基于多帧图像的视频文字跟踪和分割算法.pdf

计算机研究与发展 ISSN 1000-1239!CN 11-1777!TP Journal of Com uter Research and develo ment 43（）： ， p p 9 1523 1529 2006 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!! 基于多帧图像的视频文字跟踪和分割算法 密聪杰 刘 洋 薛向阳 （复旦大学计算机科学与工程系 上海 200433 ） （ ） 032021187#fudan .edu .cn Video texts trackin and se mentation based on multi le f rames g g p ， ， Mi Congj ie Liu Yang and xue xiangyang （ ， ， ） De artment o Com uter science and en ineerin Fudan unioersit shan hai 200433 P f P g g y g Abstract Superimposed texts bring important semantic clues for video indexing and retrieval . Texts in videos of ten s an tens or even hundreds of frames and man researchers have ex loited the tem oral p y p p redundanc of video text to im rove the text detection accurac and the text re ion ualit . described in y p y g C y this a er is a novel a roach to track and se ment static su erim osed texts b utilizin multi le video p p pp g p p y g p frame information . For text detection ， multi le frames are used to verif the a earance of the text re ions p y pp g