基于互联网搜索与反馈验证的图像自动标注  

Automatic Image Annotation Based on Internet Search and Verification of Feedback

作　　者：俞建松[1,2] 曹冬林[1,2] 李绍滋[1,2] 林达真[1,2] 

机构地区：[1]厦门大学智能科学与技术系,福建厦门361005 [2]福建省仿脑智能系统重点实验室,福建厦门361005

出　　处：《计算机工程》2012年第24期211-215,共5页Computer Engineering

基　　金：国家自然科学基金资助项目(60873179);福建省自然科学基金资助项目(2011J01367);高等学校博士学科点专项科研基金资助项目(20090121110032);深圳市科技计划基础研究基金资助项目(JC200903180630A);深圳市科技研发深港创新圈计划基金资助项目(ZYB200907110169A)

摘　　要：基于网页的图像自动标注存在图像与文本的关联假设问题,而基于内容的方法则存在数据规模小的问题。为此,将基于内容和基于网页的自动标注方法相结合,提出基于互联网搜索和反馈验证的图像自动标注方法。利用网页关联文本从互联网搜索结果中提取候选标注词,根据候选标注词在搜索得到图片的内容特征进行反馈验证。实验结果表明,该方法具有大规模标注能力,准确率比基于网页的图像自动标注方法提高了7.92%。Up to nowadays,the research of Web-based automatic image annotation is mainly about the problem of the relevance assumption of the image and text,and the main problem of the content-based automatic image annotation is the limit of the database.Aiming at this problem,this paper proposes the Internet-search-based automatic image annotation with the verification of feedback,combining the content-based and the Web-based automatic image annotation.It extracts candidate labels from the search results using Web-based texts associated with the image,and then verifies the final results by using the Internet search results of the candidate labels with the content-based image features.Experimental results show that this method can annotate the large-scale database,with high accuracy,and achieves 7.92% improvement on the basis of web-based automatic image annotation.

关 键 词：图像自动标注 互联网搜索 网页 关联文本 图像内容特征 反馈验证 

分 类 号：TN911.73[电子电信—通信与信息系统]