计算机技术与发展Issue(5)：41-45,5.DOI:10.3969/j.issn.1673-629X.2015.05.011
自然场景下基于连通域检测的文字识别算法研究
Research on Character Recognition Algorithm Based on Connected Domain Detection in Natural Scene
摘要
Abstract
Due to the complexity of text background in natural scene,it brings great difficulties to the character recognition. In this paper, propose an algorithm combined edge detection and connected component analysis to identify the text in natural scene,which improves the accuracy and recall rate of text recognition. Firstly,the ColorRoberts operator is used to detect the edge of color image directly in this al-gorithm,thus avoiding the information loss in the process of conversion from a color image to a gray image. Secondly,the subsequent pro-cessing operation,such as removing the long straight lines,removing isolated noise points and morphological processing,is actualized on the image edge which has been detected. Finally,through the connected domain labeling and analyzing,the text regions are extracted. Simulation results are given,which shows that the algorithm is reasonable and effective.关键词
边缘检测/ColorRoberts算子/连通域标记/连通域分析Key words
edge detection/ColorRoberts operator/connected domain labeling/connected domain analysis分类
信息技术与安全科学引用本文复制引用
刘新瀚,钱侃,王宇飞,朱向霄,孙知信..自然场景下基于连通域检测的文字识别算法研究[J].计算机技术与发展,2015,(5):41-45,5.基金项目
国家自然科学基金资助项目(60973140,61170276,61373135) （60973140,61170276,61373135）
江苏省产学研项目(BY2013011) （BY2013011）
江苏省科技型企业创新基金项目(BC2013027) （BC2013027）
江苏省高校自然科学研究重大项目(12KJA520003) （12KJA520003）
