An illumination angle classification method in the face recognition based on the similarity of light estimated images
Dong Ren, Zhenyu Jing, Kai Ma · 2014
To avoid destroying the face texture information, the uneven extent of illumination should be first considered before the face preprocessing. The light source direction is one of the most important factors that influence the illumination. In this paper, Discrete Cosine Transform was used to get the light estimated image. We attempt to use the PCA, 2DLDA and image similarity methods to classify the light source direction. After each method are analyzed and summarized. An illumination angle classification method based on image similarity is proposed. The experiment result shows that the proposed method achieves a better classification rate.