In Figure (a) the face is localized even though it is in a
non-frontal pose and despite the thin veil that covers it.
Figure
(b) shows the localization of a face despite numerous
other faces in the image. Figure
(a) shows the localization of
a blurry face with a large out-of-plane rotation and significant background
data. Figure
(b) shows the successful localization of a face
with a beard and glasses which do not pose a problem for the algorithm.
Additionally, the face is leaning backwards slightly.
Figure
(a) contains a localized face with a mustache and
dishevelled hair. Figure
(b) contains a face with thick
glasses. Figure
(a) shows the localization of a face with a
large in-plane rotation and unusual lighting. Figure
(b)
depicts a face which is situated in a complex background, under unusual
lighting, with in-plane rotation and at a small scale.
Figure
(c) depicts the localization of a face with dark skin
against a bright background. Note the algorithm's ability to localize the
faces in these everyday images despite variations in pose, expressions, facial
paraphernalia, lighting and background clutter.