A face portion based recognition system using multidimensional PCA

In this paper a new human face recognition algorithm based on localized face portion of an image is proposed. Extracted pure facial image is decomposed using curvelet transform and its selected subband is utilized for classification. Subband exhibiting a maximum standard deviation is dimensionally reduced using an improved dimensionality reduction technique, i.e., bidirectional two-dimensional principal component analysis to generate distinctive feature sets. These feature sets are used for training and testing an extreme learning machine classifier. Notable contributions of the proposed work include significant improvements in classification rate, speed and negligible dependence on the number of prototypes.