Hasan Dogu(e112873@ceng.metu.edu.tr)
Ilknur Kaynar(e116488@ceng.metu.edu.tr)
Motivation and
Purpose:
What technology is best
suited to determine people’s identity? There are many different identification
technologies available, many of which have been in commercial use for years. The
most common person verification and identification methods today are
Password/PIN known as Personal Identification Number, systems. The problem with
that or other similar techniques is that they are not unique, and is possible
for somebody to forget loose or even have it stolen for somebody else. In order
to overcome these problems there has developed considerable interest in
"biometrics" identification systems, which use pattern recognition techniques to
identify people using their characteristics. Some of those methods are
fingerprints and retina and iris recognition. Though these techniques are not
easy to use. For example in bank transactions and entry into secure areas, such
technologies have the disadvantage that they are intrusive both physically and
socially. The user must position the body relative to the sensor, and then pause
for a second to declare himself or herself.
Face recognition from video
and voice recognition have a natural place in these next generation smart
environments, they are unobtrusive, are usually passive, do not restrict user
movement, and are now both low power and inexpensive.
Face Recognition
is a part of a project in TUBITAK BILTEN. In our summer practice we worked on
face detection and we therefore focused our
research towards recognition
of faces. Our goal at the end of the project is to develop a face recognition
program which works efficiently and simply. In addition to this, we are planning
to improve the algorithm by using neural networks and separation of training
classes.
Our aim is to
build a face recognizer that works under varying poses, the difficult part of
which is to handle face rotations in depth. So, our recognizer is planned to be
able to identify faces which are rotated in y axis, too.
Project
Description
Our subject is to
implement a system which recognizes faces using eigenface method. The system
will take an image as input and try to find a match in its database.
Furthermore, the algorithm will be improved by using neural network and
classification of the face database.
The work to be
performed consists of two parts. First part is the construction of eigenfaces.
Second part is the identification of the given image i.e if there is a match in the database(known faces). The
algorithm will be implemented in Windows environment using Visual
C++.
In Turkey,
TUBITAK BILTEN’s Signal Processing group works on a project called GAYE which
includes face detection and recognition. The project will be used in
security
systems.
In the world,
face recognition and detection is used as a research project in universities.
Furthermore, there exist many companies working on biometric solutions for
security. The equipment used for such security systems consists of a digital camera and a PC connected to
the camera. More information about these products can be reached from the face recognition
home page (http://www.cs.rug.nl/~peterkr/FACE/face.html).
Face recognition rises from
the moment that machine started to become more and more "intelligent" and had
the advance of fill in, correct or help the lack of human abilities and senses.
The subject of face recognition is as old as computer vision and both because of
the practical importance of the topic and theoretical interest from cognitive
science.
The methods used for face
recognition are :
Eigenface method, template matching, graph matching, linear subspace method, neural network method and fisherface method.
The eigenface
approach applies the Karhonen-Loeve transform for feature extraction. It greatly
reduces the facial feature dimension and yet maintains reasonable discriminating
power.
The neural network approach , though some variants of the algorithm work on feature extraction as well, mainly provides sophisticated modeling scheme for estimating likelihood densities in the pattern recognition phase.
The template matching
method operates by performing direct correlation of image segments. Template
matching is only effective when the query images have the same scale,
orientation and illumination as the training images.
In graph matching
method , objects are represented with sparse graphs whose vertices are labeled
with a multi-resolution description in terms of a local power spectrum and whose
edges are labeled with geometrical distances.
In the linear
subspace method for each face three or more images are taken under different
lighting directions. By using these 3 images , a 3D basis for the linear
subspace is constructed.
To perform recognition, the distance of a new image to each linear subspace is calculated and the face corresponding to the shortest distance is chosen.
Fisherface method is a
refinement of the eigenface algorithm. It further reduces the eigenspace by the
fisher’s linear discriminant(FLD). FLD selects the subspace in such a way that
the ratio of the between-class scatter and the within-class scatter is
maximized. It is reported that the Fisherface algorithm outperforms the
eigenface algorithm on the facial database with wide variations in lighting
conditions [2].
As correlation
methods are computationally expensive and require great amounts of storage, it
is natural to pursue dimensionality reduction schemes. A technique now commonly
used for dimensionality reduction in computer vision-
particularly in face
recognition- is Principal Component Analysis(PCA). PCA techniques, also known as
Karhonen-Loeve methods , choose a dimensionality reducing linear projection that
maximizes the scatter of all projected samples.
Eigenface method
is based on an information theory approach that decomposes face images into a
small set of characteristic feature images called eigenfaces, which may be
thought as the principal components of the initial training set of face images.
Recognition is performed by projecting a new image into the subspace spanned by
the eigenfaces and then classifying the face by comparing its position in the
face space with the positions of the known individuals[1].
1-) Project Proposal
Report(10.10.2001)
2-) First
prototype finished. (20.10.2001)
3-) Project
Progress Report
4-) Last
prototype finished. (1.12.2001)
5-) Final report
finished.
6-) Demo of the
program
WP1: Analysis
Phase
Task1:Understanding the main idea
behind eigenfaces
Task2:Survey about
eigenfaces
Task3:Proposal Report
writing
Task1: Main design of the
program.
Task2 :Design of the user interface
Task3: Implementation of the first part(eigenface
construction)
Task4: Progress report writing
Task5: Implementation of the second part (identification of
faces)
WP3: Final
Part
Task1: Testing the program
Task2: Collecting statistical information about the program
Task3: Final report writing
Hardware:
A PC with Windows 98
operating system.
Software:
Microsoft Visual
Studio
An image processing tool( like Adobe Photoshop)
Bibliography:
[1] Turk, M., and Pentland, A.,
“Eigenfaces for Recognition”, Journal of Neuroscience, vol.3, no1
1991
[2] Belhumeur, P.,
Hespanha, J. , and Kriekman,
D., “Eigenfaces vs. Fisherfaces : Recognition using
class specific linear projection”, IEEE Trans. Pattern Analysis Machine
Intelligence, vol.19 ,no.7, pp.711-720, July 1997
[3] Pentland , A., Moghaddam, B., Starner, T., and
Turk, M., “View-based and Modular Eigenspaces for Face
Recognition” , proc. IEEE Computer Society Conf. Computer Vision and Pattern
Recognition, pp.84-91, Seattle, WA,1994
Roles:
Project Manager: Hasan Dogu
Project Coordinator: Ilknur Kaynar
Research Coordinator: Ilknur Kaynar
Implementation Manager: Hasan Dogu