On Recognizing Human Faces From Frontal Views


Yılmaz S.

Kahramanmaras Sutcu Imam University Journal of Engineering Sciences, vol.8, no.1, pp.41-44, 2005 (International Refereed University Journal)

  • Publication Type: Article / Article
  • Volume: 8 Issue: 1
  • Publication Date: 2005
  • Title of Journal : Kahramanmaras Sutcu Imam University Journal of Engineering Sciences
  • Page Numbers: pp.41-44

Abstract

 This paper consists of two main parts. In the first one, we compared L0.5, L1, and L2 metrics and showed which one is better suited for the area of face recognition for 2-dimensional gray level face images. In the second part, we have demonstrated the effectiveness of different facial expressions for the purpose of identification of individuals. We have shown that L1 norm performs as good as other two metrics but takes almost as half time to process images. We have also compared recognition performance of four different expressions on two test sets and showed that blank and angry expressions better represent a face Simulation results are reported on a dataset of 600 people selected from CNNL face database.