计算机应用研究2011，Vol.28Issue(10)：3925-3928,3933,5.DOI:10.3969/j.issn.1001-3695.2011.10.089
基于显著区域的图像自动标注
Automatic image annotation based on salient regions
摘要
Abstract
In order to improve the performance of automatic image annotation, this paper presented a novel algorithm based on image saliency analysis. Firstly,computed image saliency and obtained the salient objects. Then, the SIFT features were extracted for every image, the visual words were derived through K-Means cluster algorithm, and thus the visual Bag-of-Words model weighted by the saliency information was built to characterize images. Finally, SVM was trained to classify and annotate images automatically. Compared with unweighted algorithm, experimental results on 1255 images from Corel database show that this algorithm improves the annotation accuracy. It demonstrates the proposed approach is promising.关键词
图像自动标注/显著区域/SIFT特征/K-均值聚类/视觉词袋/支持向量机Key words
automatic image annotation/ saliency/ SIFT feature/ K-Means cluster/ bag-of-words/ SVM分类
信息技术与安全科学引用本文复制引用
尹文杰,韩军伟,郭雷,贺胜,许明..基于显著区域的图像自动标注[J].计算机应用研究,2011,28(10):3925-3928,3933,5.基金项目
西北工业大学基础研究基金资助项目(JC201041) （JC201041）
西北工业大学引进高层次人才科研启动费资助项目 （）
