An improved PCA type algorithm applied in face recognition

In image recognition, principal component analysis (PCA) and many of its variants are widely used for simplicity and good results. There are some well known algorithms of this type: two dimensional PCA, two directional two dimensional PCA and the recently proposed extended two dimensional PCA. We present a two directional approach of extended two dimensional PCA which carries the advantages of 2D2DPCA (both side projection) and the ones of E2DPCA (information retained in the scatter matrix).