【摘要】 经典Retinex算法假设场景中光照是平缓变化的,当光照变化比较强烈时,易产生"光晕"现象,为了提高光照条件变化下的人脸识别率,提出一种改进单尺度Retinex的光照人脸识别方法。采用双曲正切函数代替Retinex的对数函数对人脸图像进行亮度和对比度非线性增强;利用双边滤波代替Retinex的高斯滤波消除"光晕",采用Retinex消除光照不利影响,采用K近邻算法建立人脸分类器。结果表明,改进Retinex降低了时间复杂度,图像增强效果优于同类算法,提高了人脸识别率,很好地解决了"光晕"问题,具有光照鲁棒性,可适用于光照变化较强条件下的人脸识别。更多还原

【Abstract】 The classic Retinex algorithms hypothesize that the illumination of scene changes small. When the light changes strong, it is easy to produce a"Halo"phenomenon. In order to improve the face recognition rate in the change of illumination,this paper proposes an improved Retinex algorithm for face recognition. The face image is enhanced by nonlinear global and local, and then bilateral filtering is used instead of Gauss filtering for Retinex algorithm, and Retinex algorithm is used to eliminate the influence of illumination. K near algorithm is used to build face classifier. The results show that the proposed algorithm can reduce time complexity, enhance the image effect compared with the similar algorithms, and improve the face recognition rate. It has solved the problem of"Halo"very well.更多还原