International Journal of Engineering and Advanced Technology (IJEAT)

ISSN: 2249 – 8958, Volume-8, Issue-5S, May, 2019

Face Recognition System Based on LBPH

Algorithm
Abhishek Pratap Singh, SunilKumar S Manvi, Pratik Nimbal, Gopal Krishna Shyam

The contributions of this work tend to develop a real-time
Abstract: In this modern time, identifying a person using a face
face recognition system that will acknowledge face in poor
is a standard biometric approach to distinguishing an individual
from others. So techniques are required to identify a face must illumination condition and fetch information about that
be quick and sufficiently enough to work in real time. But there image from database
are many difficulties within the execution of face identification
in low lighting condition. In this paper, we have proposed a The paper has been divided into 7 sections. The first section
system that is using Local Binary Patterns Histogram algorithm introduces the readers to the essential of face recognition
for identifying a face. It can recognize both front and side faces technique and our contribution. The second section offers
and upgrade the value of poor enlightened picture and also existing work in details. Third section briefs regarding the
expands the recognition rate in real time. face recognition algorithm. Forth section offers the
methodology of the planned system. Fifth section deals with
Index Terms: Face recognition, LBPH, Histograms, the experimental setup. The sixth section is regarding result
Identification Process and discussion and the last is the conclusion

I. INTRODUCTION
II. RELATED WORK
Automatic face identification is an essential theme of Ojala et al. [10 ] initially used Local Binary Pattern (LBP).
computer vision and pattern recognition analysis. It will LBP refers unremarkably to displacement pixels with an
distinguish the personalities and different data as indicated 8-bit dual code from the neighborhood of the pixels. For a 3
by the optical highlights of face picture, having an square, the center component estimate is ablated from that of
exceptionally wide possibility for advancement. It is typically each of its eight neighboring pixels and is relegated to a bit
utilized in validation, criminal examination, police based on the indication of the subtraction result 1 or 0. The
investigation, automaton intelligence, and healthful science, created bits are then connected for all neighboring pixels and
etc. encoded into double strings. The dual strings inferred are
referred to as Local Binary Patterns or LBP codes. The
There are various techniques utilized in the face recognition
decimal estimate of the center component's LBP code (xc,
system. Each distinctive features underneath varied
yc). The LBP administrator LBP (P, R) produces 2p different
conditions like illumination, expression and posture
yield values, about 2p distinctive two-fold examples formed
amendment. Among them, for our motivation that is
by the P pixels within the varying R space. It has been shown
confront discovery in the video stream, we tend to focus on
that there is a lot of information in particular examples than
three noteworthy methods and that we looked their execution
others. It is conceivable to simply use a set of 2p double
underneath modified improvement condition to find an
examples to portray the surface of the photographs. These
honest coordinating and low process expenses.
examples were named by Ojala et al. as uniform examples.
As the time passed, almost several researchers have
developed a range of face detection and recognition All face images are viewed as the formation of small patterns
algorithms, as well as Deep Convolution Network that the LBP operator could recognize successfully. For face
algorithm[1],Histograms of Oriented Gradients (HOG) identification, Ahonen et al. [11 ] presented an LBP
algorithm[2],Sparse Coding(SC) algorithm[3],Gabor feature primarily based facial illustration. They have bifurcated
algorithm[4], Local Binary Pattern (LBP) algorithm[5], facial images into M minor non-covering areas R(0), R(1),
Linear Discriminate Analysis (LDA) algorithm[6], however ..., R(M) to compare the facial structure. The LBP histograms
additionally within the unceasing development. There are extracted from each sub-region are then combined into a
three different face identification techniques in the OpenCV single, spatially enhanced histogram feature..
library:, Fisherfaces[7], Eigenfaces[8] and Local Binary
Pattern Histogram (LBPH)[9 ]. In [12] as suggested by Ahonen et al, the texture
classification LBP approach describes the occurrence of LBP
codes in an image being collected into a histogram. The
Revised Manuscript Received on April 24, 2019 classification is then carried out by calculating the
Abhishek Pratap Singh, School of C & IT, REVA University, Bangalore,
Sunil Kumar S Manvi, School of C & IT, REVA University, Bangalore, similarities between straight-forward histograms. However,
Pratik Nimbal, School of C & IT , REVA University, Bangalore, India. a loss of spatial data results in
Gopal Krishna Shyam, School of C & IT , REVA University, Bangalore, considering the same
methodology for illustration of

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E10060585S19/19©BEIESP 26 & Sciences Publication
 Face Recognition System Based on LBPH Algorithm

the facial image. An approach to achieving this goal is to use

the LBP texture descriptors to create abundant local facial
descriptions and combine them into a global description.
These local characteristics are primarily based on ways in
which there are many robust causal or illumination
variations compared to traditional ways. The essential
methodology for LBP is based primarily on the facial
description proposed by Ahonen et al. dividing the facial
image into local areas and extracting the LBP texture
descriptors separately from each area. Then the descriptors
are joined together to describe the face globally. Effectively,
this histogram features a face description on three completely
different location levels: The LBP histogram labels contain
data on unit pixel-level patterns, labels are added over a
small region to integrate data at a regional level, and local
histograms are also joined together to provide a global facial
description. In [13], a face identification method with a high
rate of detecting a face has been marked into 3 stages.
Initially, a new picture is used for illustrating their
properties, namely, “Integral Image”. Then, an easy
AdaBoost classifier is utilized to choose a tiny variety of
crucial visual highlights from a really massive set of effective
properties. In the end, it is important to collaborate with the
classifiers during a cascade to extract the background regions
of the picture.
III. LOCAL BINARY PATTERN HISTOGRAM Figure 1: LBPH algorithm flowchart
The image is divided into cells (4 x 4 pixels) for the encoding
Local Binary Patterns (LBP) is a perceptible descriptor style of features. It is contrasted by using a clockwise or
used in the classification of computer vision. LBP is the counter-clockwise bearing of surrounding pixel values.
specific case of the 1990 proposed Texture Spectrum model. The value of each neighbor's intensity is compared to the
In 1994, LBP was represented for the first time. Since then, it central pixel. The location is assigned a 1 or a 0 depending on
has been found as a solid element for classifying texture. the difference whether it is higher or lower than 0. The result
More specifically, once LBP is combined with the descriptor gives a single cell an 8-bit value.
histogram of oriented gradients (HOG). It obviously Figure 2 shows the matrix calculation comparing the value of
improves the execution of identification on some datasets. the middle element of the matrix with the neighboring
Figure 1 illustrates the LBPH algorithm flowchart diagram. elements.

Figure 2: LBPH creating an 8-bit number

If the illumination condition of the image is changed, the

final result is equivalent to the previous result. Histograms
are used in larger cells as well as the frequency of values that
make system robust. Edges can be identified as the quality
changes by dissecting the results in the cell. It is possible to
obtain feature vectors by calculating the values of all cells
and connecting the histograms. Pictures can be grouped by
ID-connected handling methods. Input pictures are classified
using the same procedure and the data set is contrasted and
separation is obtained. It is very well recognized by setting a
limit value whether it is a known or obscure face. Figure 3
represents the values of the matrix when the light intensity is
wavering.

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E10060585S19/19©BEIESP 27 & Sciences Publication
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-8, Issue-5S, May, 2019

determining whether the image contains any faces in the

frame or not. When any face is detected, other facial features
are checked and a square frame is drawn on the face.
Otherwise, it starts reading other pictures. Figure 5 shows the
flow chart of the detection process.

Figure 3: If the brightness changes, the results will be the same.

IV. METHODOLOGY
The proposed face recognition approach has four main
steps: module for image acquisition, module for extraction of
features, module for training classifier database, and module
for classification. Initially, the face datasets are collected by
the image acquisition module. Then, a series of salient
features are extracted by applying feature extraction module.
These facial features are used to analyze face landmarks
which represent human identity information. In the next Figure 5: Face detection flow chart.
process, the classifier is trained for recognizing the face. In
the last module, the system recognize face image and fetch B. Feature Extraction
information about the person from the SQLite database. The
system flow chart is shown in Figure. 4 For extracting the facial features from image, the LBP
operation is used that compares the intensity value of every
component with the 8 nearest neighbour pixels values [17].
If the value of the neighboring pixel is greater than the value
of the centered pixel, it will assign 1 to its neighboring pixel,
otherwise it will assign 0. For each pixel, this task provides
an 8-bit string. A decimal value of an 8-bit pixel string
determines the LBP value. Fig. 6 shows this operation.

Figure.6 Extraction of LBP feature from the image

The input image is divided into many small sub-images after

the application of the LBP operation and the histograms of
the LBP value of each sub-images are extracted. Then all
histograms are linked to make an image-representing feature
vector and used to train a facial recognition classifier.

C. Dataset

We designed our own dataset, with 500 individual photos.

Figure 4: Face recognition system workflow Throughout the image acquisition process, face images are
cropped and converted into gray images, then these images
are saved in the same folder to make face databases for
A. Face Detection
extraction tasks. After this, the standardization technique is
For face detection, OpenCV is used that introduces a Haar applied to all images to reduce noise and set the correct
cascade classifier [15],[16]. The classifier of the Haar image scaling position to quickly obtain the result of
cascade uses the AdaBoost algorithm to locate numerous recognition. Dataset images
image facial features. Initially, it takes an input image using are shown in Figure 7
the camera and converts that color image into a grayscale
image. After this, it loads Haar cascade classifier for

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E10060585S19/19©BEIESP 28 & Sciences Publication
 Face Recognition System Based on LBPH Algorithm

Figure 7. Dataset images

D. Face Recognition Figure 8. Face detection

For the face recognition process, Local Binary Pattern In the next process, the system extracts the LBP texture
Histogram algorithm is applied. The LBP operator uses local features from every input dataset image and train them with
binary patterns to reduce the local spatial distribution of a the Haar Cascade classifier.
face image [18]. The LBP operator is a collection of binary
pixel value ratios in the center at regular pixel intervals and
is around 8 pixels. TABLE I. TRAINING IMAGES STATISTIC

It is shown in the below equation. Total Images Trained Images Training Time
500 500 10 sec

Finally, the system is performed face recognition process.

Where ic represents the index value of the middle pixel and First it detects all faces in the image, then it extracts facial
(xc, yc) shows 8 close surrounding pixels data. feature from the input test face image. After this input feature
vector is compared with the trained image dataset model
V. EXPERIMENTAL SETUP using Haar Cascade Classifier. If the input test feature vector
is matched with the trained model, it recognizes the face and
To develop this system python programming language is fetches the information about that image from the SQLite
used. OpenCV (3.0) library and other tools PyCharm and database. The system can detect multiple faces in the image if
SQLite3 are used in this system. PyCharm is an integrated any new face image comes into the camera frame system
development Environment (IDE) employed for creating recognize the face on base on facial features
computer programs, specifically for the Python language.
SQLite is an RDBMS contained in a C programming library.
As compared to other database management systems, it is not
a client-server database system. It is embedded in the end
program.
The whole project is divided into three parts facial detection,
training face images and last the recognition of the faces.
These all 3 operations are performed in PyCharm and
information regarding a person like a name, age, gender is
kept in the SQLite database.

Figure 9. Recognizing face images.

VI. RESULTS AND DISCUSSION
To develop a face recognition system, first we have to VII. CONCLUSION
prepare the face dataset for training. To develop face dataset,
we used Face detection method that detects the face in real In the proposed face recognition system, we used Local
time camera and captured face images. That captured images Binary Patterns histogram algorithm for recognizing faces.
are saved into dataset folder for feature extraction and The whole procedure is divided into three major components,
training processes. i.e. detection of faces, facial feature extraction, and
classification of the image. The Face detection process
In the initial phase, system asks information about the person describes the face of a person in input image. In feature
like name, age, gender and then the camera will open that extraction, facial landmarks are extracted and to make an
captures 500 images of a person on different face position LBPH histogram that gives the completely unique result and
and expression. The information about person is saved into then in recognition process the histogram of the input image
SQLite database and captured images are saved into dataset is compared with database histogram using the classifier.
folder with same unique id. The result shows that the system can recognize a known and
unknown person. The
problem in this proposed
method is that, if the rate of
change in the frame is very

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E10060585S19/19©BEIESP 29 & Sciences Publication
 International Journal of Engineering and Advanced Technology (IJEAT)
ISSN: 2249 – 8958, Volume-8, Issue-5S, May, 2019

high and occlusion occurs, then the system will not robustly University of Visweshwariah College of Engineering, Bangalore and Ph. D.
degree in Electrical Communication Engineering, Indian Institute of Science,
recognize faces. This problem can be considered as future Bangalore, India. He has vast experience of more than 30 years in teaching in
work. This proposed system will assist security agencies in Electronics / Computer Science and Engineering. His research interests are:
identifying criminals, for attendance purpose, to search any Software Agent based Network Management, Wireless Networks, Multimedia
Networking, Underwater Networks, Wireless Network Security, Grid and
individual in a surveillance camera. Cloud computing, E-commerce, and Mobile computing. He has published
around 160 papers in national and international conferences, 115 papers in
national and international journals, and 15 publications/books/books.chapters.
REFERENCES He received best research publication award from VGST Karnataka in 2014.
1. J. Hai qiang LONG,Tai zhe TAN. Computer Simulation, 2017, 34(1):
322-325.
Pratik Nimbal has completed M.tech with specialization
2. J. HU Liqiao, QIU Runhe. Face recognition based on adaptive weighted in computer science and Technology in year 2019 at REVA
HOG. Computer Engineering and Applications, 2017, 53(3): 164-168. University.
3. J. Olshausan B A, Field D J.Emergence of simple-cell receptive field
properties by learning a sparse code for natural images. Nature, 1996,
381(6583):607-609.
4. J. WU Qi, WANG Tang-hong, LI Zhan-li. Improved face recognition Dr. Gopal Kirshna Shyam received BE and M.Tech and
algorithm based on Gabor feature and collaborative representation. Ph.D in Computer science and engineering from VTU,
Computer Engineering and Design, 2016, 37(10): 2769-2774. Belagavi. He has handled several subjects for UG/PG
Students like Algorithms, Computer Networks, Web
5. J. CHAO W L, DING J J, LIU J Z. Facial expression recognition based on
programming, Advanced Computer architecture,
improved local binary pattern and class-regularized locality preserving
Information security, Computer Concepts and C
projection. Signal Processing, 2015, 117:1-10.
Programming. His research interest includes Cloud Computing, Grid
6. J. Yu yan JIANG, Ping LI, Qing WANG. Labeled LDA model based on computing, High performance computing etc. He has published about 10 papers
shared background topic. Acta Electronica Sinica, 2015, 2013, (9): in highly reputed National/International Conferences like IEEE, Elsevier etc.
1794-1799. and 5 papers in Journals with high impact factor like Elsevier Journal on
7. J.P. Bellhumeur, J.Hespanha, D. Knegman. Eigenfaces vs fisherfaces: Network and Computer Applications and International Journal of Cloud
Recognition using class specific linear projection.IEEE Transactions on computing (INDERSCIENCE). His research articles on Cloud computing
Pattern Analysis and MachineIntelligence. Special Issue on co-authored by Dr. Sunilkumar S. Manvi have been cited by several
FaceRecognition, 1997, 17 (1) :711-720. researchers. He is a lifetime member of CSI and is actively involved in
8. J. Turk M, Pentland A. Eigenfaces for recognition. Journal of cognitive motivating students/faculties to join CSI/IEEE/ACM societies.
neuroscience, 1991, 3(1): 71-86
9. M. Joe Minichino, Joseph Howse. Learning OpenCV 3 Computer Vision
with Python. 2016 16-82
10. T. Ojala, M. Pietikäinen and D . Harwood, "A Comparative Study of
Texture Measures with Classification based on Featured Distributions,"
Pattern Recognition, vol. 29, no. 1, pp. 51-59, January 1996
11. T. Ahonen, A. Hadid, and M. Pietika ̈inen. Face recognition with local
binary patterns. In European Conference on Computer Vision (ECCV),
pages 469–481, 200..
12. Ahonen, T., Hadid, A, and Pietikäinen, M. (2006), Face Description with
Local Binary Patterns: Application to Face Recognition. IEEE Trans.
Pattern Analysis and Machine Intelligence 28(12):2037-2041
13. Viola, Paul, and Michael J. Jones. "Robust real-time face detection."
International journal of computer vision 57.2 (2004): 137-154.
14. Varun Garg, Kritika Garg, Face Recognition Using Haar Cascade
Classifier, Journal of Emerging Technologies and Innovative Research
(JETI),, December 2016
15. Zheng Xiang, Hengliang Tan, Wienling Ye. The excellent properties of
dense gird-based HOG features on face recognition compare toGaborr
and LBP, 2018
16. Md. Abdur Rahim, Md. Najmul Hossain, T. Wahid, Md. Shafiul Azam,
"Face recognition using Local Binary Patterns (LBP)", Global Journal of
Computer Science and Technology Graphics &Vision,, Global Journals
Inc. (USA), Volume 13 Issue 4 Version 1.0 Year 2013.
17. T. Chen, Y. Wotao, S. Z. Xiang, D. Comaniciu, and T. S. Huang, “Total
variation models for variable lighting face recognition” IEEE
Transactions on Pattern Analysis and Machine Intelligence, 28(9):2006
18. Zhao and R. Chellappa "Robust face recognition using
symmetricshape-from-shadingg" Technical Report, Center for
Automation Research ,University of Maryland,, 1999

AUTHORS PROFILE

Abhishek Pratap Singh has completed M.Tech with

specialization in Computer Science and Technology in
year 2019 at REVA University Bangalore India.

Dr. Sunilkumar S. Manvi received his B. E. Degree from

Karnataka University, M. E. degree in Electronics from the

Published By:
Blue Eyes Intelligence Engineering
Retrieval Number: E10060585S19/19©BEIESP 30 & Sciences Publication