场景图像中文本区域字符提取方法研究  

Research on Text Region Character Extraction in Scene Image

作　　者：李海浩[1] 顾滨兵[1] 刘艳平[1] LI Haihao;GU Binbing;LIU Yanping(No.91404 Troops of PLA,Qinhuangdao 066000)

机构地区：[1]91404部队

出　　处：《计算机与数字工程》2018年第12期2580-2584,共5页Computer & Digital Engineering

摘　　要：随着计算机技术的发展,光学识别技术的发展已较为成熟,但自然场景图像包含较复杂背景,字符的提取和识别技术难度相对较大。在自然场景中文本区域提取技术的基础上,提出了一种字符提取模型,详细介绍了基于局部重叠阈值的图像二值化算法,通过字符的水平、垂直投影技术,实现了单个字符边界的划分,并根据字符的某些特征对非字符进行了过滤,最终实现了字符的提取。With the development of computer technology,the optical recognition technology has been developed very mature,however,the natural scene images contain more complex background,the extraction and recognition of characters are relatively dif?ficult.A character extraction model is proposed on the basis of text area extraction in natural scene,the two value algorithm of image based on local overlap threshold is introduced in detail,the division of the boundary of a single character is realized by the horizon?tal and vertical projection techniques of characters,the non-character is filtered according to some characteristics of the character,finally,the character extraction is realized.

关 键 词：场景图像 字符提取 方差阈值 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]