目的 为了改善哈希序列对篡改内容的正确识别率和鲁棒性，提出基于颜色矢量角度直方图耦合离散余弦变换压缩的鲁棒哈希算法。方法 结合插值运算和Gaussian滤波器，对图像实施预处理，使其对于任意的认证目标均可产生一个固定长度的哈希序列；引入极坐标变换LPT（Log-polar transform），对规范尺寸图像完成坐标变换，输出二次图像；借助SVD（Singular Value Decompostion）方法来处理二次图像，输出全局鲁棒特征；提取规范尺寸图像中所有像素分量的颜色矢量角度，并计算其对应的直方图，形成色彩感知特征；将全局与色彩感知特征组合，将其视为1D哈希序列；基于离散余弦变换DCT（Discrete Cosine Transform），对1D哈希序列实施压缩，输出一组交流系数；引入Logistic映射，根据其输出的混沌序列来设计差异加密方法，对交流系数实施扩散，从而形成目标哈希序列。基于l2范数距离和优化的识别阈值，对待检测图像内容的真伪实施判别。结果 试验结果显示，相对于已有的哈希机制而言，所提方案具备更强的鲁棒性，对各类几何攻击均有更高的检测准确度。结论 所提哈希方法具有较高的鲁棒性与识别正确率，对商标检索、信息防伪等领域具有较好的参考价值。

to the work aims to improve the correct recognition rate and robustness of hash sequences for tampered content by proposing a robust hash algorithm based on color vector angle histogram and coupled discrete cosine transform compression. Combining interpolation operation with Gauss filter, the image was preprocessed to produce a fixed-length hash sequence for any authentication target. The polar coordinate transformation LPT (Log-polar transform) was introduced to transform the coordinates the image in standard size to output the secondary images. SVD method (Singular Value Decomposition) was used to process the secondary image to output global robust features. Subsequently, the color vector angles of all the pixel components in the normalized size image were extracted, and the corresponding histograms were calculated to form color perception features. Combined global and color perception features were regarded as 1D hash sequences. Based on discrete cosine transform (DCT Discrete Cosine Transform), the 1D hash sequence was compressed to get a set of AC coefficients. By introducing the logistic map, the differential encryption method was designed by the chaotic sequence to diffuse the AC coefficients to form the target hash sequence. Based on the optimal decision threshold and l2 norm distance, the authenticity of detected image content was dis-criminated. From the test results, the proposed scheme had stronger robustness and higher detection accuracy for various geometric attacks than the current hash mechanism. This hash method has high robustness and recognition accuracy, which has good reference value for trademark retrieval, information anti-counterfeiting and other fields.