Abstract: In order to improve the performance of image hashing algorithm with resistance to geometric distortions, a hashing method based on image normalization and Slant transform is proposed in this paper. The affine invariant image is obtained by using geometric moments invariants with image normalization method. The image is then blocked and Slant transform is applied, and middle-frequency coefficients are obtained. In addition, it further generates the image hashing sequence by using the user-key. Test results indicate that the proposed method is robust to common image procession and geometric distortions, and sensitive to the user-key, so the method proposed in this paper is effective.

Key words: image normalization, Slant transform, image hashing, normalized Hamming distance

摘要： 为了提高图像哈希算法抵抗几何攻击的性能，提出了一种结合图像归一化和Slant变换的图像哈希算法。使用几何不变矩将原始图像进行归一化操作得到仿射不变图像，仿射不变图像分块并进行Slant变换，根据用户密钥选择Slant变换中频系数极性并生成最终的图像哈希序列。实验结果表明该算法对常见图像操作处理和常见几何攻击鲁棒，对原始图像内容和用户密钥敏感，是一种有效的图像哈希算法。

关键词: 图像归一化, Slant变换, 图像哈希, 归一化汉明距离