中文摘要：

为了解决多模生物认证中生物模板安全传输问题，提出把人脸图像隐藏嵌入到指纹图像中，用于多模生物特征认证，以提高生物体征识别的安全性和准确性。利用指纹图像归一化、中心点检测，获取嵌入区域的几何失真不变域，提出在几何失真不变域基于奇异值分解（SVD）的多重嵌入算法，嵌入相应的人脸图像。检测者通过相关性优化算法，盲提取人脸图像，再结合载体指纹图像进行多模认证。实验结果表明，该算法能够抵抗旋转、缩放、平移等几何失真，也能抵抗压缩、滤波、噪声等攻击，提高了生物模板的传输安全性，指纹与人脸双模生物认证相比于单模认证具有更高的正确识别率。

英文摘要：

In order to solve the bio-template secure transmission in multimode biometric authentication, a novel al- gorithm was presented that hiding face watermarks in fingerprint carrier image to improve security and recognition accuracy. RST invariant embedding region was obtained by carrier image normalization and the core point detection. Multiple embedding based on SVD in the region and watermark optimizations in extraction process were presented in hiding scheme. The security of the proposed algorithm and its robustness against different kinds of attacks were verified by simulations.