Two Robust Perceptual Image Hashing Schemes Based on Discrete Wavelet Transform

Abstract

Robustness, discrimination and security are essential characteristics of perceptual image hashing in cloud computing. In this paper, we propose two image hashing schemes based on block discrete wavelet transform (BDWT) to achieve these three characteristics. The basic scheme combines block discrete wavelet transform, singular value decomposition and statistics feature to extract the features of each overlapping image block. And chaotic cryptography is applied to encrypt these features. Furthermore, image correction is added to the basic scheme to realize a more comprehensive image hashing IC-BDWT, which can resist rotation attack. Meanwhile, we conduct experiments on 400 images from four public databases to analyze the proposed schemes. The receiver operating characteristic (ROC) curve of the experimental results demonstrates that the proposed hashing algorithms exhibit superior robustness and discrimination compared to the latest hashing algorithms under common content-preserving attacks. And the IC-BDWT hashing is far better than other algorithms in resisting rotation attack. Experimental results show that our schemes are secure and sensitive to secret keys.