To search, Click below search items.


All Published Papers Search Service


2D&3D-ComFusFace: 2D and 3D Face Recognition by Scalable Fusion of Common Features


Juan Zhou, Yongping Li, Jingyan Wang


Vol. 12  No. 3  pp. 30-36


In traditional 2D and 3D face recognition systems, different features are extracted from 2D and 3D face images, and then are fused to improve the recognition performance. The shortage of these methods is that they neglect the intrinsic complementary features between 2D and 3D data. In this paper, we investigate the possibility of extracting and scalable fusing common features from 2D intensity and 3D depth face images, and develop a novel 2D and 3D face recognition method-- 2D&3D-ComFusFace, which represent and fuse some common global and local features of 2D and 3D data. A novel pose normalization method for 3D range data is also proposed before transiting them to be depth image. After preprocessing, two global features--2D Principle component Analysis (2DPCA), 2D Fisher Linear Discriminate Analysis (2DFLD), and a local feature--Local Binary Pattern (LBP) are extracted from both 2D intensity image and 3D depth image. Then the matching scores are computed and fused by weighted sum rule to get a further improved performance. The experiments are carried out on CASIA3D database, and significant improvements of both recognition rate and EER are achieved.


2D intensity images, 3D depth images, global feature, local feature, Fusion