摘要
为了在复杂背景下提取字符,采用了基于图像理论的降维算法和彩色游程算法,对复杂图像进行颜色聚类,通过阈值控制相近颜色的聚类,同时生成平均颜色的彩色游程信息,有效地合并了颜色种类,为字符区域的二值化奠定了良好的基础.实验结果表明,采用以上算法在复杂背景下提取字符,其识别率达到89%.
In order to detect characters in a complex background, two algorithms respectively of dimensionality reduction based on graph-theory and of color run-length are adopted to process image. The above algorithms cluster similar colors to generate average color run-length for image binarization by a threshold. The experiment result shows that the detected characters of eighty-nine percentages in a complex background are successfully recognized by OCR.
出处
《东北师大学报(自然科学版)》
CAS
CSCD
北大核心
2008年第1期40-44,共5页
Journal of Northeast Normal University(Natural Science Edition)
基金
湖北省自然科学基金资助项目(2004ABA029)
湖北省教育厅高等学校教学研究项目(20050232)
关键词
复杂背景
字符提取
图像理论聚类
彩色游程
complex background
character detection
graph-theoretical cluster
color run-length