提出一种基于支持向量机( SVM)和遗传算法(GA)的离散余弦变换(DCT)域盲数字图像水印方法. 该方法能自适应于图像的局部特征. 依据图像块的局部特性,利用SVM对图像块分类,自适应地确定水印嵌入强度, GA用来优化水印嵌入位置. 实验结果表明该方法有较好的不可见性和较强对抗攻击的鲁棒性.-Based on support vector machine (SVM) and genetic algorithm (GA), discrete cosine transform (DCT) domain blind digital image watermarking method. The method adaptive to local image features. Based on the local characteristics of image block using the SVM the image block classification, adaptive to determine the embedding strength, GA to optimize the watermark embedding location. The experimental results show that the method has good invisibility and robustness of the strong against the attacks.