Recognition of Non-symmetric Faces Using Principal Component Analysis.
Krishnan Nallaperumal, G. Jothi, G. Lakhs Aaron · 2004
All the human faces are symmetric approximately up to 90 %- 95 % only. No face is 100 % symmetric. Based on this property that 10 % difference is present in any human face is given as the input to the face recognition system. The intensity variations of the faces are equalized first. Then the left and right face difference is given as the input to the database and to the face recognition system. This will reduce the storage half the size compared to any other method and provides better recognition for the nonsymmetric and non-frontal faces. It also works well for the symmetric faces. Here eigenface method is used for recognition.