Author(s)

Ling Gan, Tianzhen Zhang

Corresponding Author

Tianzhen Zhang

ABSTRACT

All Deep quantization learning based hashing methods have been proven to be effective in the field of image retrieval recently. How to improve the description ability of hash codes is still a challenging problem. In this paper, we propose a new deep quantization network architecture for supervised hashing called Supervised Deep Quantized Hashing (SDQH)for image retrieval. The main contribution is to reduce the redundancy between the hash code and the network parameters, Moreover, a block coding module is proposed and the similarity is learned through a specific loss function and joint quantization method to represent the image features as compact binary codes. Experimental results on three benchmark data sets show that the image retrieval performance of the hash code obtained by the proposed method outperforms the current main methods.

KEYWORDS

Image retrieval, Hashing, Deep quantization learning