Perceptual image hashing via dual-cross pattern encoding and salient structure detection

In this paper, we propose a robust image hashing scheme based on perceptual texture and structure features. Through pre-processing, the input image with arbitrary size is regularized as a secondary image to alleviate the influence of noises. Then, the encoding of dual-cross pattern (DCP) is conducted on the secondary image to produce two coded maps representing textural information in horizontal-vertical and diagonal directions, respectively, and the DCP-based textural features can be extracted with the assist of histogram composition. On the other hand, salient structural features can be extracted from the frequency coefficients and position information of selective-sampled blocks containing the richest corner points. The final hash can be obtained after dimension reduction for the two types of extracted features. Experimental results demonstrate that our scheme has better performances with respect to robustness, anti-collision, and efficiency than some of state-of-the-art schemes.