Abstract: Aiming at human profile recognition，a statistical learning method is proposed based on linear correlation assumption between profile and frontal face images，then the parameters matrix is solved by means of multi-variable linear regression model，finally the frontal face image can be synthesized from its profile.Experimental results demonstrate that the lifelike effect can be achieved，and the method is easy to be realized without any auxiliary equipments and prior knowledge and therefore has a better practicability.

摘要： 针对侧面人脸的识别问题提出了一种基于统计的方法。基于人脸正面和侧面图像的线性相关性假设，通过对样本的学习建立多变量线性退化模型求解出系数矩阵，从而由输入的人脸侧面照片合成得到其正面的图像。实验表明这种方法所获的正面像接近于真实的照片，且无需任何的辅助设备和先验知识，具有一定的实用性。