The Negative Effects of Whitening Transformation in Face Recognition

Whitening transformation has widely been used in face recognition. Typically it is used in one of four ways: as a data preprocessing tool before feature extraction, to calculate statistically uncorrelated discriminant vectors, as a transitional step in calculating discriminant vectors, or as a data post-processing tool after feature extraction. The experimental studies presented in this paper demonstrate, however, that researchers´ faith in this technique is misplaced and that in fact whitening transformation is inappropriately used, and actually has a negative effect in face recognition without proper analysis.