|
|
|
| A Method Based on Two-Dimensional Feature Extraction in the Complex Domain for Face Recognition |
| Han Ke①;Zhu Xiu-chang①; Wang Hui-yuan② |
| ①College of Telecom. & Info. Eng., Nanjing University of Posts & Telecom., Nanjing 210003, China; ②College of Info. Science & Eng., Shandong University, Jinan 250100, China |
|
|
|
|
Abstract A new method based on two-dimensional feature extraction in the complex domain is proposed for face recognition in this paper. First, face images are performed by mirror transform, and the original face samples and the corresponding mirror samples are used to compute the even symmetry samples and the odd symmetry samples, respectively. The even symmetry samples and the odd symmetry samples are used to form complex samples by an odd-even weighted factor. Then the complex image within-class scatter matrix and the complex image between-class scatter matrix are defined in the complex domain, respectively, to calculate a family of optimal complex projection axes, and complex face samples are projected onto the family of optimal complex projection axes to extract the face features. Finally, a nearest neighbor classifier is employed to classify the extracted features. The method in the paper is evaluated on the NUST603 face image database. Experimental results show the proposed method achieves better performance.
|
|
Received: 13 June 2006
|
|
|
|
|
|
|
|
|
2008, Vol. 30 
