Automatic image annotation using community detection in neighbor images

摘要: Abstract Automatic image annotation is useful in automatic management and, understanding their contents. refers to the process of assigning tags/labels images so that they clearly indicate content which meant be conveyed. In this paper, a new method presented. first phase proposed method, neighbor are similar query image, retrieved using low-level features. second phase, network (a relative graph) tags created finally, densest community among all communities selected for as final tags. order show efficiency some metrics such precision, recall, and f-score used three standard datasets namely Corel5k, IAPR TC12 Mir Flickr. The results more efficient than state-of-the-art approaches.