Robust Image Hashing Using Histogram Reconstruction for Improving Content Preservation Resistance and Discrimination

This paper proposes a new image hashing method, which uses histogram reconstruction to solve the problem of the histogram not being sensitive to the change of pixel position, while ensuring the robustness of the hashing algorithm against common content preservation attacks (such as blurring, noise addition and rotation). The proposed algorithm can resist arbitrary angles of rotation, possibly because the reconstructed histogram leverages the rotational symmetry and its own invariance to rotation operations. We measure the similarity between different images by calculating the Hamming distance of the hash vectors of different images. Our experiments show that the proposed method performs well in robustness and discrimination compared with other established algorithms. In addition, we conduct a receiver operating characteristic curve analysis to further verify the superior overall performance of our image hash method.

Automated summary

This paper proposes a new image hashing method based on histogram reconstruction to enhance the sensitivity of histograms to changes in pixel position and ensure robustness against common content preservation attacks. The algorithm leverages rotational symmetry and invariance to rotation operations, enabling resistance to arbitrary angles of rotation. Experimental results demonstrate the method's strong performance in robustness and discrimination compared to established algorithms. Additionally, a receiver operating characteristic curve analysis further confirms the superior overall performance of the proposed image hash method.