IRJET- Face Recognition by Additive Block based Feature Extraction

International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 06 Issue: 03 | Mar 2019 www.irjet.net p-ISSN: 2395-0072
© 2019, IRJET | Impact Factor value: 7.211 | ISO 9001:2008 Certified Journal | Page 5294
FACE RECOGNITION BY ADDITIVE BLOCK BASED FEATURE
EXTRACTION
Krishnaja M1, S Mohan2, Dr. V Jayaraj3
1PG Scholar, Dept. of Electronics & Communication Engineering, Nehru Institute of Engineering & Technology,
2Assistant Professor, Dept of ECE, Nehru Institute of Engineering & Technology.
3Professor and Head, Dept of ECE, Nehru Institute of Engineering & Technology.
---------------------------------------------------------------------------------***-------------------------------------------------------------------------------
Abstract - Face recognition is the main process for
biometric recognition process. The basic fundamental
procedure for pose demonstration and illumination variation
method. To overcome this problem the proposed method
consists of Chirp Z-Transform (CZT) and Goertzel algorithm.
These processes are used for pre-processing stages. Themajor
matching of the recognized process will be based on the
feature extraction method. Gray Level Co-occurance Matrix
(GLCM) is used in the proposed work to achieve feature
extraction method. Each stages of the Face recognition
method is overcome with the new accuracy value. Thus the
segmentation of facial region is done using combination of
CZT and Geortzel algorithm. These algorithm are used for
increase in the illumination normalization value and the
intensity range. Thus the proposed feature extraction
technique is the block based additive fusion of the input face
image. Thus, the face is selected and trained using classifier.
These trained images are classified using Euclidean distance
classifier. The proposed approach has been tested on four
benchmark face databases, viz., Color FERET, HP, Extended
Yale B and CMU PIE datasets, and demonstrates better
performance compared to existing methods in the presence of
pose and illumination variations.
Key Words: CZT and Geortzel algorithm, FERET, HP,
Extended Yale B and CMU PIE datasets, Euclidean classifier
1. INTRODUCTION
In view of expanding dangers concerning security
observation ID verification and diverse endless logical
purposes face recognition for plays a crucial and mitigating
role into present world. There are a number of methodsthat
right now exist for FRL database. For methods may likewise
be generally partitioned into threesubdivisionsinparticular
preprocessing capacity extraction and have choice. Efficient
face cognizance procedures depend on great component
extraction and highlight choicestrategies. Work extractionis
a strategy that gets rid of the repetitive learning saving the
data that is basic. Trademark extraction may likewise be
most likely classified into geometric arranged techniques
and measurable arranged systems. factual systems like
practically identical to statute factor examination PCA free
part assessment ICA and straight discriminate investigation
LDA make utilization ofmathematical methodologiesthough
the geometric based strategies handle the face as auxiliary
substance through speaking to the face as set of separations
and points between the characteristic face components.
1.1 Chirp Z-turninto(CZT)andGoertzelalgorithms
Preprocessing
The proposed novel preprocessing strategy utilizes a blend
of CZT and Goertzel calculation connected to singular
squares of picture. This assigned mix performs
enlightenment standardization of the picture.
1.2 Block Based Additive Fusion for Feature
Extraction
The proposed added substance square arranged system
includes partitioning the picture into squares of equivalent
size and superimposing (added substance combination)
them on the whole to type one resultant square. This
resultant square includes the dominating features of all the
individual squares, inserted into the size of a solitary square
and therefore causes the emergency to measurements of
solitary square blocks of each pixels.
2. LITERATURE SURVEY
Change focused capacity extraction has created in most
recent events. A standout amongst the most standard ways
includes the usage of DCT and DWT2. Various techniques
have been employed to optimize the DWT based feature
extraction. One such way is thresholding [3]. Modifications
have been moreover made to the DCT strategy through
partitioning the picture into squares and making utilization
of DCT to individual blocks [4]. One more trademark
extraction way alluded to as Stationary Wavelet turn into
(SWT) was used to overcome the pose related problems [5].
Frequency spectrum can also be used for feature extraction
[6]. A method based on facial symmetry and DCT can also be
used as a feature extractor [7]. Innovative preprocessing
strategies had been furthermore utilized to sustain the FR
methodology. Foundation disposal fixated on entropy is a
powerful preprocessing technique [8]. History expulsion
using k- implies grouping will likewise be utilized as a
preprocessing technique8. An additional efficient
preprocessing framework was once to make utilization of a
Laplacian Gradient covering process [9]. To battle
brightening renditions, best the dim part of the previewwas

specifically increasingly reasonable, keeping up the more
splendid areas immaculate. This is called specific more
noteworthy enlightenment technique3.DWT,Singularvalue
Decompostion (SVD) and CZT used to be utilized for
watermarking [10]. This calculation ﬁrst changes over the
time space photograph into recurrence sub-bands using
DWT, at that point these sub-groups are changed to Z area
through making utilization of CZT and ﬁnally watermarked
by methods for SVD. CZT can likewise be connected to
process MRI data [11].
3. METHODOLOGY
3.1 FACE RECOGNITION
The face is a primary part of who you might be and the way
people identify you. Except withinthecaseofidentical twins,
the face is arguably a man or woman'smostparticularbodily
characteristic. Whilst humans have the innate capability to
respect and distinguish exceptional faces for millions of
years, computers are simply now catching up. For face
attention there are two varieties of comparisons. The
primary is verification. This is where the process compares
the given character with who that charactersaysthey'reand
gives a sure or no determination. The 2d is identification.
That is the place the process compares the given individual
to all the different contributors within the database and
gives a ranked list of suits. All identification or
authentication applied sciences operate utilizing the
following four levels:
Determine in levels in Face recognition
A. Seize: A bodily or behavioral pattern iscapturedbymeans
of the system throughout enrolment and likewise in
identification or verification process. B. Extraction: certain
information is extracted from the sample and a template is
created. C. Evaluation: the template is then when compared
with a brand new sample. D. Suit/non healthy: the
procedure decides if the aspects extracted from the brand
new samples are a fit or a non-match.
Face cognizance technology analyze the particular shape,
sample and positioning of the facial aspects. Face
consciousness is an extraordinarily intricatetechnologyand
is basically software situated. This Biometric Methodology
establishes the analysis framework with tailor-made
algorithms for each style of biometric gadget. Face
recognition begins with a photograph, searching for a
individual in the snapshot. This may capture be entire
making use of several approaches, together withmovement,
epidermis tones, or blurred human shapes. The face
realization approach locates the top and sooner or later the
eyes of the individual. A matrix is then developed centered
on the traits of the character’s face. The approachofdefining
the matrix varies in keeping with the algorithm (the
mathematical approach used by the pc to participate in the
comparison). This matrix is then compared to matrices that
are in a database and a similarity score is generated for each
and every assessment. There are two predominant
techniques to the face cognizance quandary: Geometric
(feature situated) and photometric (a view founded). As
researcher curiosity in face consciousness persisted, many
special algorithms were developed, three of which had been
well studied in face consciousness literature. Consciousness
algorithms may also be divided into two predominant
procedures:
1. Photometric stereo: Used to recuperate the form of an
object from a number of photos taken beneath
extraordinary lights conditions. The form of the recovered
object is outlined by using a gradient map, which is made
from an array of surface common. (determine 3.4)
Figure 3.1.Four: Photometric Stereo photo
2. Geometric: Is headquartered on the geometrical
relationship between facial landmarks, or in other phrases
the spatial configuration of facial aspects. That means that
the fundamental geometrical aspects of the face such
because the eyes, nose and mouth are first located after
which the faces are labeled on the groundwork of quite a lot
of geometrical distances and angles between aspects.
Figure3.2: Geometric Facial cognizance
4. PROPOSED WORK
The proposed system consists of following block diagram as
shown in the figure 4.1

Figure 4.1 Proposed Work
4.1 PRE-PROCESSING
CZT computes the Z develop into at M facets in a Z-aircraft.
The Goertzel algorithm is most valuable when an N point
DFT is to be computed using less number of coefficients.The
CZT algorithm transforms the image into Z area. When
Goertzel algorithm is applied to this convertedphoto,actsas
a reconstruction algorithm to the image. The reconstruction
produces an image that's inverted with admire to its usual.
The mixture of CZT and Goertzel algorithm performs
illumination normalization of photograph.
Figure 4.1 input image
Figure 4.2 Grey Scale Conversion
4.2 ADDITIVE BLOCK BASED FEATURE
EXTRACTION
This feature extraction approach,thephotographissplitinto
blocks of equal size and then added. The number of blocks
and the dimensions of every block are dependent on the
quantity of rows of the preprocessed photograph. For a
photo of measurement m×n the number of blocks will also
be arbitrarily chosen to be multiples of m keeping the size of
each and every block is equal. In case the numberof blocksis
just not a particular more than one of the quantity of rows,
additional rows with zero values can be addedtoacquirethe
detailed multiple.
Figure 4.3 gamma correction
The notion can also be to keep an top-quality quantity of
facets that must be extracted. In this paper,threemethods of
dividing the snapshot is employed, i.e., division into four,
eight and 16 blocks. The input snapshot is first divided into
designated quantity ofblocksofequal dimension.CZTisthen
applied in my view to those blocks. That is adopted with
application of Goertzel algorithm on every block.

Figure 4.4 CZT and Goertzel algorithm
This mixture is used as preprocessing manner to enhance
the image. These blocks are then superimposed to provide a
single block from which the aspects are chosen. In view that
the character blocks are all added in the end the outcomesof
padding zeros does to not make contributions to any
anomaly in the characteristic extraction approach.
Photographs with pose variancehavelesscorrelationamong
their blocks. Thus, to beat the pose variance trouble, images
ought to be divided into 16 blocks. Outcome exhibit that for
pose variant databases, the highest ARR is got for sixteen
blocks. This is due to the fact that 16 blocks has mixed lotsof
the aspects present individually in each and every of the
blocks.
On the other hand, the illumination variant pics have
excessive correlation among the blocks considering the fact
that there is no variant in pose. Thus, dividingthesegraphics
into 4 blocks proved to be ample. In this case, dividing the
picture into sixteen blocks proved to have scale back ARRas
compared to division of picture into four blocks. The
correlation among the blocks of the photographismisplaced
when the columns of the photo are divided. As a result, to
continue symmetry of the face only rows are divided. The
stem plots of the snap shots after software of block situated
system that are close to equivalent, underlines the fact that
Additive Block situated method can be utilized as feature
extraction procedure for recognition in both pose variant
and illumination variantportraits.Nowthepictureisdivided
into eight blocks and each and every block is processed in
my view. These are then brought to receive the resultant
block. This resultant block is then passed to the feature
selector for feature resolution. The more the quantity of
blocks, lesser the quantity of elementsextractedconsidering
the fact that the dimensions of every block and accordingly
the consequent block reduces. However it is alsoessential to
keep an ideal number of features and not increase the
number of blocks randomly. This might in the end fail to
provide excellent consciousness. The facets that are
extracted in the above procedure are subjected to feature
decision method bygray-degreeco-incidencematrix(GLCM)
which helps in extra reduction of points.
4.3 GRAY LEVEL CO-OCCURANCE MATRIX (GLCM)
GLCM is defined as thegreydegreeco-incidencematrix.Here
the feel aspects of photos are extracted and saved in a
matrix. GLCM is without doubt one of the simplest matrix
ways to extract the feel elements. GLCM elements are
extracted for all of the images in the database and the input
picture are saved for performingaffine moments. The4 most
often used homes akin to vigor, Entropy, distinction and
Inverse difference second are used to cut down the
computational complexity. The co-prevalence matrix is a
statistical mannequin and is valuable in a type of photo
analysis applications akin to in biomedical, remote sensing,
industrial defect detection techniques, and so forth. Grey
level Matrix is used to extract features situated on the gray
degree worth of pixels. The points are fundamental forevery
classification algorithms. Here texture elements of photos
are extracted.
The GLCMs features are stored in a matrix, the place the
number of GLCM is calculated. The GLCM aspects are
extracted by way of the variance and difference of entropy
know-how. Utilizing the affine moment invariants process
the feature extraction is finished to extract points akin to
eyes, eyebrows and lips. It is accomplished by means of
utilizing facial expression awareness of exceptional feelings
like irritated, worry, sad, pleased, surprise and traditional.
Making use of these facial expressions the images are
converted in to binary graphics for extracting the facts.
4.4 EUCLIDEAN CLASSIFIER
To measure the extent of matchingbetweenthetrainandthe
experiment photographs, Euclidean distance components
are used. Euclidean distance between two facets is defined
because the straight line distance between the features.
3. CONCLUSION
There are multiple methods in which facial recognition
systems work, but in general, they work by comparing
selected facial features from given image with faces within a
database. It is typically used in security systems and can be
compared to other biometric such as fingerprint or iris
recognition systems. An efficient algorithm for face
recognition has been proposed.

Figure 4.5 Face recognized
Thus the proposed work consists of accuracy rate of 90% in
the classifier stage.
REFERENCES
[1] A.S. Syed Navaz, T. Dhevi Sri & Pratap Mazumder.
(2003) ‘Face Recognition Using Principal Component
Analysis and Neural Networks’, International Journal of
Computer Networking, Wireless and Mobile
Communications (IJCNWMC) ISSN 2250-1568Vol. 3,
Issue 1, 245-256.
[2] Derzu Omaia ,JanKees v. d. Poel,Leonardo V.
Batista.(2010), ‘2D-DCT Distance Based Face
Recognition Using a Reduced Numberof Coefficients’.
[3] Fatma Zohra Chelali, A. Djeradi and R. Djeradi. (2009),
‘Linear Discriminant Analysis for Face Recognition’ in
Proc. Third International Conference on Automatic Face
and Gesture Recognition, pp 336-341, Nara Japan.
[4] Gaurav Kumarand Pradeep Kumar Bhatia.(2014),
‘ADetailed Review of Feature Extraction inImage
Processing Systems’,FourthInternational Conferenceon
AdvancedComputing and CommunicationTechnologies,
pp.5-12.
[5] Hossein Sahoolizadeh, B.ZarghamHeidari,and C.Hamid
Dehghani. (2008) ‘A New Face Recognition Method
using PCA, LDA and Neural Network’, International
Journal of Computer Science and Engineering.pp. 2-4.
[6] Marian Stewart Bartlett, Javier R. Movellan, and
Terrence J. Sejnowski. (2002), ‘Face Recognition by
Independent Component Analysis’ , IEEE Transactions
On Neural Networks, Vol. 13, No. 6,pp.1450-1464.
[7] Mohamed Rizon, Muhammad Firdaus Hashim, Mohd
Rozailan Mamat. ‘Face Recognition using Eigen-faces
and Neural Networks’, American Journal of Applied
Sciences 2 (6): 1872-1875, 2006, ISSN 1546-9239.
[8] Muzammil Abdulrahman, Yusuf G. Dambatta, A. S.
Muhammad, and Abubakar S. Muhammad.(2014),‘Face
Recognition Using Eigenface and Discrete Wavelet
Transform’, International Conference on Advances in
Engineering and Technology (ICAET), pp.510-513.
[9] Surya Kant Tyagi and Pritee Khanna.(2012), ‘Face
Recognition Using Discrete Cosine Transform and
Nearest Neighbour Discriminant Analysis’,IACSIT
International Journal of Engineering and Technology,
Vol. 4, No. 3.
[10] Urvashi Bakshi, Rohit Singhal.(2014), ‘A Survey on Face
Detection Methods and Feature Extraction Techniques
of Face Recognition’, International Journal of Emerging
Trends & Technology in Computer Science Vol.3,
pp.223-237.
[11] Yong Chen, Hao Feng, Xianbao Wang.( 2008), ‘Delong
Zhou ,Face Recognition Using Cubic B-spline Wavelet
Transform’,IEEE Pacific-Asia Workshop on
Computational Intelligence and Industrial Application.
[12] Dr. S. Vijayarani , S. Priyatharsini,(2015), ‘Facial Feature
Extraction Based On FPD and GLCM Algorithms’,
International Journal of Innovative Research in
Computer and Communication Engineering, Vol. 3,
pp.1514-1521.
[13] Hassan M, (2011) , ‘Smart Human Face Detection
System’, International Journal ofComputers,vol.5,no.2,
pp. 210-216.

IRJET- Face Recognition by Additive Block based Feature Extraction

Recommended

Recommended

More Related Content

What's hot

What's hot (20)

Similar to IRJET- Face Recognition by Additive Block based Feature Extraction

Similar to IRJET- Face Recognition by Additive Block based Feature Extraction (20)

More from IRJET Journal

More from IRJET Journal (20)

Recently uploaded

Recently uploaded (20)

IRJET- Face Recognition by Additive Block based Feature Extraction