Tag quality improvement for social images

摘要: Online social media sharing web sites like Flickr allow users to manually annotate images with tags, which can facilitate image search and organization. However, the tags provided by are often imprecise incomplete, severely limits application of browse. In this paper, we propose a scheme improve poorly annotated associated images. Two properties exploited integrated in an unified optimization framework: (1) consistency between visual semantic similarities, where similarity is estimated using tags; (2) compatibility before after improvement, since initial user carry valuable information. An iterative bound method derived solve problem. Experimental results on dataset show that proposed significantly quality tags.