Latent Structure Preserving Hashing

Ziyun Cai, Li Liu, Mengyang Yu and Ling Shao

Abstract

Aiming at efficient similarity search, hash functions are designed to embed high-dimensional feature descriptors to low-dimensional binary codes such that similar descriptors will lead to the binary codes with a short distance in the Hamming space. It is critical to effectively maintain the intrinsic structure and preserve the original information of data in a hashing algorithm. In this paper, we propose a novel hashing algorithm called Latent Structure Preserving Hashing (LSPH), with the target of finding a well-structured low-dimensional data representation from the original high-dimensional data through a novel objective function based on Nonnegative Matrix Factorization (NMF). Via exploiting the probabilistic distribution of data, LSPH can automatically learn the latent information and successfully preserve the structure of high-dimensional data. After finding the low-dimensional representations, the hash functions can be acquired through multi-variable logistic regression. Experimental results on two large-scale datasets, i.e., SIFT 1M and GIST 1M, show that LSPH can significantly outperform the state-of-the-art hashing techniques.

Session

Search and Detection

Files

Extended Abstract (PDF, 237K)
Paper (PDF, 2M)

DOI

10.5244/C.29.172
https://dx.doi.org/10.5244/C.29.172

Citation

Bibtex

@inproceedings{BMVC2015_172,
	title={Latent Structure Preserving Hashing},
	author={Ziyun Cai and Li Liu and Mengyang Yu and Ling Shao},
	year={2015},
	month={September},
	pages={172.1-172.11},
	articleno={172},
	numpages={11},
	booktitle={Proceedings of the British Machine Vision Conference (BMVC)},
	publisher={BMVA Press},
	editor={Xianghua Xie, Mark W. Jones, and Gary K. L. Tam},
	doi={10.5244/C.29.172},
	isbn={1-901725-53-7},
	url={https://dx.doi.org/10.5244/C.29.172}
}