This document summarizes a student project on face recognition. It begins with an introduction to face recognition, its applications, and common challenges. It then reviews literature on existing face recognition methods and identifies problems related to tilted poses and variations in illumination and expression. The proposed method will work to improve recognition rates under these conditions in two phases - training and testing. The method aims to enhance the preprocessing and feature extraction steps to make the system more robust. A basic flowchart of the proposed approach is provided, and the document concludes with references.

1. FACE RECOGNITION
Given By:
Satyendra Rajput(132112205)
Guided by:
Dr. Jyoti Bharti

2. CONTENT
Introduction
Literature survey/ Review
Problem Identification
Proposed method
Conclusion
References

3. INTRODUCTION
Identification of person using face.
It is one of the most user friendly and frequently
used methods in biometrics.
 Face image contain rich information.
 Facial feature point(eyes, nose, mouth)
 Facial features make human recognition an
automated process.
It is very easy and simple process for recognition.

4. INTRODUCTION
 Face recognition is an active research area with a wide
range of applications in the real world.
Information security,
Authentication,
Access control,
Law enforcement surveillance[8], Etc.
 Most face recognition algorithms are designed to work
best well aligned, well illuminated, and fontal pose face
image[5].
 Small variation of face size and orientation can be
effected the result.

5. LITERATURE SURVEY
 Basically we referred Eight research paper.
 All these paper are working on image processing
field, pattern recognition, signal processing and
experts system.
 During a survey we mainly focus on face pose
recognition problem.

6. BASIC STEP FOR FR
 An automatic face recognition system are mainly
comprised of three steps.[3]
Detection
Feature
Extraction
Face
recognition
Input image(image
sequence)
Result Output
Basic flowchart of a face recognition

7. FACE REGION DETECTION AND LOCALIZATION
 Detection may include face edge detection, segmentation
and localization, namely obtaining a pre-processed
intensity face image from an input scene, either simple or
cluttered, locating its position and segmenting the image
out of the background[3].
 Face detection may fall into two categories.
(1) Local feature based ones
(2) Global methods
o Their Face detection region are required by the
comparative matching between detecting region and
constructed template based on modeling.

8. APPROACHES BASED ON FEATURE EXTRACTION
 Geometrical method
 Color based or texture based method[3]
 Motion based method
 Holistic based method: feature derived from the
hole images. It generate a general template for
whole image.
 Eigen face base method
 Feature based: LGS,LBP, EGBM(elastic bunch
Graph matching)[8].
 Other Hybrid method:
combine both local & global feature to produce a
more complete facial representation.[3]

9.  Eigen face based method [2] :-
system decomposes an entire input image into
subband images which contain discriminant feature.
Multiple sliding windows within different subbands
are aligned to the same spatial location. Feature
are selected and calculate likelihood ratios. Ratios
exceed a fixed threshold than face location is
reported.

10. SPATIAL MATCHING DETECTOR METHOD [2]
 This approach embraces SVM, various template
matching methods, other discriminable kernel cost
function methods.

11. NEURAL NETWORKS METHOD [2]
 Use of three layers of weights allows to evaluate
the distance between an input image and the set of
face image.

12. FACE COMPONENT EXTRACTION USING
SEGMENTATION METHOD ON FRS[4]
 Face Components extraction process
1. Face skin model detection.(pre processing)
2. Face detection process on normal still image.
3. Face cropping process on normal static image.
4. Extraction process and Measurement of distances
between face components.
 Dataset: 150 subject images(local)
 Result : more face component produces very
good accuracy.
Method properly on frontal single human face with
relatively different lighting condition.

13. GENETIC BASED LBP FEATURE EXTRACTION AND
SELECTION FOR FACIAL RECOGNITION[1]
 Feature extraction using LBP method.
 Initially image is segment into a number of uniform
evenly distributed patches that cover entire image
Gray Scale image Segmented

14. CONT...
 Step wise diagram of face component extraction

15. CONT...
 Matrix value find using equation.
 Pattern Matric:
Computing LBP
String LBP value
01100110  102 (decimal)

16. CONT...
 Feature vector or template is a concatenation of all the
histograms corresponding to the patches on an image.
 Recognition is performed by comparing a captured
probe template p, with all the vector in a gallery set H
={h0, h1 ,.....hq-1 } using Manhattan distance metric. If
subject hj from the gallery set that is closest to p is
considered to be its match.
 SSGA is used to evolve a population of candidate
feature extractors.
 Candidate FE fei is a 6 tuple <Xi, Yi ,Wi ,Hi ,Mi ,Fi>
 Result: No. of Patches required for recognition is less
& accuracy enhance over SLBPM.
 Dataset: ERCG dataset(105 subject per 3 image)

17. FACE RECOGNITION WITH SLGS[8]
 Each pixel is represented with a graph Structure of
its neighbor pixels.
 Histogram of SLGS were used for recognition by
using NN classifiers that include Euclidean
distance, correlation coefficient and chi square
distance measure.
 Dataset : AT&T and Yale face DB
 Result: Improve recognition rate over LBP and
LGS.
SLGS is robust to variation in term of facial
expression, facial details and illumination.

18. CONT...
 Relationship in symmetric
structure consist of same
number of neighbor pixels
on both sides.
 Drawback: This approach may produce low
performance for rotated face image.

19. FACE RECOGNITION WITH LBP, SPATIAL PYRAMID
HISTOGRAMS AND NAIVE BAYES NN
CLASSIFICATION[5]
 Pre-Processing :applying Tan & Triggs’ illumination
normalization algorithm.
 LBP operator: LBP are computed for each pixel,
create a fine scale textual description of the image.
 Local feature extraction: Local features are create
by computing histograms of LBP over local image
regions
 Classification: Each face image in test set is
classified by comparing it against the face images
in the training set. Comparison is perform using
local features obtained in the previous step.

20. CONT...
 Dataset: AT&T-ORL, Yale, Georgia tech and extended
yale B.
 Evolution methodology:
 Algorithm parameter: regions size 8X8.
 Result: NBNN give better accuracy over different classifier
and holistic algorithm.
 Future work: NBNN increase computational cost relative
to original LBP based algorithm.
 Replace LBP histogram descriptors with other local
descriptors.
 Find a better alternative to the grid based regions. Grid
partition has no natural relation to shape of face.

21. MOTIVATION
 Small variation of face size and orientation can
effect the result.
 Many algorithm still not work efficiently variation in
pose, illumination and facial expression of
image[8].

22. PROBLEM IDENTIFICATION
1) Face recognition on Tilted face pose.
2) Improve the recognition rate under illumination and
facial expression of image [7].

23. PROPOSED METHOD
 FR method work in 2 phase
Training & Testing phase.
 May be I work on pre-
processing phase to make
system more robust on
illumination & noise.
 I would work on FE phase to
improve the accuracy.
Framework for Face recognition

24. FLOWCHART OF PROPOSED METHOD
 This are the basic step we are follow
to recognition the face.

25. CONCLUSION
 Using proposed method we improve the face
recognition system under illumination variation
and non-frontal view.
 Proposed approach is very simple in term of
calculation.
 Improve Speed of recognition.
 It required only one scanning without any need
to a complicated analysis.

26. REFERENCES
[1] Joseph shelton and Gerry Dozier(2011) “Genetic based
LBP feature Extraction and selection for facial
Recognition”.
[2] Rahimeh Rouhi, Mehrari and Behzad(2012) “Review on
feature extraction techniques in face recognition.”
[3] Yongzhong Lu, Jingli and Shengsheng “A survey of
face detection, extraction and recognition.”
[4] Dewi agushinta and Adang(2010-11) “Face component
extraction using segmentation method on face
recognition system.”
[5] Daniel and Domigo(2011) “face recgnition with LBP,
Spatial Pyramid Histograms and Naïve bayes Nearest
Neighbor classification.”
[6] Poonam Sharma, KV Arya(2013) “Efficient FR using
wavelet based generalized neural network.”

27. REFERENCES
[7] Bhumika G bhat, Zankhana H shan(2011)
“face feature extraction techniques: A survey”.
[8] Mohd Filkri abdullah, Md shohel Sayeed(2014)
“face recognition with sysmetric LGS”.

28. Thanks