A comprehensive comparative performance analysis of Laplacianfaces and Eigenfaces for face recognition

Abstract

This paper provides a comprehensive comparative analysis of the performance of locality preserving projections (LPPs)‐based Laplacianfaces, which is a recently introduced algorithm with the more traditional, principal component analysis (PCA)‐based Eigenfaces. All possible combinations of neighbourhood defining distance metrics, classifier distance metrics and number of retained eigenvectors have been tried on different imaging environments. The FERET facial database was chosen which provides enough diversity in illumination, facial expressions and aging. CsuFaceIdEval, an open source platform, is used for this comparison and recognition rates are studied in detail. As a result of our detailed analysis, we provide best combination of selected parameters to extract the best results from these two algorithms.

Keywords:

• face recognition
• locality preserving projections
• principal component analysis
• machine vision
• pattern recognition

Portions of the research in this paper use the FERET database of facial images collected under the FERET programme, sponsored by the DOD Counterdrug Technology Development Program Office. The authors are thankful to Mr Xiaofei He for his valuable comments and suggestions.