Abstract: In order to solve the problem as low robustness and tamper detection ability induced by ignoring the color feature and angle feature of the image in current image Hashing technology, which mainly uses the image gray information to generate the Hash sequence, the robust color image Hashing algorithm based on mixed features and color vector angles is designed by extracting the color information and angle features of color images. Firstly, the linear interpolation and Gauss low-pass filter are introduced to preprocess the initial image for enhancing the robustness of content preserving operations such as scaling and noise. Then, the filtered image is divided into a series of non overlapping sub-blocks, and the secondary image is obtained by using the singular value decomposition method to process each sub-block. By designing the robust feature extraction method, the color vector angle of two images is calculated to obtain the perceptual features of the image for determining the salient feature points. A mixed feature is formed by combining the extracted two features, and a new feature sequence is outputted based on quantization method. The encryption mechanism is designed based on the Logistic map to confuse the feature sequence, and it can obtain the robust Hash sequence. Finally, the Hamming distance is used to detect the authenticity of image content. The experimental results show that the proposed algorithm has higher robustness and tamper detection accuracy compared with the current Hash technique.

Key words: image Hash, mixed feature, angle of color vector, secondary image, salient edge points, Logistic mapping, encryption mechanism

摘要： 为了解决当前图像哈希技术主要利用其灰度信息来生成哈希序列，忽略了图像的色彩特征，使其对感知鲁棒性与篡改检测能力不佳的问题，通过提取彩色图像的颜色信息与角度特征，设计了混合特征与颜色矢量角度的鲁棒彩色图像哈希算法。首先，引入线性插值与高斯低通滤波器，对初始图像完成预处理，增强哈希序列的对缩放、噪声等内容保留操作的鲁棒性；随后，将滤波图像分割为一系列的非重叠子块，利用奇异值分解方法处理每个子块，获取二次图像，改善其对旋转的鲁棒性；再计算二次图像的颜色矢量角度，通过设计两个鲁棒特征提取方法，确定显著性边缘像素点，获取图像的感知特征；通过将提取的两个特征进行组合，形成混合特征，利用量化方法，输出新的特征序列；并基于Logistic映射，设计加密机制，对混合特征进行混淆，获取鲁棒哈希序列。最后，利用Hamming距离，对图像内容的真伪进行检测。实验结果表明：与当前哈希技术相比，所提算法具备更高的鲁棒性与篡改检测精度。

关键词: 图像哈希, 混合特征, 颜色矢量角度, 二次图像, 显著性边缘像素点, Logistic映射, 加密机制