改进单尺度 Retinex 的复杂光照人脸识别算法

摘要

Under complex illumination conditions,face recognition performance declines dramatically,the single-scale Retinex algorithm is easy to produce "halo"phenomenon.Aiming at this,we propose a complex illumination face recognition method which is based on im-proved single-scale Retinex algorithm.First,the Laplace gradient operator is used to enhance the edge information of lightened face images, and then based on single-scale Retinex algorithm the centre adaptive adjustment fitting function in accord with human visual system is intro-duced to enhance each colour component of the image and to improve the contrast of face image.Finally Yale B and CMU-PIE face database are used for simulation test.Results show that the proposed method enhances the contrast of face image,improves face image visual effect, and is conducive to improving face recognition rate and recognition speed,it can meet the requirements of face recognition under complex illu-mination conditions.%在光照复杂条件下人脸识别性能明显下降。针对单尺度 Retinex 算法易产生“光晕”现象，提出一种改进单尺度 Retinex算法的复杂光照人脸识别方法（ISSR）。首先利用拉普拉斯梯度算子增强光照人脸图像的边缘信息，然后在单尺度 Retinex 算法的基础引入符合人眼视觉特性的中心自适应调节拟合函数，以增强图像各个颜色分量，提高人脸图像对比度，最后采用 Yale B 和CMU-PIE 人脸库进行仿真测试。结果表明，ISSR 算法增强了人脸图像的对比度，改善了人脸图像的视觉效果，有助于提高人脸图像识别率和识别速度，能满足复杂光照条件下的人脸识别要求。