目的 为了解决当前彩色图像水印技术的水印容量和不可感知性不佳等问题，设计基于YCoCg-R颜色空间和离散余弦变换的自适应图像水印算法。方法 首先，将载体图像从RGB空间转换到YCoCg-R空间，计算Y, Co, Cg这3个颜色分量；随后，引入离散余弦变换DCT对Y分量进行处理，获取相应的DCT系数；引入模糊聚类分析，将DCT低频系数视为局部特征，利用SVM机制对DCT低频系数进行训练分类，确定出适合嵌入水印的DCT系数；设计水印嵌入机制，将水印数据隐藏到合适的DCT低频系数中；借助可逆DCT方法，将Y分量从频域变换到空域；联合初始的Co与Cg分量，将其变换到RGB空间，输出水印图像；最后，通过相应的水印提取方法，复原初始水印信息。结果 测试数据表明，与当前水印方法相比，所提技术具有更高的不可感知性与水印容量，其水印图像的PSNR值为43.67 dB，且其允许的水印容量可超过1 bits，在多种攻击下，复原水印的质量更高，相关系数NC均超过了0.92。结论 所提技术能够较好地将水印信息隐藏在载体图像中，具有较高的鲁棒性和安全性。

The work aims to design the adaptive image watermarking algorithm based on YCoCg-R color space and discrete cosine transform, in order to solve the defects like poor watermark capacity and imperceptibility in current watermarking technology of color images. Firstly, three color components (Y, Co and Cg) were calculated by transforming the carrier image from the RGB space into YCoCg-R space. Then, the corresponding DCT coefficients were obtained by introducing the discrete cosine transform (DCT) to process the Y component. The fuzzy clustering analysis was introduced to regard the DCT low frequency coefficients as local features. The DCT coefficients suitable for embedding watermarks were determined by training DCT low frequency coefficients with SVM mechanism. The watermark embedding mechanism was designed to hide the watermark data into the appropriate DCT low fre-quency coefficients. The Y component was transformed from the frequency domain to the airspace by means of reversible DCT method. The watermark image was outputted by combining the initial Co and Cg components and transforming them into RGB space. Finally, the initial watermark information was restored by the corresponding watermark extraction method. The test data showed that the proposed technology had higher imperceptibility and watermark capacity compared with the current watermarking method. The watermark image had a PSNR value of 43.67 dB, and the watermark capacity of its allowable watermark capacity could be more than 1 bits. Restoration watermark had higher quality and the correlation coefficient NC exceeded 0.92 under a variety of attacks. The proposed technology can better hide the watermark information in the carrier image, and it has higher robustness and security.