中文摘要：

为解决当前图像哈希算法难以兼顾较高的抗旋转鲁棒性与正确识别精度的问题,提出非负矩阵分解耦合环形区域分割的图像哈希认证算法。引入双线性插值算子,对输入图像进行预处理,增强哈希对图像缩放的鲁棒性,使其对不同尺寸图像具有相同长度的哈希序列;设计环形区域分割机制,将预处理图像分割为环形像素块,形成二次图像,获取旋转不变矩阵;利用非负矩阵分解,联合乘性更新规则,获取二次图像的系数矩阵,构建紧凑的图像哈希;依据哈希序列元素,设计相关系数模型,计算初始图像与用户接收图像的哈希值的相似度,通过优化确定决策阈值,完成图像内容的真伪认证。实验结果表明,与当前图像哈希算法相比,所提算法具有更强的感知鲁棒性与安全性,呈现出更好的ROC曲线特性,能够精确识别出旋转、噪声、缩放以及伽马校正等数字操作。

英文摘要：

To solve the difficulties of coping with both the high robustness for rotation tampering and the correct recognition ac-curacy in the current image hashing algorithm, the color image hash authentication algorithm based on non-negative matrix de-compose and circle region segmentation was proposed. The bilinear interpolation operator was introduced to pre-process the in-put image to enhance the scaling robustness for image hashes of different size images having the same length. The annular region segmentation mechanism was designed to divide the preprocessing image into annular pixel blocks for obtaining secondary image and rotation invariant matrix. The coefficient matrix of the secondary image was constructed based on non-negative matrix de-compose for obtaining the compact image hash. The correlation coefficient model was designed according to hash sequence ele-ments to calculate the hash value similarity between the original image and the user receiving image for completing the authentici-ty of the image content authentication. Experimental results show that the proposed algorithm has stronger perception robust-ness and safety with better ROC curve for accurately identifying the rotation, noise,and scaling and gamma correction tampe-ring attacks.