Basics of face detection and facial recognition

Basics of face detection and facial recognition
Roland Matusinka
Óbuda University, John von Neumann Faculty of Informatics
Phone: (36)-(30)-782-4470, [email protected]
Abstract – Face detection and recognition is one of
the most used technologies today. From social sites to
CCTV systems it can be found almost anywhere. The
goal of this paper is to introduce the reader to the
basics of these technologies.
The aim of face detection algorithms is to determine the
size and location of human faces in digital images,
ignoring anything else, like trees or other objects in the
image [1]. This can be achieved either by using a reference
image containing known faces, or by finding key features
of a face, like the eyes, the nose or the mouth. In the first
case the faces are matched bitwise, meaning that any slight
alteration in the processed image (like a different facial
expression) can cause the matching to fail. After we found
the face in the image, we can use facial recognition the
determine who that face belongs to. Facial recognition
systems extract landmarks and unique features from a face,
and try to match them to a facial database. This facial
database can consist of geometric (structural) data of faces
or a gallery of images used as reference data.
Given an arbitrary image, the goal of face detection is to
determine whether or not there are any faces in the image
and if present, return the image location and extent of each
face. Some challenges of face detection are the following
 Pose: There is no guarantee that a person will
look straight into a camera every time a picture
is taken of him/her. Therefore, more often than
not, a human’s face will be at an angle with the
camera (frontal, 45-degree, profile, upsidedown, etc.) and thus the proportions of the face
will be altered.
 Presence or absence of features: Facial
features such as beards, mustaches, and glasses
may or may not be present, and there is a great
variability of these features in size, color, and
 Facial expression: The appearance of a face are
affected by the person’s facial expression.
 Occlusion: Face images can be occluded by
other objects. In an image of a group of people
some faces may be partially occluded by other
Image orientation: Face images vary for
different rotations about the camera’s optical
Image condition: When the image is created
factors such as sources, and distribution of
lighting, and properties of the camera affect the
appearance of a face.
There are many closely related problems of face
detection. Face localization aims to determine image
position of a single face. This is a simplified detection
problem with the assumption that an image contains only
one face. The goal of face feature detection is to detect the
presence and location of certain facial features such as
eyes, nose, nostrils, eyebrow, lips, mouth, ears, etc., with
the assumption that there is only one face in the image.
Face recognition or face identification compares an input
image (probe) against a database (gallery) and reports a
match, if any. Face authentication, face tracking, and facial
expression recognition is also worth mentioning as
common problems. Face detection is the first step in any
automated systems which solves the above problems [2].
There are different approaches to face detection. Often
they are divided into four categories. These categories may
overlap, so an algorithm could belong to two or more
categories [3].
 Knowledge-based
methods that encode our knowledge of human
faces. They try to capture our knowledge of
faces and translate them into a set of rules.
 Feature-invariant methods: Algorithms that
try to find structural features of a face that exist
in any condition (different pose, different
lighting, etc.).
The above methods are mainly used for face
 Template
algorithms compare input images with stored
patterns of faces or features. These methods are
used for both face localization and detection.
 Appearance-based methods: A specialization
of template matching methods where the pattern
database is learnt from a set of training images.
They are designed to use in face detection.
Face recognition’s core problem is to extract
information from photographs. This feature extraction
process can be defined as the procedure of extracting
relevant information from a face image. This information
must be valuable for the later step of identifying a subject
with an acceptable error rate.
There are many feature extraction algorithms. Most of
them are used in other areas than face recognition.
Researchers in face recognition have used, modified and
adapted many algorithms and methods to their purpose.
Feature selection algorithms aim to select a subset of the
extracted features that cause the smallest classification
error. The importance of this error is what makes feature
selection dependent to the classification method used. The
most straightforward approach to this problem would be to
examine every possible subset and choose the one that
fulfills the criterion function. However, this can become
an unaffordable task in terms of computational time.
Therefore researchers make an effort towards creating a
satisfactory algorithm, rather than an optimal one [3].
Once features are extracted and selected, the next step is
to classify the image. Appearance-based face recognition
algorithms use a wide variety of classification methods.
Sometimes two or more classification methods are
combine to achieve better results. On the other hand
model-bases methods match the sample with a template or
model, then a learning method can be used to improve the
algorithm. Either way, classifiers have a big impact in face
There are three concepts that are key in building a
classifier – similarity, probability, and decision boundaries
 Similarity: Patterns that are similar should
belong to the same class. The idea is to establish
a metric that defines similarity and a
representation of the same class.
 Probability: Some classifiers are build based
on a probabilistic approach. Bayes decision rule
is often used. The rule can be modified to take
into account different factors that could lead to
miss-classification. Bayesian decision rules can
give an optimal classifier, and the Bayes error
can be the best criterion to evaluate features.
Therefore, probability functions can be optimal.
Decision boundaries: This approach can
become equivalent to a Bayesian classifier. It
depends on the chosen metric. The main idea
behind this approach is to minimize a criterion
(a measurement of error) between the candidate
pattern and the testing patterns.
Illumination. Many algorithms rely on color
information to recognize faces. Features are extracted
from color images, although some of them may be grayscale. The color that we perceive from a given surface
depends not only on the surface’s nature, but also on the
light upon it. As many feature extraction methods relay on
color/intensity variability measures between pixels to
obtain relevant data, they show an important dependency
on lightning changes. Not only light sources can vary, but
also light intensities may increase or decrease, new light
sources added. Entire face regions be obscured or in
shadow, and also feature extraction can be impossible
because of solarization. The big problem is that two faces
of the same subject but with different illumination can
show more differences between them than compared to
another subject [3].
Pose. Pose variation and illumination are the two main
problems researchers have to face in face recognition. The
majority of face recognition methods are based on frontal
face images. Recognition algorithms must implement the
constraints of the recognition application. Many of them,
like video surveillance, video security systems or
augmented reality home entertainment systems take input
image data from uncontrolled environment. The
uncontrolled environment constraint involves several
obstacles for face recognition.
Other problems worth mentioning are:
 Occlusion,
 Optical technology (format of input images),
 Facial expressions.
[1.] Face detection – Wikipedia.
Last visited: 2014. 04. 27.
[2.] Ming-Hsuan Yang, David J. Kriegman, Narendra Ahuja Detecting Faces in Images: A Survey
Last visited: 2014. 04. 27.
[3.] Ion Marqués – Face Recognition Algorithms
Last visited: 2014. 04. 27.
Related flashcards
Create Flashcards