Semi-supervised manifold-ranking-based image retrieval with low-rank Nyström approximation. (Chinese. English summary) Zbl 1265.68138

Fu, Xianghua; Li, Jianqiang; Wang, Zhiqiang; Du, Wenfeng

Acta Autom. Sin. 37, No. 7, 787-793 (2011).

Summary: In the real image retrieval, there is an abundance of unlabeled images whereas there only exist few labeled images. To address this issue, based on our previous work on semi-supervised manifold image retrieval, this paper proposes a novel learning method, called semi-supervised manifold ranking based image retrieval (\(\mathrm{S}^2\)MRBIR). The images are assumed to be always embedded in low-dimensional sub-manifolds. In particular, \(\mathrm{S}^2\)MRBIR adopts the manifold regularization framework to rank the retrieved images while regarding the relevant feedback process of image retrieval as an online learning process and treating the image retrieval as a classification problem. The manifold regularization framework is capable of taking into account both labeled and unlabeled data, the classification performance, the geometric structures of the data distribution, and the complexity of the classifier. Moreover, an accelerating algorithm based on low-rank Nyström approximation is proposed to improve the computing procedure of \(\mathrm{S}^2\)MRBIR. Experimental results on the Corel image database demonstrate the effectiveness of \(\mathrm{S}^2\)MRBIR.

MSC:

68T05	Learning and adaptive systems in artificial intelligence
68U10	Computing methodologies for image processing
68P20	Information storage and retrieval of data

Keywords:

image retrieval; manifold learning; Nyström approximation; semi-supervised learning

Software:

Corel