中文摘要：

针对利用单幅参考图像进行人脸光照迁移时存在光照与人脸细节特征分离不彻底、效率低等问题,提出-种快速的人脸光照迁移算法. 该算法充分利用人脸的结构特征和阴影检测的先验知识,引导目标图像与参考图像进行大尺度层与细节层的分离; 然后将目标细节层与参考图像大尺度层进行合成以获得光照迁移后的亮度图像. 实验结果表明, 该算法可实现光照效果的快速迁移, 并且能较好地保持目标图像的人脸细节特征.

英文摘要：

Existing single-image-based lighting transfer algorithms have shortcomings in thoroughly sepa-rating lighting and facial details, and they are generally inefficient. With regard to these problems, a fast lighting transfer method is proposed based on some priors. At first, the prior knowledge of facial structure feature and shadow detection are used to well decompose the information of the large scale layer and the detail layer of the reference and target images, respectively. Then, the detail layer of the target image and the large scale layer of the reference image are synthesized to obtain the final transferred brightness image. Ex-periments demonstrate that with only a single reference image as input the method can efficiently transfer the lighting effects of the reference image to the target image, and the facial detail features of the target im-age can be well preserved.