非压缩域数字视频中文字的检测与提取  

Text Detection and Extraction in Uncompressed Digital Videos

作　　者：高平利[1] 任金昌[1] 赵荣椿[1] 

机构地区：[1]西北工业大学计算机学院,西安710072

出　　处：《中国体视学与图像分析》2004年第4期244-248,共5页Chinese Journal of Stereology and Image Analysis

摘　　要：作为一种高级语义特征 ,视频中的文字信息对视频内容的理解、索引和检索具有重要意义。本文针对非压缩域中视频文字的检测与提取技术做了详尽的分析和讨论 ,涉及的内容包括 :文字事件的检测、帧图像中文字区域的检测与分割以及视频文字识别 (VideoOCR)等。对于不同的典型算法 ,分析了其理论基础和应用特点 ,并且通过对比指出其不足和应用的局限。最后 ,本文还展望了该技术未来发展的方向。As a high-level semantic feature, text information of videos is very important for understanding, indexing and retrieval of video contents. In this paper, comprehensive analysis and discussions are given on detection and extraction of texts from videos in uncompressed domain, in which several topics are involved: text event detection, text region detection and segmentation in video frames as well as video OCR, etc. As for different typical algorithms, their theoretical basis and applications are discussed, along with shortcomings and limitations in applications are pointed out through comparisons. Finally, some relevant prospects are also proposed.

关 键 词：非压缩域 文字事件检测 文字分割与识别 视频索引与检索 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]