribution was analyzed,and the neighborhood denoising method was adopted to correct mistakes pixels.The experimental results show that this algorithm can effectively separate the target pixel and the background pixel of uneven illumination of text image,and the visual rendering after processing can be used for reading and improving the recognition rate of OCR system.

关 键 词：光照不均 自适应 图像特征 投影算法 区域分类 

分 类 号：TP301[自动化与计算机技术—计算机系统结构] TP39[自动化与计算机技术—计算机科学与技术]