Full-Text Search:
Home|Journal Papers|About CNKI|User Service|FAQ|Contact Us|中文
《Optical Technique》 2018-05
Add to Favorite Get Latest Update

The text extraction algorithm based on multi-feature detection and support vector regression

YANG Jun;ZHAO Lin;WuHan Polytechnic;Guangxi Electrical Polytechnic Institute;  
It is difficult to extract scene text from complex background,and a text automatic detection and extraction scheme based on multi-feature detection and support vector regression was proposed.In order to distinguish the text from the non text edges effectively,three text features were extracted based on the image edge.And the three text features by multi-scale fusion,feature fusion using text detection candidate text boundaries helps detect text with different sizes of different types of degraded image to improve the robustness.For each candidate text boundary detected,the local threshold of each pixel is estimated according to the pixels in the neighborhood window,and the candidate characters are adaptively segmented by local threshold.The support vector regression(SVR)model is introduced to separate the text pixels from the image background,eliminate the non text boundaries and extract the real characters and words.Experiments show that the proposed method has better robustness and can be applied to text extraction in various complex scenes.Compared with other algorithms,this algorithm has better Precision-Recall curve and F measurement value.
【Fund】: 湖北省教育厅科学技术研究计划指导性项目(B201658);; 广西省自然科学基金(2014GXNSFCJ053172)
【CateGory Index】: TP391.41;TP391.1
Download(CAJ format) Download(PDF format)
CAJViewer7.0 supports all the CNKI file formats; AdobeReader only supports the PDF format.
©2006 Tsinghua Tongfang Knowledge Network Technology Co., Ltd.(Beijing)(TTKN) All rights reserved