Multiple Feature Fusion Image Retrieval Method Via Graph Method

With the development of image retrieval technology,retrieval result re-ranking and multiple features fusion technology have been widely concerned.Most previous studies mainly consider similarity measure method and iterative algorithm to improve the retrieval accuracy.However,image manifold in real world is always complex,which may not be suitable for image retrieval.Images that are closer to the query image may not have higher correlation with the query image.In order to solve the above problems,this paper proposed a novel approach for image unsupervised re-rank of single feature on graph and illustrated rationality of multiple feature fusion ranking.In this paper,we first analyze graph structure and multiple feature fusion re-ranking from manifold aspect.Then,Three Tiers Neighborhood Graph(TTNG)is constructed to re-rank the original ranking list by single feature and to enhance precision of single feature.Furthermore,we propose Multi-graph Fusion Ranking(MFR)for multiple feature ranking which considers the correlation of all images in multiple neighborhood graphs.Finally,Evaluations are conducted on UKBench,Corel-1k,Corel-10 k,Cifar-10,Holiday and Oxford-5k benchmark datasets.The experimental results show that our TTNG and MFR outperform other state-of-the-art methods.For example,we achieve competitive results NS score 3.94,precision 94.30%,precision 70.56%,precision 44.21%,MAP 92.52% and MAP 80.64% on UKBench,Corel-1k,Corel-10 k,Cifar-10,Holiday and Oxford-5k benchmark datasets respectively.