基于GLS模型的光照鲁棒性人脸的识别

针对线性子空间模型在处理具有阴影的人脸图像时出现的不足之处,提出了GLS模型,并将其用于不同光照下的人脸识别。按照测试图像与正确模型间距离尽可能小的原则构造了一个确定最优分组数和子空间维数的标准;采用SVD方法和K平均聚类法将像素分组,并确定每个分组的线性子空间模型;计算测试图像到每个GLS模型中所有分组的线性子空间模型的距离之和,进而识别人脸图像。经假设检验统计表明,基于该模型的方法在处理不同光照下的人脸图像时,效果明显优于其他方法。

Robust face recognition under various illuminations based on grouped linear subspace model

Grouped linear subspace model was presented to make up disadvantages of linear subspace model while dealing with face images with shadows.According to the principle that the distance between the test image and right model is as small as possible,a criterion was constructed to determine the optimal number of groups and dimensions.SVD method and K-means clustering method were used to group the pixels and determined a linear subspace model for each group;distances between test image and subspace models of each group were added up for each GLS model and the test image was recognized as the minimum one.Experimental results with hypothesis testing show that face recognition system based on GLS model is superior to other methods in dealing with face images under various illuminations.