IRJET- A Review on Various Approaches of Face Recognition

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 06 | June 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 3188
A Review on Various Approaches of Face Recognition
Abhishek Kumar1, Dr. S. A. Khan 2, Er. Ashish Chand3, Er. Parwinder Kaur4
1M.Tech Scholar Department of Electronics and Communication SSIET, Patti Tarn Taran-Punjab(India)
2Associate Professor Department of Electronics & Communication SSIET, Patti Tarn Taran-Punjab(India)
3Assistant Professor Department of Electronics & Communication SSIET, Patti Tarn Taran-Punjab(India)
4Assistant Professor Department of Electronics & Communication SSIET, Patti Tarn Taran-Punjab(India)
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - Face recognition based on low resolution
images provide low accuracy as compare to various
approaches that provide better results under high-resolution
parameters. A method for face recognition by using the
principal component analysis with Eigen vectors have been
used in recent research. Euclidean distance classifier is used
for the matching between the trainingandtesting images. The
purpose of the research is to improve the accuracy for the low
resolution images. By analysing various approaches for face
recognition there is need to develop a new approach which
can provide better results using texture features for blurred
images.
Key Words: Face Recognition, DCT, DWT, LTP, PCA, ELTP, FAR,
1. INTRODUCTION
The face plays a major role in our social intercourse in
conveying identity and emotion. The human ability to
recognize faces is remarkable. We can recognize thousands
of faces learned Throughout our lifetime and identify
familiar faces at a glance even after years of separation. The
skill is quite robust, despite large changes in the visual
stimulus due to viewing conditions, expression, aging, and
distractions such as glasses or changes in hairstyle.
Computational models of faces have been an active area of
research since late 1980s, for they can contributenotonlyto
theoretical insights but also to practical applications,suchas
criminal identification, security systems, image and film
processing, and human-computer interaction, etc.However,
developing a computational model of face recognition is
quite difficult, because faces are complex, multidimensional,
and subject to change over time. Generally, there are three
phases for face recognition, mainly face representation, face
detection, and face identification.
1.1 Face representation
Face Representation is the first task, that is, how to model a
face. The way to represent a face determines the successive
algorithms of detectionandidentification.Fortheentry-level
recognition (that is, to determine whether or not the given
image represents a face), a face category should be
characterized by generic properties of all faces; and for the
subordinate-level recognition (in other words, which face
class the new face belongs to), detailed features of eyes,
nose, and mouth have to be assigned to each individual face.
There are a variety of approaches for face representation,
which can be roughly classified into three categories:
template-based, feature-based, and appearance-based. The
simplest template-matching approaches represent a whole
face using a single template, i.e., a 2-D array of intensity,
which is usually an edge map of the original face image. In a
more complex wayoftemplate-matching, multipletemplates
may be used for each face to account for recognition from
different viewpoints. Another important variation is to
employ a set of smaller facial feature templates that
correspond to eyes, nose, and mouth, for a single viewpoint.
The most attractive advantage of template-matching is the
simplicity; however, it suffers from large memory
requirement and inefficient matching. In feature-based
approaches, geometric features, such as position and width
of eyes, nose, and mouth, eyebrow's thickness and arches,
face breadth, or invariant moments, are extracted to
represent a face. Feature-based approaches have smaller
memory requirement and a higher recognition speed than
template-based ones do. They are particularlyuseful forface
scale normalization and 3D head model-based pose
estimation.
1.2 Types of Face Recognition
Face recognition scenario has been classified into two
different categories. These categories arebasicallymatching
of face image with single image or group of images. These
two scenario of face recognition has been explained below.
1.2.1 Face verification (or authentication)
Face verification is the process to verify person’s identity
that has been claimed to be matched with template. Face
verification is the process of one to one match that
comparing a query face with claiming face. Verification is to
be done on the basis of features of templateimageandquery
image. To evaluate performance of the face verification
different parameters have to be classifiedthathasbeenused
for different ROC curves. Falseacceptanceandfalserejection
rate has to be computed to compute verification rate of the
claimed query. A good verification system should balance
these two rates based on operational needs.

1.2.2 Face identification (or recognition)
Face identification is the process of matching of single
person image with multiple imagesavailableinthedatabase.
This face identification process is also known as one too
many matching process. In this process query face image is
compared with all the template images available in the face
image database. The image that is closest match with the
database images is most identifying image that match with
test image. The query face image features has been
compared with the database face images so that can identify
that maximum matched image on the basis of distance. The
distance has been computed with all the images available in
the database of facial images. These distances have been
arranged numerically in ascending order. The top level
image distance is maximum matched image with the test
image available in the database. If the top arranged distance
is minimum then that define maximum matched image has
been found with test image. On the basis of these test results
the parametersforfacerecognitionsystem.FalseAcceptance
Rate (FAR) and False Rejection Rate (FRR) had to be
computed for performance evolution of face identification
system.
1.3 Challenges In Face Recognition
 Scale: The scale of a face can be handled by a rescaling
process. In Eigen face approach, the scaling factor can be
determined by multiple trials. The idea is to use multi
scale Eigen faces, in which a test face image is compared
with Eigen faces at a number of scales. In this case, the
image will appear to be near face space of only the closest
scaled Eigen faces. Equivalently, we can scale the test
image to multiple sizes and use the scaling factor that
results in the smallest distance to face space.
 Variation in Poses: Varying poses result from the change
of viewpoint or head orientation. Different identification
algorithms illustrate different sensitivities to pose
variation.
 Variation in Iluminance: To identify faces in different
luminance conditions is a challenging problem for face
recognition. The same person, with the same facial
expression, and seen from the same viewpoint,canappear
dramatically different as lighting condition changes. In
recent years, two approaches, the fisher face space
approach and the illumination subspace approach, have
been proposed to handle different lighting conditions
 Disguise: Disguise is another problem encountered by
face recognition in practice. Glasses,hairstyle,andmakeup
all change the appearance of a face. Most researchwork so
far has only addressed the problem of glasses.
1.4 Parameters Used
False acceptance rate (FAR) is the measure of rate that a
biometric system will incorrectly accept the access
attempted by an unauthorised user. False acceptance rate is
computed by using the ratio between the false accepted and
identification attempts done.
False rejection rate (FRR) is the measure in the biometric
security system is that when a system incorrectly rejectsthe
authentication done by an authorised person. FRR is
computed by using the ratio between the false rejections
done by system that are made by authorised person to the
total attempts done.
2. APPROACHES USED
2.1 Appearance Based Recognition Approaches
Appearance based face recognition process consist of Eigen
faces that preserves global structure of image sub space,
Fischer face that preserves discriminating information and
laplacian face that preserves local structure of image
subspace. These approaches have been used for face
recognition based on appearance based recognition.
2.1.1 Principal Component Analysis (PCA)
PCA was invented by Karl Pearson for reduction of
dimensions of the dataset that contain redundant
information. This leads to reduction in variables from
dataset that known as principal components from the
dataset which accounts most variation occurred in different
variables of dataset. Eigen faces are the principal
components from the distribution of the faces or Eigen
vectors are the 2-dimessional feature subspace from N*N
covariance image of facial part. Each face image is a linear
combination of different face images in Eigen sub space.
Eigen face cosmists mean of all the images that are available
in the dataset images that has been used for matching
process. Eigen values of query image have been matched
with dataset Eigen values for recognition process.
In this process of Eigen face mean image has been computed
from all the images available in the dataset that has been
represented by X1, X2………………….Xn.
(1)
After computation of mean image dataset imagessubtracted
image has been reconstituted for the images that has been
used for development of covariance matrix.
(2)
Eq. (2) represents subtracted image the group of images
have been used for reconstructions of facial images
covariance matrix
M=A.AT (3)

Fig -1: Training dataset images
After this process the Eigen features and theEigenvaluesfor
the face image was computed The Eigen value µi and Eigen
vector vi has been computed on the basis different equations
that are represented as.
Fig -2: Mean image constructed from dataset images
By using the Eigen values and Eigen vectors Eigen face
matrix has been generated that is used as features for
matching purposes.
Fig -3: Eigen faces from all images
After the generation of Eigen space, the matrix is computed
by using different face samples. The database features has
been stored. These feature sub spaces have been used for
recognition purpose using distance classifier represented in
eq. (4)
(4)
2.1.2 Local Discriminant Analysis (LDA)
LDA was an approach that had been developed for
preserving local discriminating information from facial
image. Face images can be divided intotwodifferentdivision
that are spatial and frequency division. Intheprocessofface
recognition various factors have been considered for
recognition that varies due to variation in age, gender and
person’s characteristics. These issues are not comprised in
PCA so to overcome this major issue in recognition new
approach had been purposed that works on the principal of
Fischer-face.
In this process face images available in the dataset are
belongs to different classes. Multiple images under different
condition must be available in a single class. At least one
image must be available in test image dataset.
After this instances each available image with two-
dimensionalm X narray of intensity valuesI(x,y), LDA
approach construct the lexicographic vector expansion ϕ ϵ
RmXn. This vector represents values of initial faces. Thus the
set of all faces in the feature space is treated as a high-
dimensional vector space.
(5)
(6)
(7)
(8)
By using these different equation LDA approach computes a
transformation coefficients that increases between class
scattering and decreases within class scattering.
(9)
The linear transformation is given by a matrix U whose
columns are the eigenvectors of Sw
-1Sb(called Fischerfaces).
These fischerfaces have been computed using eq. (9).
(10)
By using above mathematical expression fischerfaces
features for dataset has been a measure that has been used
for computation of face recognition accuracy based on
threshold values.

Fig -4: Separation of feature sub space cluster
Fig 4 represents separation of different person’s classes in a
clustered view using ORL face dataset.
2.1.3 Texture Feature Based Recognition
Approaches
Face modalities have been recognized using different
approaches that compute texturefeaturesfromfacial images
that have been used for face matching. In this process facial
part contain images have beenusedandfeaturespointsfrom
whole face image has been computed for computation of
texture features that has been used with machine learning
approaches or distance classifier approachesforrecognition
process. In the process of texture feature analysis Local
Binary Patterns (LBP), Local Ternary Pattern (LTP) and
Equalized Uniform Local Binary Pattern (EULBP) have been
used for feature extraction.
3. BINARY PATTERN (LBP)
Local Binary pattern operator was designedfor extraction of
texture description from images. This operator assigns a
label to every pixel available in the image using a window of
3*3 moving on all the images pixels. This window use center
pixels and neighbor pixel values for assigning a label to each
pixel available in the image. These labels have been
considered for result in binary format.Lateronresearchhad
been done to increase window size for huge image so that
time [] complexity for features extraction can be reduced. In
this extension window size can be used a set of pixels that
are evenly spaced on a circle centeredatthepixel thatallows
any radius and number of sampling points. To find out
optimized solution bilinear interpolation has been used
where neighboring points does not fall into the center pixel.
In the following, the notation (P, R) will be used for pixel
neighborhoods which mean P sampling points on a circle of
radius of R.
Fig -4: (a) basic LBP operator, (b) the circular
Figure 7(a) represents binary code computation using LBP
operator on single image mask using 3 * 3 neighbor pixel
values and fig 7(b) repents extended (P,R) used for radius of
selection of higher number of sampling point in a single
window.
Let S (u) is the binary matrix that has to be computed based
on center pixel values that has been donatedbyIc,Ni denotes
the neighbor pixel values under a window.
(11)
(12)
After computation of all thebinarycodesfortheentireimage
these codes have been histogram concatenated to compute
feature sub space using binary labels. These feature sub
spaces have been used for recognition process. Various
contents available in the face images can affect recognition
accuracy so that weight age can be associated to all the
images contents that play vital role in recognition process.
4. CONCLUSION
Face recognition under different condition is a challenging
problem that causesmismatchingininvestigationprocess.In
this paper novel approaches have been discussed that has
been used based on the basis of appearance and texture
based face recognition. By analyzing feature of various
approaches and texture based approaches LDA is an
approach that is exhausted over the PCA in the process of
whole image based recognition. LDA consistoflowerfeature
dimensions that cause less time complexity and provides
better accuracy, whereas in texture based face recognition
LTP is much better than other approaches because prone to
noise, and uniform regions.

REFERENCES
[1] Tomasz Marciniak, AgataChmielewska,
RadoslawWeychan, Marianna Parzych, Adam
Dabrowski, (2015), “Influence of low resolution of
images on reliability of face detection and recognition”,
International Journal of Multimedia tools applications,
Vol. 74, pp. 4329-4349
[2] MuweiJian, Kin-Man Lam (2015), “Simultaneous
Hallucination and Recognition of Low-Resolution Faces
Based on Singular Value Decomposition”, IEEE
Transactions on Circuits and Systems for Video
Technology, pp. 1-14
[3] Rizwan Ahmed Khan, Alexandre Meyer, Hubert Konik,
SaidaBouakaz (2013), “Framework for reliable, real-
time facial expression recognition for low resolution
images”, international journal of pattern recognition,
pp. 1-34
[4] Hae Min Moon and Sung Bum Pan (2013), “The LDA-
based Face Recognition at a Distance using Multiple
Distance Image”, International Conference
onInnovative Mobile and Internet Services in
Ubiquitous Computing, pp. 249-255, DOI
10.1109/IMIS.2013.50.
[5] Shengcai Liao, Jain, A. K. and Li, S. Z. (2013), “Partial-
Face Recognition: Alignment Free Approach” IEEE
Transactions on Pattern Analysis and Machine
Intelligence, Vol. 35, Issue 5, pp. 1193-1205, ISSN
0162-8828.
[6] Nazari, S. and Moin, M.-S. (2013) “Face Recognition
Using Global and Local Gabor Features”, Iranian
Conference on Electrical Engineering, pp. 1-4, DOI
10.1109/IranianCEE.2013.6599704.
[7] Baohua Yuan, Honggen Cao and Jiuliang Chu (2012),
“Combining Local Binary Pattern and Local Phase
Quantization for Face Recognition” International
Symposium on Biometrics and Security Technologies,
pp. 51-53, ISBN 978-1-4673-0917-2.
[8] Rashid, R.D., Jassim, S.A. and Sellahewa, H. (2013) “LBP
Based on Multi Wavelet Sub-Bands Feature Extraction
Used for Face Recognition”, IEEE International
Workshop on Machine Learning for Signal Processing,
pp. 1-6, ISSN 1551-2541.

IRJET- A Review on Various Approaches of Face Recognition

More Related Content

What's hot (19)

Similar to IRJET- A Review on Various Approaches of Face Recognition (20)

More from IRJET Journal (20)

Recently uploaded (20)

IRJET- A Review on Various Approaches of Face Recognition