Color space MS-based feature extraction method for face verification

In the last years, face verification has gained a great interest in the pattern recognition community and in many application fields. It is among the most attractive research areas because face images can be captured in a non-intrusive way. Many algorithms have been developed in this area, among them the Principal Component Analysis (PCA) is a typical face based technique which considers face as global feature. However, PCA method suffers the disadvantage in terms of discriminant ability and large computational load. Its performance deteriorates especially in the present of varying lighting condition and facial expression. This paper proposes a face authentication method using color information based on a new simple feature extraction technique using face image Mean and Standard deviation (MS) features. The results of evaluation carried out on XM2VTS face database show that MS-based features outperforms PCA-features for all tested color spaces, with best score for the MS-based features using the S component of the HSV (90.44%). In addition, to be stable in different color spaces, MS-based technique proves to be quite stable between validation and tests conditions, which showed strength and robustness of MS-features. The Robustness and ease of extraction make MS-features ideal candidates for a real time application with limited resources.