Dwivedi Saloni, Gupta Nitika.

, International Journal of Advance Research, Ideas and Innovations in Technology

ISSN: 2454-132X
Impact factor: 4.295
(Volume3, Issue3)
Available online at www.ijariit.com
A New Hybrid Approach on Face Detection and Recognition
Saloni Dwivedi Nitika Gupta
M.Tech Student Assistant Professor (CSE Deptt.)
BBDU, Lucknow BBDU, Lucknow

Abstract: Face detection and recognition is an important paradigm when we consider the biometric based systems. Among
various biometric elements, the face is the most reliable one and can be easily observed even from a distance as compared to iris
or fingerprint which needs to be closely observed to use them for any kind of detection and recognition. Challenges faced by face
detection algorithms often involve the presence of facial features such as beards, moustaches, and glasses, facial expressions,
and occlusion of faces like surprised or crying. Another problem is illumination and poor lighting conditions such as in video
surveillance cameras image quality and size of the image as in passport control or visa control. Complex backgrounds also make
it extremely hard to detect faces. In this research work, a number of methods and research paradigms pertaining to face detection
and recognition is studied at length and evaluate various face detection and recognition methods, provide a complete solution
for image-based face detection and recognition with higher accuracy, a better response rate as an initial step for video
surveillance.

Keywords: Face Detection, Face Recognition, Eigen Faces, Viola Jones.

I. INTRODUCTION
Face detection is an easy visual task for human vision, however; this task is not easy and is considered to be a challenge for any
human computer interaction approach based on computer vision because it has a high degree of variability in its appearance. How
can computers detect multiple human faces present in an image or a video with the complex background? That is the problem. The
solution to this problem involves segmentation, extraction, and verification of faces and possibly facial features from a complex
background.

Face recognition from image or video is a popular topic in biometrics research. Many public places usually have surveillance
cameras for video capture and these cameras have their significant value for security purpose.
The actual advantages of face based identification over other biometrics are uniqueness and acceptance. As human face is a dynamic
object having a high degree of variability in its appearance, that makes face detection a difficult problem in computer vision. In this
field, accuracy and speed of identification are the main issues.
Face detection is a technology that persuades the sizes and the locations of human faces in the digital images. It recognizes faces
and neglects anything else, such as trees, bodies and the buildings. Face detection might be recognized as a more general illustration
of the face confinement. It is the center of all the facial analysis, e.g., face localization, face authentication, face recognition, facial
feature detection, face tracking, and facial expression recognition.

II. LITERATURE REVIEW

Face detection and recognition have always been an area of interest for the researchers worldwide. There are several types of research
in psychophysics and neuroscience studying about how we human performs recognition processes, and many of them have direct
relevance to engineers interested in designing algorithms or systems for machine recognition of faces. There are certain arguments
pertaining to this area. The first argument in these disciplines is that whether face recognition is a dedicated process against other
objects or pattern recognition tasks. There has been certain evidence that proves that face recognition clearly has many advantages
over other biometric features. Another argument which comes from a section of researchers is the holistic-based viewpoint which
claims that human recognizes faces by the global appearances, while the feature-based viewpoint believes that important features
such as eyes, noses, and mouths play dominant roles in identifying and remembering a person.
An Automated face recognition is an interesting area of computer vision problem with many commercial and law enforcement
applications. Mug shot matching, user verification and user access control, crowd surveillance, enhanced human computer
interaction all become possible if an effective face recognition system can be implemented. While research into these areas dates

© 2017, www.IJARIIT.com All Rights Reserved Page | 485

 Dwivedi Saloni, Gupta Nitika., International Journal of Advance Research, Ideas and Innovations in Technology

back to the 1960's, it is only very recently that acceptable results have been obtained. However, face recognition is still an area of
active research since a completely successful approach or model has not been proposed to solve the face recognition problem. The
inadequacy of automated face recognition systems is especially apparent when compared to our own innate face recognition ability.
Faizan Ahmad et.al [1] have discussed the various challenges in the area of image-based face detection and recognition. In this
paper, they have evaluated a number of face detection and recognition methods existing and went on to develop a system for the
said method’s evaluation as the first milestone for video-based face detection and recognition for surveillance. As mentioned in this
paper a general system overview is as shown below:

Figure 1: System Overview [1]

The developed system to evaluate the face detection and recognition methods which are considered to be a benchmark. Some
methods performed consistently over different datasets whereas other methods behave very randomly however based on average
experimental results performance is evaluated, five datasets been used for this purpose. As shown in the paper, Haar-like features
reported relatively well but it has much false detection than LBP which could be considered being a future work in surveillance to
reduce false detection in Haar-like features and for the recognition part gabor is reported well as it’s qualities overcomes datasets
complexity.
Amr El Maghraby et.al [2] in their paper have discussed a hybrid face detection method using Viola-Jones method and skin color
detection. In this paper, a fast, reliable automatic human face and facial feature detection is one of the initial and most important
steps of face analysis and face recognition systems for the purpose of localizing and extracting the face region from the background.
This paper presents a Crossed Face Detection Method that instantly detects low resolution faces in still images or video frames.
Experimental results evaluated various face detection methods, providing a complete solution for image-based face detection with
higher accuracy, showing that the present method efficiently decreased the false-positive rate and subsequently increased the
accuracy of face detection system in still images or video frames especially in complex backgrounds. The proposed method can
process different kinds of images and under different lighting conditions. The experimental results showed that our new approach
was able to achieve a higher detection rate than any of the 2 methods mentioned prior, and clearly improved Viola - Jones face
detection accuracy and decreasing false negative rates.
K M Poornima and Ajit Danti [3] have proposed human face recognition of still images using face detection by AdaBoost face
detector, region of interest (ROI) extraction, feature extraction using discrete wavelet transform (DWT), dimensionality reduction
by employing independent component analysis (ICA) and classification using k-Nearest Neighborhood (k-NN) classifier. The
proposed system is evaluated by conducting the experiments on Faces 94 database. For experimenting the proposed system, 40
classes from “Faces94” database are chosen. For training and testing, 40 class each and each class containing 10 images are
considered. Out of 10 images, 5 images from each class are used for training and remaining are used for testing. Out of 40 classes,
13 classes are taken from the female dataset, 21 classes are from the male dataset and 6 classes are from male staff dataset. Therefore,
total 400 images are considered, out of which 200 images are used for training and remaining 200 images are used for testing. The
proposed system achieved about 83.5% of recognition rate when k value is 1.Recognition rate decreases ask value increases.
Zahra Sadri Tabatabaie1, [4] present a hybrid face detection system using a combination of appearance-based and feature-based
methods. They have combined Viola and Jones face detection method with a color-based method to propose an improved face
detection method. They discuss a pixel-based skin detection methods, that classify each pixel as skin or non-skin individually,
independently from its neighbors and combine it with Viola and Jones based face detection to improve the performance of face
detection systems in terms of increasing the face detection speed and decreasing false positive rate. The results as shown that the
proposed method efficiently increases face detection speed as well as decreases false positive rate.
Abhishek Maity et.al [5] have suggested a novel approach to face detection using image parsing and morphological analysis. The
main objective of the paper as mentioned by the authors is to propose an algorithm for extraction of some fundamental information
of an image efficiently and then finally use that to detect the human face on the image. The method which they propose is based on
assumption that the image contains the frontal face. They describe a method which widely-used edge-based features. Firstly, the
skin region is detected using a colour based algorithm and Brightness preserving histogram equalization techniques operated on
RGB to determine the presence of human and the head region is calculated. Lastly, the probable calculated head region is extracted
using Sobel Edge-Detection and matched with our trained database files for further classification of the segmented image.

© 2017, www.IJARIIT.com All Rights Reserved Page | 486

 Dwivedi Saloni, Gupta Nitika., International Journal of Advance Research, Ideas and Innovations in Technology

Sayantan Thakur et. Al [6] have proposed a face detection method using skin tone segmentation. They have proposed an algorithm
which ingeniously uses a novel skin color model, RGB-HS-CbCr for the detection of human faces. Skin regions are extracted using
a set of bounding rules based on the skin color distribution obtained from a training set. The segmented face regions are further
classified using a parallel combination of simple morphological operations. Experimental results on a large photo data set have
demonstrated that the proposed model is able to achieve good detection success rates for near-frontal faces of varying orientations,
skin color and background environment. The model proposed by the authors utilizes the additional hue and chrominance information
of the image on top of standard RGB properties to improve the discrimination between skin pixels and non-skin pixels.Skin regions
are classified using the RGB boundary rules and also additional new rules for the HS and CbCr subspaces. These rules are
constructed based on the skin color distribution obtained from the training images. The classification of the extracted regions is
further refined using a parallel combination of morphological operations.

III. FACE DETECTION

Early activity in the face detection has passed over as promptly as the start of the 1970s, where the basic heuristic and anthropometric
systems were used. These systems are generally rigid due to different assumptions, for example, frontal face, plain foundation, a
common visa photo situation. For all these frameworks, any change in the picture conditions ability means a fine-tuning, if not the
complete overhaul. Regardless of all these issues, the development of an exploration investment stayed continuous until the 1990s,
when a handy and genuine face differences and feature coding frameworks began to change into an actuality.

Figure 2: Detection System Overview

The objective of the face detection is to find out the existence of any appearances in the picture and, if it is present, gives back where
it’s owed and the degree of the each one face. While the face detection is an inconsequential work for the human vision, it is a test
for the machine vision due to the varieties in scale, area, facial articulation, light condition, introduction, posture and other different
appearance characteristics.
There exist two different types of face detection problems:

1) Face detection in images - Most of the face detection frameworks endeavors to focus a small quantity of an entire face, by
allocating with the majority of the foundation and various zones of a singular's head, for example, hair that is not important for the
face identification. With static pictures, this is regularly processed by executing a sliding "window" over a picture. The face detection
scheme then searches if a face is present inside the window. Regrets, with static pictures there is a large hunt space of the conceivable
areas of a face in the picture. The Faces can be expensive and be placed anywhere from upper left to an easier right of the picture.
2) Real-time face detection – The Continuous-face recognition involves the analysis of the face from an arrangement of casings
from any feature-catching gadget. The Real-time face detection is really a bit more straight forward process than recognizing a face
in the static pictures. It is on the findings that not at all like a huge portion of our nature, an individuals can dependably keep moving.
Applications of Face Detection
 Facial recognition- The Face detection is used within the biometrics, like as a piece of or together with a facial
identification framework. Similarly, it is utilized as a part of the human machine interface, the video surveillance and an
image database management.
 Marketing- The Face detection is selecting the enthusiasm of all the advertisers. A webcam power to be coordinated into
a TV and to identify any face that strolls by. At that point the groundwork calculates the sexual orientation, race and age
extent of face.
 Photography- Some recent digital cameras applies the face detection technique for the autofocus. The Face detection is
very useful for selecting regions of an interest in photo slideshows that can utilize a pan-and-scale Ken Burns effect.
 Smart captcha- It is the mixture of an effectively existing captcha which uses sound and realistic pictures.

IV. VIOLA JONES FACE DETECTION METHOD

© 2017, www.IJARIIT.com All Rights Reserved Page | 487

 Dwivedi Saloni, Gupta Nitika., International Journal of Advance Research, Ideas and Innovations in Technology

The basic rule of the Viola-Jones calculation is to sweep a sub-window to fit for discovering faces over a given info picture. The
fundamental picture transforming the methodology capable of rescaling the information picture to various sizes and after that to run
the changed size locator through these pictures. This technique ends up being monotonous because of the picking of the diverse size
pictures. In opposition to the standard techniques, Viola-Jones rescaled the detector quieter than the data picture and execute the
locator commonly terminate the picture each time with an alternate size. This locator is built by utilizing a supposed fundamental
picture and some of the basic rectangular properties reminiscent of the Haar wavelets. The Viola-Jones face indicator has three
recognized key commitments, which precise high handling rate and the discovery rates. These important commitments are the basic
picture, an efficient taking in computation-focused around the Adaboost, and a course structure.
The scale invariant detector
The first step in the Viola-Jones face detection algorithm is to change the input image into an integral image. This is accomplished
by making each pixel equivalent to the entire sum of all the pixels above and to the left of the concerned pixel. This illustrated in
Figure 2.

Figure 3: Integral Image

This permits for the calculation of the sum of all the pixels inside any given rectangle utilizing only four values. These four values
are the pixels in an integral image that coexist with the corners of a rectangle in an input image.

V EIGEN FACES
This method is based on the eigen faces technique in which the Principal Component Analysis (PCA) is used. Principal Component
analysis was developed by Karl Pearson in PCA is a statistical method which is used to reduce the high dimensional data space to
the low dimensional characterized space.This method is best suited for data compression and removal of redundancy.It is the most
successful technique that is prevalent in image recognition and compression.
The foundation of using Eigen faces in face recognition is based on the fact that each image can be represented as a matrix. A matrix
has a set of eigenvectors that represents the principal components of the matrix. Eigenfaces are the eigenvectors of the covariance
matrix of all faces. Similar faces can be described in a space with lower dimensionality. In mathematical terms, Eigenfaces are the
principal components that divide the face into feature vectors (or we can say that basic approach of using PCA in a face recognition
is to put across the large 1-D vector of pixels built from the 2-D facial image into its discrete components of the feature space. This
can be said to be as an Eigen Face projection. A covariance matrix gives us the information about these feature vectors. These
eigenvectors are the basis for measurement of variation among several faces. The faces are described by a linear combination of
highest Eigenvalues. Each face can be considered as a linear combination of the eigen faces. Then the face can be estimated by
using the eigenvectors occupying the largest eigenvalues. The best N eigen faces define an N-dimensional space, known as the “face
space”. Principal Component Analysis has been employed by L. Sirovich & M. Kirby (1987) to efficiently represent pictures of
faces. They defined that face images could be approximately reconstructed using a small collection of weights for each face and a
standard face picture. The weights describing each face are obtained by projecting the face image onto the eigen picture. Eigenface
also found its application in the research work of M.A Turk & Alex Pentland who produced a comparatively more efficient technique
in the face recognition system by using PCA.

VI. PROPOSED ALGORITHM

The below algorithm has been incorporated in the research work:
(a)Given image is in the form of (𝑋1 , 𝑌1 )…..( 𝑋𝑛 , 𝑌𝑛 )
(b) yi=0,1….for positive and negative value.
(c) initialize the weight w1,i=−1⁄2𝑚 , for yi=0,1 respectively.(where m,n are positive and negative integer respectively.
1
2𝑙
(d) for t= 1,…….T.
(1) normalize the weight wt..
𝑤𝑡,𝑖
Wt,i= ∑𝑛 (1)
𝑗 =1 𝑤𝑡,𝑖
(where wt is probability distribution).
(2)For each feature j, train a classifier hj which is restricted to use a single feature.
The error is evaluated with respect to
𝑤𝑡,𝑒𝑡 =∑I 𝑤𝑖 /hj (xi) ,yi (2)
(3) Choose the classifier ht with lowest erroe rate 𝑒𝑡 .
(4)Update the weight

© 2017, www.IJARIIT.com All Rights Reserved Page | 488

 Dwivedi Saloni, Gupta Nitika., International Journal of Advance Research, Ideas and Innovations in Technology

W t+1,i= wt,𝑖 𝛽 𝑡 1−𝑒𝑖 (3)

Where ei =0 if example 𝑥𝑖 is classified correctly.
𝑒𝑖=1 Other wise
And
𝑒
𝛽𝑡 = 𝑡⁄1 − 𝑒 (4)
𝑡
FINAL STORAGE CLASSIFIER:
h(ϰ)=1 ∑𝑡 𝑇 =1 𝑎𝑡 ℎ𝑡(𝑥)≥1/2 ∑𝑡 𝑇 =1 𝑎𝑡 (5)
1
where 𝑎𝑡= log
𝛽𝑡

The system consists of real-time face detection and recognition implementation. The various phases of the system implementation
have been discussed below:
Face Detection Phase:
Step 1: Read the input image
Step2: Read Frontal Face Object
Step 3: Obtain face co-ordinates
Step 4: Use coordinates to locate face on complex background using Viola Jones
Step5: Detect face

Training phase
Step 1: Crop face for detection
Step 2: Make bounding box of 240 x 320 pixels
Step 3: Grayscale conversion reduce the dataset.
Step 4: Apply Image Enhancement to enhance the quality of obtained Image by adjusting intensity value of pixels.
Step 5: Create Eigen faces from this image
Computing the average face
image m = (1/P)*sum(Tj's) (6)
for i= 1 to Number of rows
Computing the difference image for each image in the training set
A i= Ti – m (7)
temp = double(T(i) – m (8)
Merging all centered images
A = [A temp] (9)
end
Obtain covariance matrix
L = A'*A (10)
where A is center matrix Compute eigenvectors of L matrix
For i=1:eigen vectors
if( D(i,i)>1 ) , where D is the eigen vector of L matrix.
Eigen L=[Eigen_N,Eigen_values] (11)
end
eigen_ face= A * Eigen L (12)
Step 4: Form the database of all training sets.
RECOGNITION PHASE:
Step 1: Get the image to be recognized i.e. test image
Step 2: Apply Eigen faces algorithm on the test image
Step 3: Calculate Euclidean Distance from all eigen faces in database
Step 4: Minimum Euclidean Distance data set is recognized.

VII SIMULATION RESULTS AND ANALYSIS

The algorithm discussed above is implemented in MATLAB 2013a, using various available functions and mathematical
techniques. The below discussion shows a step by step implementation of the algorithm.

© 2017, www.IJARIIT.com All Rights Reserved Page | 489

 Dwivedi Saloni, Gupta Nitika., International Journal of Advance Research, Ideas and Innovations in Technology

Figure 4: Face Detection

Figure 5: Face Area Detected, cropped and converted to grayscale

Figure 6: Enhanced Image

Figure 7: Training Dataset of 10 Images (Person 1)

© 2017, www.IJARIIT.com All Rights Reserved Page | 490

 Dwivedi Saloni, Gupta Nitika., International Journal of Advance Research, Ideas and Innovations in Technology

Figure 8: Training Dataset of 10 Images (Person 2)

Figure 9: Recognised Face

The face detection and recognition were done with a number of faces. The following table shows the recognition rate of different
faces:
Table I: Recognition rate
Person No. of correct No. of Percentage
Recognition Incorrect
Recognition
Person 1 26 4 86.6

Person 2 25 5 83.33

Person 3 27 3 90.00

Person 4 26 4 86.6

CONCLUSION
This paper presents a real-time face detection and recognition algorithm implementation and its subsequent results on a number of
test datasets of live images. The detection is performed using Viola-Jones Algorithm and the recognition phase uses eigen faces and
Euclidean faces method. Image enhancement after detection of a face region from complex background has been used to improve
the quality of the recognition system. The system performs satisfactorily well with a number of complex backgrounds tested in a
number of illumination conditions.

REFERENCES
[1] Faizan Ahmad, Aaima Najam and Zeeshan Ahmed,“Image-based Face Detection and Recognition:“State of the Art””, IJCSI
International Journal of Computer Science Issues, Vol. 9, Issue 6, No 1, November 2012.
[2] Amr El Maghraby, Mahmoud Abdalla, Othman Enany and Mohamed Y. El Nahas, “Hybrid Face Detection System using
Combination of Viola - Jones Method and Skin Detection”, International Journal of Computer Applications (0975 – 8887) Volume
71– No.6, May 2013.
[3] Ajit Danti, K.M. Poornima, Narasimhamurthy, “Detection of Multiple Faces in Color Images using Haar Wavelets”,
International Conference on VLSI, Communication & Instrumentation (ICVCI), Proceedings published by International Journal of
Computer Applications (IJCA), 6-11,2011.
[4] Zahra Sadri Tabatabaie1, Rahmita Wirza Rahmat, Nur Izura Binti Udzir and Esmaeil Kheirkhah,”A Hybrid Face Detection
System using a combination of Appearance-based and Feature-based methods”, IJCSNS International Journal of Computer Science
and Network Security, VOL.9 No.5, May 2009.

© 2017, www.IJARIIT.com All Rights Reserved Page | 491

 Dwivedi Saloni, Gupta Nitika., International Journal of Advance Research, Ideas and Innovations in Technology

[5] Abhishek Maity, Sayan Dasgupta, Debjit Paul,”A Novel Approach to Face Detection using Image Parsing and Morphological
Analysis”, International Journal of Computer Trends and Technology (IJCTT) – volume 23 Number 4–May 2015.
[6] Thakur, S.; Paul, S.; Mondal, A.; Das, S.; Abraham, A., "Face detection using skin tone segmentation," Information and
Communication Technologies (WICT), 2011 World Congress on, vol., no., pp.53,60, 11-14 Dec. 2011 doi:
10.1109/WICT.2011.6141217
[7] Hatem, H., Beiji, Z., Majeed, R., Lutf, M. and Waleed, J., 2015. Face Detection and Pose Estimation Based on Evaluating Facial
Feature Selection. International Journal of Hybrid Information Technology, 8(2), pp.109-120.
[8] Bakshi, U. and Singhal, R., 2014. A survey on face detection methods and feature extraction techniques of face recognition.
International Journal of Emerging Trends & Technology in Computer Science (IJETTCS), 3(3), pp.233-237. [9] Rath, S.K. and
Rautaray, S.S., 2014. A Survey on Face Detection and Recognition Techniques in Different Application Domain. International
Journal of Modern Education and Computer Science, 6(8), pp.34-44.
[10] Fernandes, S. and Bala, J., 2013. Performance Analysis of PCA-based and LDA-based Algorithms for Face Recognition.
International Journal of Signal Processing Systems, 1(1), pp.1-6.
[11] Gayathri, S., Mary Jeya Priya, R., and Dr.Valarmathy, S., 2014. Face Recognition by Using Distance Classifier Based on PCA
and LDA, International Journal of Innovative Research in Science, Engineering, and Technology, 3(3), pp.1121-1126.
[12] Kamerikar, Umesh Ashok, and Dr. M.S. Chavan. 2014. Experimental Assessment of LDA and KLDA for Face Recognition,
International Journal of Advance Research in Computer Science and Management Studies, 2(2), pp.137-146.
[13] Gawande, M.P., and Agrawal, D.G., Face recognition using PCA and different distance classifiers. IOSR Journal of Electronics
and Communication Engineering, 9(1), pp.1-5.
[14] Barnouti, Nawaf Hazim. 2016. Improve Face Recognition Rate Using Different Image Pre-Processing Techniques, American
Journal of Engineering Research (AJER), 5(4), pp.46-53.

© 2017, www.IJARIIT.com All Rights Reserved Page | 492