Guan,Y., Fei,X., GuoCan,F., Ching.Y.S.
This paper addresses face recognition for de-lighting variation problem via 3D reconstruction based on SFS. Firstly, we improve the Geometric-based SFS by introducing the integrability constraint as one of the regular terms. This improvement preserves preferably the local curvedness of the recovered surface. Secondly we propose a novel method to investigate human face recognition in the illumination varying case using local topography information extracted from intensity images by SFS algorithms, such as curvedness and shape index. The experimental results have shown the curvedness and shape index has good ability to represent 3D local features, and also it is insensitive to lighting variations since only 3D information is involved. Comparing with typical face recognition based on PCA+LDA, the proposed method has demonstrated the good performance. This implies local topology characteristics are effective attributes for face recognition in de-light variations by using SFS.