摘要： 在自动标注系统中，底层特征转换成高层标注的准确度较低。为此，将自动标注系统中的底层视觉特征和社会标注系统中的高级语义相结合，提出一种新的图像语义标注算法——FAC算法。从自动标注系统和flickr网站用户中得到候选标注，利用图像标注推荐策略获取推荐标注，根据WordNet语义词典中的语义关系，精简出最终的标注集合。实验结果表明，与传统的自动标注算法相比，FAC算法的准确度较高。

关键词: 图像检索, 社会标注, 图像分割, 图像标注, 模糊评判, 范数

Abstract: In the system of image automatic annotation, the accuracy of low-level feature and senior semantic annotation is low. Aiming at this problem, combined with low visual features in automatic notation system and senior semantic feature in socialized annotation system, this paper proposes FAC algorithm based on image automatic and socialized annotation. It obtains several candidate annotations from automatic notation system, and users in the website flickr. It uses the image annotation recommending strategy to get recommending annotations. The final aggregate of label results are made full use of the semantic relations from the WordNet semantic dictionary. Experimental result shows that, compared with traditional automatic notation algorithm, the accuracy of FAC algorithm is higher than traditional semantic retrieval.

Key words: image retrieval, socialized annotation, image segmentation, image annotation, fuzzy evaluation, norm

中图分类号: 

• TP391.4