Image annotation based on manifold structure by fusion of multiple dissimilarity spaces

摘要

Automatic image annotation has been an active research topic in recent years due to its potential impact on both image understanding and web image retrieval. In many situations the similarity between two image feature vectors could not be found correctly by the Euclidean distance between feature vectors. The purpose of this study is to reduce the semantic gap in automatic image annotation by learning the intrinsic structure collectively revealed by known labeled and unlabeled images.We learn a semantical dissimilarity graph based on fusion of dissimilarities in multiple spaces. The experiments showed that the geodesic distances between the samples on the learned manifold structure are closer to their semantic distance. So, the continuity between the instances of a semantic at the semantic space is kept in manifold space. The proposed method has been compared to the other well-known approaches by Corel data set. The results confirmed the effectiveness and validity of the proposed method.