Locality Preserving Hashing Method for Image Retrieval

Qiang Li,Yahong Han,Jianwu Dang
DOI: https://doi.org/10.1145/2632856.2632898
2014-01-01
Abstract:Binary hashing is widely used in efficient similarity computing for image/video search. In this work, we propose a Locality Preserving Hashing (LPH) method to improve the image retrieval performance, we first project high-dimensional feature space into low dimension using locality preserving method, in which the locality properties can be well maintained, then binaries the low-dimensional feature into binary index code using a heuristic projection strategy; in retrieval stage, we compute the Hamming distance of the cell indices by implementing XOR and bit-count operations, obtain the similarity result and finally get the target images. We evaluate the algorithm on four public image datasets and compare it with some popular methods, and the experimental results show the validity of our proposed method.
What problem does this paper attempt to address?