Abstract: Illumination analysis plays an important role in face recognition and tracking systems.A new approach is proposed to solve the parameters of the illumination includes energy and direction，which is developed based on the model of diffuse reflection and ambient reflection.A large number of experiments show the algorithm converges quickly，and the light could fit the image precisely，the estimated light direction averagely deviated from the angle of true azimuth is 5.43°.Another advantage of the algorithm is it could cope with the different pose，which means the capability of synthesizing the face image in an arbitrary pose and illuminating condition.

Key words: illumination estimation, illumination fitting, image synthesis

摘要： 光照分析对人脸图像的识别与跟踪有重要意义，提出一种人脸图像光照参数准确估计算法。该算法根据光照模型构造光照参数求解的迭代方程，光照参数根据不同需要可同时包含光照能量，光照方向矢量和环境光能量等信息。实验显示迭代过程能迅速收敛，并且光照角度估计准确，yaleB实验数据的水平角度估计平均错误率仅为5.43°。由于算法隐含了人脸姿态描述，因此其另一优势是可以同时对人脸姿态和光照条件进行综合分析，大量实验测试说明该算法用于光照估计有效可行。

关键词: 光照估计, 光照拟合, 图像合成