照明变化人脸图像在独立成份分析空间中的分布

Distribution of Images of Same Face under Variant Light Conditions in Independent Component Analysis Subspace

应用独立成份分析(ICA)方法研究了不同照明条件下同一姿势人脸的图像,并用图像在ICA空间中任意三个独立成分的组合系数模拟得出:1)强度变化的同一幅人脸图像分布在一条直线上;2)照明方向变化下的同一姿势人脸的图像按一定规则集中分布.这两个结论可以分别用来判断两幅图像是否源于同一幅图像和一幅图像是否源于照明角度变化下的同一个姿势物体.

Independent component analysis （ICA） has been used to investigate the same face images under variant light conditions and three combination coefficients are randomly selected to simulate the distribution of these images in ICA subspace. Images that are originated from the same face image but with different intensities are observed to distribute along a line in the ICA subspace. And images that are originated from the same face but under different light directions are observed to distribute collectively in ICA subspace according to a certain rule. These two results can be used to effectively judge the problems of whether two images are originated from the same image and whether an image is belonged to the images of one object but under variant lighting angles,respectively.