Study on the de-watermark algorithm based on grayscale text  

作　　者：Huang Guoquan Chen Zhipeng Sun Xiaocui 黄国权;Chen Zhipeng;Sun Xiaocui(Department of Computer Science,Guangdong Pharmaceutical University,Guangzhou 510006,P.R.China;Medicinal Information&Real World Engineering Technology Center,Guangdong Pharmaceutical University,Guangzhou 510006,P.R.China)

机构地区：[1]Department of Computer Science,Guangdong Pharmaceutical University,Guangzhou 510006,P.R.China [2]Medicinal Information&Real World Engineering Technology Center,Guangdong Pharmaceutical University,Guangzhou 510006,P.R.China

出　　处：《High Technology Letters》2021年第1期95-102,共8页高技术通讯（英文版）

基　　金：Supported by the National Entrepreneurship Training Fund Project(No.201810573007X);China Scholarship Council Project(No.201708440547);GDPU Higher Education Innovation Strong School Project(No.2017GXJK079,51359055)。

摘　　要：When using the current popular text recognition algorithms such as optical character recognition(OCR)algorithm for text images,the presence of watermarks in text images interferes with algorithm recognition to the extent of fuzzy font,which is not conducive to the improvement of the recognition rate.In order to pursue fast and high recognition rate,watermark removal has become a critical problem to be solved.This work studies the watermarking algorithm based on morphological algorithm set and classic image algorithm in computer images.It can not only remove the watermark in a short time,but also keep the form and clarity of the text in the image.The algorithm also meets the requirements that the higher the clarity of image and text,the better the processing effect.It can process the Chinese characters with complex structure,complicated radicals or other characters well.In addition,the algorithm can basically process ordinary size images in 1 s,the efficiency is relatively high.

关 键 词：de-watermark text recognition character recognition optical character recognition(OCR)application 

分 类 号：TP309.7[自动化与计算机技术—计算机系统结构]