The quality of the analyzed approaches for the feature extraction varies a lot. The
integral projection is the best choice for finding the center of the eyes. Almost 90
percent of the given eyes could be detected. But for the other two feature points the
algorithms are too vulnerable to disturbing factors like facial hair or bad exposure. So,
the differences in brightness between nose and face are mostly too small to detect the
edges of the nose. The extraction of the mouth has similar problems. The shape of the
mouth that greatly varies from face to face as well as beards impede useful results.