Face recognition using selected 2DPCA coefficients

Face recognition based on principal component analysis (PCA) has provided successful results. This leads researchers to propose several variants of PCA such as the two-dimensional PCA (2DPCA). The results reported using this technique have demonstrated that it has an enormous potential as feature extractor for face recognition. However, the main drawback is the high number of coefficients produced. In this paper we propose to use a feature selection algorithm to analyze and to discard coefficients that are not relevant to the face recognition task. Experimental results on the ORL and the Yale databases have shown that the number of coefficients extracted by the 2DPCA can be reduced in about ten times while improving recognition rate. Keywords-Feature selection, PCA, MOGA.