Compressed Hashing for Neighborhood Structure Preserving

Abstract

Hashing methods play an important role in large-scale image retrieval, and have been widely applied due to fast approximate nearest neighbor search and efficient data storage. However, most existing hashing methods are not taken the low-dimensional manifold into account in nearest neighbors search. In this paper, we propose an effective hashing method to preserve the intrinsic structure of high-dimensional data points in the low-dimensional manifold space. In particular, we introduce a compressed algorithm to learn a smaller synthetic data set represent the database in the original space so that the approximated nearest neighbors can be quickly discovered. To this end, we exploit the manifold learning generate appropriate binary code. Experimental results on benchmark data sets show that the proposed approach is effective in comparison with state-of-the-art methods.