中文摘要：

为补偿人脸识别中的年龄变化，结合人脸形状与纹理信息，提出了一种新的人脸年龄图像合成方法来模拟年龄变化．首先将人群划分成8个年龄段，选取不同年龄段人脸的多个特征点，依据径向基函数变形技术，得到脸形变形函数，对人脸进行形状变形；接着对目标图像进行多方向滤波，提取原始纹理信息，然后根据使用神经网络模拟得到的年龄因子，选择相应的模板纹理信息，在原始纹理信息基础上进行置换和融合，得到新的纹理信息特征，并将得到的纹理特征叠加在已经变形的人脸图像上，合成年龄模拟图像．另外为进行更细微的纹理控制，使用高斯滤波器调节滤波程度．试验结果表明该方法简单易行，而且可以生成更接近目标年龄的人脸图像．

英文摘要：

In order to compensate the aging process in the face recognition, a new method was proposed for composing aged face images, considering both face shape and texture. First, a new warped face was obtained using the function trained according to radial basis function, and getting eight types of human faces in different ages; second, the image was filtered by multi-angled filters to extract the skin texture feature, which was translated and fused with the corresponding sub-images of the template image decided by the added age factor obtained by neural network, and the age-related texture information was added into the warped face. In addition, Guass filter was used to control the degree of the texture information of human face. finally a new aged face image was synthesized. Results of experiments show that this method can get a more actual face image.