Abstract: A method based on single-scale Retinex and skin color model is proposed for distortion of skin color in non-uniform illumination face image enhancement. The face image is converted to YCbCr color space, and the intensity and color components are enhanced by different methods. First, in order to compress the dynamic range of face image and enhance the details of shadow areas, the method based on single scale Retinex is applied on the intensity component images. Then, to improve the color quality of skin region, the value of chrominance components（Cb and Cr）of skin pixels are adjusted according to the aggregation characteristics of skin color in CbCr space. Experiments in the CAS-PEAL face database show that the proposed method can improve the image details and color quality of the facial region.

Key words: face image, image enhancement, Retinex theory, skin model

摘要： 针对非均匀光照彩色人脸图像增强中肤色失真问题，提出了一种基于单尺度Retinex和肤色模型的方法。将人脸图像转换至YCbCr颜色空间，并采用不同的方法分别处理Y分量和CbCr分量。针对亮度分量（Y）采用单尺度Retinex方法压缩图像的动态范围，增强图像暗处的细节信息；针对图像中肤色区域，根据肤色在CbCr空间具有聚集性的特点，调整亮度分量增强后肤色像素点色度分量（Cb和Cr）的值，改善肤色区域的颜色质量。在CAS-PEAL人脸库中进行实验，该方法与传统的人脸图像增强方法相比，在图像细节呈现能力和面部色彩真实程度方面均有提高。

关键词: 人脸图像, 图像增强, Retinex理论, 肤色模型