Face Recognition for Partial Occluded Image: A Survey
1
Priya Matte, 2Rohit Raja, 3Yamini Chouhan
Department of Computer Science,
Shri Shakracharya Technical Campus, 490009
1
2
3
priya.mate@gmail.com
id-rohitraja4u@gmail.com
id-yaminichouhancs@gmail.com
Abstract
Over the years biometrics has gained unparalleled popularity in digital medium and has proven
its usefulness for several applications concerned with the threats and crime or security purposes.
Face Recognition is a widely emerging biometric for automating the surveillance, as it has aid in
strengthening the security from several types of terrorist or criminal threats. Though, there are
several face recognition techniques which are categorized based on its error rates in recognition
but there are few that gives the marginal rate for sufficient and validated recognition rates for
occlude images. This survey illustrates the precise overview of the major face recognition
techniques which paves strong foothold for the partially occlude images.
Keywords: Face Recognition methods, feature extraction, partial occlusion, error in recognition
rates.
1. Introduction
Occlusion in context of facial imaging refers to the obstruction exhibit by wearable glasses,
scarfs, hairs, or alternative accessories. Efforts created on face recognition below appropriate
condition has already been achieved in past studies however to date the issues like illumination
field, facial features, cause estimation & partial occlusion has been the present concern of
analysis for effective face recognition below variable condition. we've got classified the literature
in awe of an equivalent issue and so enlist the few of the effective makes an attempt that are
created to resolve the difficulty. This ways like half primarily based ways that includes of ways
like Principal element Analysis (PCA), Linear Discriminate Analysis (LDA), Non-negative
Matrix resolving (NMF), native Non-Negative Matrix resolving (LNMF), freelance element
Analysis (ICA) and therefore the alternative variations of it's shown higher recognition rates.
Also, alternative ways like feature primarily based technique or form technique sometimes
considers the options round the essential facial elements like that of eyes, ears, or mouth region
is incorporated within the recognition section of this algorithm; tho' such ways show numerous
set of research results. Occlusion is split into 2 classes i.e., natural occlusion and artificial
occlusion. Natural occlusion refers to the non-intentional blockade of read of the objects whereas
the artificial occlusion to the intently sitting of artificial obstruction between the views. Partial
occlusion has generally severely have an effect on the image process of statistics because it tends
to disrupt the identification of the actual image. For example: in cases of retinal statistics
eyelashes occluded the iris, earrings may ugly the ear lobe and thereby causing the error rates
within the biometric recognition method [1]. Now, once it involves the face recognition there
square measure many accessories which may be advisedly or accidentally wont to obstruct the
pictures [2].
Now, there are important rise in reports wherever the crime committer has exhibit disguised that
it fails the digital security system to observe and stop the threat at preliminary stages itself or
maybe in things wherever folks amendment their look to trick the face recognition systems. This
face recognition systems unsuccessful fully during this criteria to beat such difficulties [3]. tho'
the face recognition technology has achieved varied feats however the issues of illumination,
cause estimation, partial occlusion has stay however to completely resolved with high accuracy
in recognition rates [4-9]. Hence, there square measure few ways adopted within the past studies
to beat matters [10-15]. ways like Principal element Analysis (PCA) [16], Linear Discriminate
Analysis (LDA) [17], neural networks [18] and several other variations with the combo of them
had been adopted to beat this things however none has proved its effectiveness for the resolution
of this conditional situations like non-linear cases as every has its own limitation of operations.
There square measure alternative ways by the likes Kernelmachine- primarily based
Discriminate Analysis (KDA) [19], neural networks [20], versatile Discriminate Analysis (FDA)
[21], and Generalized Discriminate Analysis (GDA) [22], that square measure appropriate for
non-linear cases however haven’t mature in application due to its high machine price of
operation for each coaching & testing of datasets.
The face recognition technique primarily depends on 2 feature sets i.e., world and native options
[23,24]. This options square measure simply influenced by occlusions or noises of bound sorts
and thereby degrading the lustiness of such ways. There square measure studies that reports the
high effectiveness of native options once utilized for face recognition [25-29]; however at an
equivalent time this options square measure principally at risk of the affects by partial occlusion.
However, if used understandably such ways show exceptional results. makes an attempt created
by Martinez that employs his strong recognition system by merging the native options supported
similarities [30]; or of the makes an attempt on Support Vector Machines (SVM) classified the
native options in to teams that deals with the options of facial elements like eyes, ear, nose to
determine a geometrical interpretation of corresponding feature sets. [31-33]. Now, there square
measure alternative classes of face recognition that focuses on a lot of of the holistic options of
facial regions i.e., wishing on world options. Thus, within the following section we have a
tendency to presents the review of identifiable ways that cites higher ends up in performance
take a look at specifically for part obstruct pictures.
2. Face Recognition Techniques for Partial Occlusion
The primary strategies until currently for partitioning the difficulty of partial occlusion is
split into 3 categories namely: Feature primarily based strategies, half {based|based
mostly|primarily primarily based} strategies and shape based strategies. (As shown in
Figure 1).
.
Figure 1. Enlisted techniques for face recognition in occlude cases.
2.1.Part Based Methods
The primary ways until currently for resolution the difficulty of partial occlusion is split into 3
categories namely: Feature primarily based ways, half {based|based mostly|primarily primarily
based} ways and FrThe half based methodology contains of many bits of method that involves
the division of facial image into many overlapping & non-overlapping pel sets that ar then
used for the matching task [34]. Tan et al. had bestowed an idea of eliminating the non-matching
details from the non-parametric partial similarity between 2 sets of facial images; that helps in
extracting social details [35]. Here, the take a look at and coaching image separated into blocks
so stick-out to make a self-organizing map (SOM); this maps within the partial neighboring
distances and after helps out mechanically finding the foremost similar components between the
2 pictures. This methodology is additionally named as non-metric distance live because of the
actual fact that doesn’t satisfies the self-similarity of alternative metric properties like triangular
difference. Thus, distance live involve within the methodology is compared with alternative
distance measures, for example: Simplified Mahalanobis (SM), changed square geometer
distance (SE), Weighted Angle-based distance (WA) and Angle primarily based Distance
between colourless feature vectors [36]. The experiment had administrated over the AR face info
and ORL info [37]; on each info the popularity rates (RR) of ninety seven & seventy
four.6% severally had reportable. In another work by Kim et al. 2D-PCA methodology were
wont to handle the matter of partial occlusion [38]. This methodology depends over K-NN and 1NN classifiers that allows it to eliminate the jam components and check for partial similarity.
thenceforth many tried together with supervised & unattended coaching had administrated
with this classifiers. the tactic shows some approximate recognition potency of ninety eight and
ninety nine for eyeglasses and scarf severally. Jia & Martinez employs SVM as a classifiers
that was changed for missing options within the jam region of facial imagination [39]. This
experiment is conducted over AR face info & Face Recognition Grand Challenge (FRGC)
datasets [40].
For facial pictures with varied facial features and part jam pictures a probabilistic methodology
was tried [41]. For this the computing region of facial pictures is divided in to k-regions so every
region is analyzed exploitation Mahalanobis distance; that orderly provides a performance vary
of 75-88% recognition results. Similarly, native Gabor Binary Pattern (LGBP) is employed
together with Kullback-Leibler Divergence (KLD) to live the likelihood for an equivalent case
[42]. The LGBP methodology forms the multi-scaled pictures followed by merging of native part
to make up one world feature vector from the bar graph of every of its elements. a really similar
effort by Wang et al. & Meng et al. uses world half detectors by facilitate of adjusted
Gradients (HOG) and native Binary Pattern (LBP) and develop Gabor feature primarily based
strong illustration and Classification (GRRC) system [71-78].
Latter, native Salient (LS-ICA) technique was introduced that employs ICA design so as to work
out the native basis of facial image [43]. there have been instances wherever the tactic was
compared with native Non-Negative Matrix factoring (LNMF) & native feature Analysis
(LFA). In his methodology the regions of facial imagination with special options like eyes, nose,
lips etc., ar maintained and latter aligned in orderly fashion for the popularity method. There was
another methodology that uses Posterior Union Model (PUM) and also the live of similarity is
decided among category|the category} containing samples per class [44]. they need conjointly
tried overcoming the problem of huge feature vectors from single coaching image. the popularity
sampling performed over XM2VTS & AR databases provides the performance of ninety
eight.8% and 91.5% severally. Thereupon, a Probabilistic call primarily based Network
(PDBNN) approach for jam faces together with PUM is introduced; wherever the most focus is
on increasing the strength of the system by accentuation on the matched options [45]. This
methodology is tested on XM2VTS & ORL face datasets with the performance of eighty
four.4% and 83.5% of recognition rates, severally. In a trial the feature vector enthusiastic about
the sets of similarities ar fashioned on the take a look at & probe pictures [46]. Followed by
it the LDA technique is used as a classifier for separating distinguished categories with ninety
seven.41% of recognition rates over FERET info. There also are few ways that follows learning
primarily based approach exploitation neural network or fuzzy scaling as feature classifiers.
Among such ways most notable is that the Auto-Associative Neural Network for face
reconstruction; here the classifier used has 3 layer multilayer perceptron to evoke original face
from distorted ones [47]. Experiments carried over AR datasets with ninety three pictures for pel
wise coaching of size love eighteen x twenty five pixels.
Lately, 3 variants of PCA ways has evidenced its effectiveness against recognizing jam pictures
[48]. The ways ar as follows:actal primarily based ways. (As shown in Figure 1).
(1) Component primarily based PCA (CPCA), whereby totally different elements of the facial
expression square measure handled individually.
(2) Holistic methodology, wherever the whole facial region is reborn into sub-spaces [49,50].
during this methodology the topological space learning square measure enforced or weighted
factors square measure allotted to distinct regional areas over the facial imaging. tho' the formula
will perform effectively once solely the two hundredth of the face is occluded. Since, it's
supported reconstruction of the image by accounting the distinction of arrangement from the
reconstructed image
(3) Lophoscopic PCA (LPCA), wherever solely those regions square measure individually
known wherever the probabilities of occlusion square measure doubtless to occur [51]. tho' it
increase the performance however its major downside is that the high process value.
Following such tasks a fusion mechanism is employed to seek out the choice sets of spate
categories of every image belongs to; this ways offers seventy two, sixty nine and forty seventh
of recognition rates for the on top of mentioned AR information. There comparison with
different ways particularly ICA, LNMF, NMF and Neural Associators (NA) was studied in [52].
The studies has additionally been tried towards the handling of expressions supported LDA
& PCA methods; followed by social control of gray intensity levels through mean variance
and bar graph leveling to eliminate the results of illumination and partial occlusion [53]. all told
of this ways there was associate degree assumption taken prior to before experimenting that the
given image is occluded and also the reference image is accessible with the probe image being
regionally occluded below the condition of partial similarity [54, 55]. For Selective native NonNegative Matrix factoring (S-LNMF) based mostly methodology depends over divided PCA and
one NN classifier used to discover distortions, wherever LNMF was used for comparison over
identical AR face recognition datasets cropped to sixty four x eighty eight constituent sets [56].
A face recognition methodology in PCA class for addressing the illumination downside
employing a localized non-linear feature choice methodology is claimed to be sturdy in varied
light-weight coefficient of reflection fields and partial occlusions [57]. For this methodology
kernel-eignspace is employed for feature extraction with merging of neighboring sub-parts in
part congruency with relevance the reference image. This accounts for associate degree overall
performance of ninety nine.5% on AR datasets
.
2.2.Fractal Based Methods
Partitioned Iterated operate System (PIFS) based mostly face recognition conferred
associate degree algorithmic rule computing self-similarities between the given image
subspaces and therefore the at the same time establishing the link between many square
grid regions of the facial imagination (as shown in fig a pair of below) [58]. owing to its
sensitiveness towards occlusions it's handled regionally handled regionally and thereby
more distortions area unit} eliminated through ad-hoc distance measure.
Figure.2: Range block for major detected subfractal areas (eyes, nose, mouth, etc) for an
arbitrary image & that of with rotation.
There area unit solely a couple of major work tried with this kind of methodology. The
experiments carried over AR face information show dissimilar results [59]. during this style of
methodology the image is preprocessed and size to possess the fastened dimension whereas
guaranteeing the eyes and mouth area unit centrical with relevance the time; thenceforth the
amount of sub areas area unit accessed iteratively. Different similar ways area unit made public
in [60].
2.3.Feature Based Methods
2.4.
Feature based mostly strategies takes thought of individual options like eyes, mouth, eyebrows,
nose etc., in account for the popularity method and ignores the opposite options [61]. Since, the
likelihood is measured to search out the proper image; consequently it deals with inexact location
and partial occlusions [62]. during this variety of strategies the image is split into distinct native
components and area unit then analyzed in isolation with the eignspaces. this permits the training
of alignment of subspaces within the controlled lighting surroundings or occlusion. The
experiment is carried over the A facial datasets for a hundred grayscale pictures akin to fifty men
& fifty ladies of size a hundred and twenty x one hundred seventy pixels; giving the
popularity rate of a variety around 85-95% & ~80% for scarf occluded and specs occluded
pictures severally.
Figure.3: Feature division results on frontal and rotated imagery of facial regions.
Another similar tries area unit illustrated in [63]; whereby during this technique soft masking is
employed as a primary process tool to classify the outliner in occlusion pictures. Experiments
over Massachusetts Institute of Technology face recognition info provides the performance
estimate of ninety fifth. The graph primarily based techniques wherever facial mental imagery is
reconstructed mistreatment graph model with label learning approach may be a hot analysis topic
[64]. This technique uses distributed illustration & graphical model for recognition of
partly obstruct pictures. Notably, AN illustrative example of such ways is Active look Models
(AAM) that build a mesh-like structure over the facial mental imagery so as to mark visible
distinctive options of the facial region. within the past this technique is tried with adjustive fitting
strategy over many databases that provide the high performance vary for pictures 10-40% partly
obstruct [65]. This technique has AN favorable position over the antecedently cited ways
because the easy image are often obstruct by facial hair or hands so by distribution the adjustive
options with graph modeling may be a nice job during this direction.
In terms of ability this graph-mesh primarily based approach is enforced with alternative
adjustive formula like SVM (Support Vector Machine) together to influence the native distortion
options [66-68]. The advantage of this ways is that it satisfies Mercer’s theorem that alternative
have didn't. Since, the errors shaped by occlusion area unit usually distributed in nature ANd
didn’t have an effect on the pixels in an adversely. Thus, the performance vary of such ways area
unit effective. Since occlusion or errors area unit distributed in nature so that they don’t
sometimes have an effect on the pixels greatly. Performance on numerous databases shows that
formula executes effectively.
3. Discussion
After reviewing completely different principal ways within the literature of face recognition for
halfial block pictures it are often over that the foremost of the work is finished underneath part
primarily based} ways tho' the tries that ar supported feature based ways show guarantees to
bring higher leads to the close to future. The half based mostly ways typically converts the image
into topological spaces then attempt overlapping with the matching pictures for locating the
higher chance of matching with many subspace reduction technique [69]. Whereas those ways
that belongs to the feature primarily based} models embody Ada-Boost and SVM based
methodologies victimisation mathematician mixtures or eignspaces. Like already declared within
the on top of text none of the tactic handles all the matter effectively within the race of sturdy
face recognition system. the assorted ways of the researchers for partitioning the difficulty of
partial occlusion has been mentioned and therefore the said mentioned techniques ar summed up
within the table one below. Summarizing the general techniques has its execs & cons and
variant field of usage. Some schemes become complicated however offer higher results however
it will increase the hardness of computation and creating the system functionally complicated is
very undesirable inside the similar trade-off. Our effort of grouping all the literature inside one
document can let the analysisers to own a chance to determine their research accomplishments
thereupon of the additional needed to furnish within the literature inside this regard.
S.No. Methods Training
1
[35]
Testing
Recognition
Publication
Sets
Sets
Results (%)
Year
700 in AR
1900 in AR dataset 5
97.0% in AR and
2006
dataset 5
images/person in
74.6% in ORL
ORL dataset
dataset
100 sunglass images
96.00%
2007
N/A
88.9%
2009
50 randomly
remaining
75-85%
2002
chosen images
used for testing
images/ person
in ORL dataset
2
[38]
35 sunglass
and 100 scarf images
images and 35
for testing
scarf images of
20 men and 15
women
3
[39]
1200 images
from AR and
800 images from
FRGC
4
[41]
of persons in AR
dataset
5
[42]
50 randomly
N/A
75%
2007
chosen images
of persons in AR
dataset
6
[43]
200
600
88.3%
2005
7
[70]
100 persons
4 images / person. of
83.5%
2009
97.41 %.
2008
87.1%
2006
which one or two for
chosen
testing
randomly 4
images/ person.
of which one or
two for training
8
[45]
100 persons
remaining used
selected
of testing
randomly four
images for each
person used
9
[46]
1002 front view
FA has 1196 subjects
and the FB set has
face images
selected from
1195 subjects
training
10
[47]
93
N/A
72%
2003
11
[48]
10 images of 37
10 images of 37
~88-99% with
2008
individuals
individuals
100
300
different
12
[55]
variations
84-96% with
2005
certain variation
due to
expressions
13
[56]
270
200
85 - 95% and
2008
~80% for scarf &
sunglasses
4. Conclusion
This literature survey deals with numerous strategies that overcome the matter of hinder pictures
underneath uncontrolled conditions, thought there ar makes an {attempt|tries} that perform
effectively for controlled conditions however within the pursuit to create the system extremely
sturdy and adaptative is needed to create the attempt underneath the uncontrolled condition that
is far difficult than the previous. during this paper we've classified the face recognition
techniques into 3 elements. quantity} amount of labor has been done over the pattern based
mostly} approach whereas the feature based approach guarantees to eliminate many loopholes
within the partial occlusion imagination to provide higher ends up in the close to future.
However, it's conjointly contains of less quantity of contribution within the literature. Since, the
occlusion is related to corresponding distorted pixels and for it half based mostly technique has
evidenced to be more practical it appears that the sector is saturating among this region. 3d face
recognition is effective although it's higher procedure complexness that is undesirable for an easy
period of time application.
References
[1] Benjamin C., David B., Philip M., and Jitendra M., “A Real-Time Computer Vision System
for Vehicle Tracking and Traffic Surveillance,” Transportation Research Part C: Emerging
Technologies, vol. 6, no. 4, pp. 271-288, 1998.
[2] Sharif M., Sajjad M., Jawad J., Younas J., and Mudassar R., “Face Recognition for Disguised
Variations using Gabor Feature Extraction,” Australian Journal of Basic and Applied Sciences,
vol. 5, no. 6, pp. 1648-1656, 2011.
[3] Yang M., “Kernel Eigenfaces vs. Kernel Fisherfaces: Face Recognition using Kernel
Methods,” in Proceedings of the 5th IEEE International Conference on Automatic Face and
Gesture Recognition, USA, vol. 14, pp. 215-220, 2002.
[4] Chellappa R., Wilson C., and Sirohey S., “Human and Machine Recognition of Faces: A
Survey,” in Proceedings of IEEE, vol. 83, no. 5, pp. 705-740, 1995.
[5] Zhao W., Chellappa R., Phillips P., and Rosenfeld A., “Face Recognition: A Literature
Survey,” ACM Computing Surveys, vol. 35, no. 4, pp. 399-458, 2003.
[6] Zisheng L., Jun-ichi I., and Masahide K., “Block- Based Bag of Words for Robust Face
Recognition under Variant Conditions of Facial Expression, Illumination, and Partial
Occlusion,” IEICE Transactions on Fundamentals of Electronics, Communications and
Computer Sciences, vol. E94-A, no. 2, pp. 533-541, 2011.
[7] Zhiwei Z. and Qiang J., “Robust Pose Invariant Facial Feature Detection and Tracking in
Real- Time,” in Proceedings of the 18th International Conference on Pattern Recognition, pp.
1092- 1095, 2006.
[8] Zihan Z., Andrew W., Hossein M., John W., and Yi M., “Face Recognition with Contiguous
Occlusion Using Markov Random Fields,” in Proceedings of IEEE 12th International
Conference on Computer Vision, Kyoto, pp. 1050-1057, 2009.
[9] Sharif M., Sajjad M., Jawad J., Younas J., and Mudassar R., “Face Recognition for Disguised
Variations using Gabor Feature Extraction,” Australian Journal of Basic and Applied
Sciences, vol. 5, no. 6, pp. 1648-1656, 2011.
[10] Amirhosein N., Esam A., and Majid A., “Illumination Invariant Feature Extraction and
Mutual-Information-Based Local Matching for Face Recognition under Illumination Variation
and Occlusion,” Pattern Recognition, vol. 44, no. 10-11, pp. 2576-2587, 2011.
[11] Shermina J. and Vasudevan V., “Face Recognition System with Various Expression and
Occlusion Based on a Novel Block Matching Algorithm and PCA,” International Journal of
Computer Applications, vol. 38, no. 11, pp. 27- 34, 2012.
[12] Guoliang L., Kudo M., and Toyama J., “Robust Human Pose Estimation from Corrupted
Images with Partial Occlusions and Noise Pollutions,” in Proceedings of IEEE International
Conference on Granular Computing, Japan, pp. 433-438, 2011.
[13] McCloskey S., Langer M., and Siddiqi K., “Removal of Partial Occlusion from Single
Images,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 33, no. 3, pp.
647- 654, 2011.
[14] Struc V., Dobrisek S., and Pavesic N., “Confidence Weighted Subspace Projection
Techniques for Robust Face Recognition in the Presence of Partial Occlusions,” in Proceedings
of the 20th International Conference on Pattern Recognition, Istanbul, pp. 1334-1338, 2010.
[15] Aleix M., “Recognizing Imprecisely Localized, Partially Occluded, and Expression Variant
Faces from a Single Sample Per Class,” IEEE Transaction on Pattern Analysis and Machine
Intelligence, vol. 24, no. 6, pp. 748-763, 2002.
[16] Zhao H., Yuen P., and Kwok J., “A Novel Incremental Principal Component Analysis and
its Application for Face Recognition,” IEEE Transactions on Systems, Man, and Cybernetics,
Part B: Cybernetics, vol. 36, no. 4, pp. 873-886, 2006.
[17] Zhao H. and Yuen P., “Incremental Linear Discriminant Analysis for Face Recognition,”
IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics, vol. 38, no. 1,
pp. 210-221, 2008.
[18] Lawrence S., Giles C., Tsoi A., and Back A., “Face Recognition: A Convolutional NeuralNetwork Approach,” IEEE Transactions Neural Network, vol. 8, no. 1, pp. 98-113, 1997.
[19] Chen W., Yuen P., Huang J., and Dai D., “Kernel Machine-Based One-Parameter
Regularized Fisher Discriminant Method for Face Recognition,” IEEE Transactions on Systems,
Man, and Cybernetics, Part B: Cybernetics, vol. 35, no. 4, pp. 659-669, 2005.
[20] Yann A. Le Cun ,Patrice Y. Simard, John S. Denker, Bernard Victorri,” Transformation
Invariance in Pattern Recognition: Tangent Distance and Propagation”, Neural Networks: Tricks
of the Trade, Springer-Verlag, Berlin, pp 239–274, 1998.
[21] Hastie T., Tibshirani R., and Buja A., “Flexible Discriminant Analysis by Optimal Scoring,”
Journal of the American Statistical Association, vol. 89, no. 428, pp. 1255-1270, 1994.
[22] Baudat G. and Anouar F., “Generalized Discriminant Analysis using a Kernel Approach,”
Neural Computation, vol. 12, no. 10, pp. 2385- 2404, 2000.
[23] Guo G., Li S., and Chan K., “Face Recognition by Support Vector Machines,” in
Proceedings of the 4th IEEE International Conference on Automatic Face and Gesture
Recognition, Grenoble, pp. 196-201, 2000.
[24] Pontil M. and Verri A., “Support Vector Machines for 3D Object Recognition,” IEEE
Transactions on Pattern Analysis and Machine Intelligence, vol. 20, no. 6, pp. 637-646, 1998.
[25] Brunelli R. and Poggio T., “Face Recognition: Features Versus Template,” IEEE
Transactions on Pattern Analysis and Machine Intelligence, vol. 15, no. 10, pp. 1042-1052, 1993.
[26] Florent P., Jean-Luc D., and Kenneth R., “A Probabilistic Model of Face Mapping with
Local Transformations and its Application to Person Recognition,” IEEE Transactions on Pattern
Analysis and Machine Intelligence, vol. 27, no. 7, pp. 1157-1171, 2005.
[27] Hotta K., “A View-Invariant Face Detection Method Based on Local PCA Cells,” Journal of
Advanced Computational Intelligence and Intelligent Informatics, vol. 8, no. 2, pp. 130-139,
2004.
[28] Ying Weng, Aamer Mohamed, Jianmin Jiang and Stan Ipson,” Face Detection based Neural
Networks using Robust Skin Color Segmentation”, 2008 5th International Multi-Conference on
Systems, Signals and Devices,pp.1-5,2008.
[29] Schneiderman H. and Kanade T., “A Statistical Method for 3D Object Detection Applied to
Faces and Cars,” in Proceedings of the IEEE Computer Society Conference on Computer
Vision and Pattern Recognition, Pittsburgh, pp. 746-751, 2000.
[30] Martinez A., “Recognizing Imprecisely Localized, Partially Occluded and Expression
Variant Faces from a Single Sample Per Class,” IEEE Transactions on Pattern Analysis and
Machine Intelligence, vol. 24, no. 6, pp. 748-763, 2002.
[31] Cristianini N. and Shawe-Taylor J., An Introduction to Support Vector Machines,
Cambridge University Press, UK, 2000.
[32] Vapnik V., Statistical Learning Theory, John Wiley & Sons, 1998.
[33] Sharif M., Sajjad M., Mughees A., Younas J., and Mudassar R., “Using Nose Heuristics for
Efficient Face Recognition,” SINDH University Research Journal, vol. 43, no. 1-A, pp. 63-68,
2011.
[34] Sharif M., Sajjad M., Jawad J., and Mudassar R., “Illumination Normalization
Preprocessing for Face Recognition,” in Proceedings of the 2nd Conference on Environmental
Science and Information Application Technology, China, pp. 44-47, 2010.
[35] Tan X., Chen S., Zhou Z., and Liu J., “Learning Non-Metric Partial Similarity Based on
Maximal Margin Criterion,” in Proceedings of IEEE Computer Society Conference Computer
Vision and Pattern Recognition, New York, pp. 138- 145, 2006.
[36] Perlibakas V., “Distance Measures for PCABased Face Recognition,” Pattern Recognition
Letters, vol. 25, no. 6, pp. 711-724, 2004.
[37] Martinez A. and Benavente R., “The AR Face Database,” Technical Report, Computer
Vision Center, The Ohio State University, 1998.
[38] Kim T., Lee K., Lee S., and Yim C., “Occlusion Invariant Face Recognition using TwoDimensional PCA,” in Proceedings of International Conferences Advances in Computer
Graphics and Computer Vision, Berlin, pp. 305- 315, 2007.
[39] Jia H. and Martinez A., “Support Vector Machines in Face Recognition with Occlusions,”
in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Miami, pp.
136- 141, 2009.
[40] Phillips P., Flynn P., Scruggs T., Bowyer K., Chang J., Hoffman K., Marques J., Min J., and
Worek W., “Overview of the Face Recognition Grand Challenge,” in Proceedings of IEEE
Computer Society Conference on Computer Vision and Pattern Recognition, USA, vol. 1, pp.
947-954, 2005.
[41] Martinez A., “Recognizing Imprecisely Localized, Partially Occluded and Expression
Variant Faces from a Single Sample Per Class,” IEEE Transactions on Pattern Analysis and
Machine Intelligence, vol. 24, no. 6, pp. 748-763, 2002.
[42] Zhang W., Shan S., Chen X., and Gao W., “Local Gabor Binary Patterns Based on
Kullback- Leibler Divergence for Partially Occluded Face Recognition,” IEEE Signal Processing
Letters, vol. 14, no. 11, pp. 875-878, 2007.
[43] Kim J., Choi J., Yi J., and Turk M., “Effective Representation using ICA for Face
Recognition Robust to Local Distortion and Partial Occlusion,” IEEE Transactions on Pattern
Analysis and Machine Intelligence, vol. 27, no. 12, 2005.
[44] Sharif M., Sajjad M., Younas J., and Atif A., “Single Image Face Recognition using
Laplacian of Gaussian and Discrete Cosine Transforms,” International Arab Journal of
Information Technology, vol. 9, no. 6, pp. 562-570, 2012.
[45] Lin J., Ming J., and Crookes D., “A Probabilistic Union Approach to Robust Face
Recognition with Partial Distortion and Occlusion,” in Proceedings of IEEE International
Conference on Acoustics, Speech and Signal Processing, Las Vegas, pp. 993-996, 2008.
[46] Liu Q., Yan W., Lu H., and Ma S., “Occlusion Robust Face Recognition with Dynamic
Similarity Features,” in Proceedings of the 18th International Conference on Pattern
Recognition, Hong Kong, vol. 3, pp. 544-547, 2006.
[47] Kurita T., Pic M., and Takahashi M., “Recognition and Detection of Occluded Faces
by a Neural Network Classifier with Recursive Data Reconstruction,” in Proceedings of the
IEEE Conference on Advanced Video and Signal Based Surveillance, pp. 53-58, 2003.
[48] Rama A., Tarres F., Goldmann L., and Sikora T., “More Robust Face Recognition by
Considering Occlusion Information,” in Proceedings of the 8th IEEE International Conference on
Automatic Face & Gesture Recognition, Amsterdam, pp. 1-6, 2008.
[49] Jia H. and Martinez A., “Face Recognition with Occlusions in the Training and Testing
Sets,” in Proceedings of the 8th IEEE International Conference on Automatic Face & Gesture
Recognition, Amsterdam, pp. 1-6, 2008.
[50] Yan S., Wang H., Liu J., Tang X., and Huang T., “Misalignment-Robust Face Recognition,”
IEEE Transaction on Image Processing, vol. 19, no. 4, pp. 1087-1096, 2010.
[51] Tarres F. and Rama A., “A Novel Method for Face Recognition under Partial Occlusion or
Facial Expression Variations,” in Proceedings of the 47th International Symposium, pp. 163-166,
2005.
[52] Ciocoiu B., “Occluded Face Recognition using Parts Based Representation Methods,” in
Proceedings of the European Conference on Circuit Theory and Design, Romania, vol. 1,
pp. 315-318, 2005.
[53] Faaya T. and Toygar O., “Illumination Invariant Face Recognition under Various Facial
Expressions and Occlusions,” in Proceedings of the 3rd International Conference on Image and
Signal Processing, Berlin, vol. 5099, pp. 304-311, 2008.
[54] Tan X., Chen S., Zhou Z., and Liu J., “Face Recognition under Occlusions and Variant
Expressions with Partial Similarity,” IEEE Transaction on Information Forensics and Security,
vol. 4, no. 2, pp. 217-230, 2009.
[55] Tan X., Chen S., Zhou Z., and Zhang F., “Recognizing Partially Occluded, Expression
Variant Faces from Single Training Image Per Person with SOM and Soft kNN Ensemble,” in
Proceedings of IEEE Transaction on Neural Networks, vol. 16, pp. 1-13, 2005.
[56] Oh H., Lee K., and Lee S., “Occlusion Invariant Face Recognition using Selective Local
Non- Negative Matrix Factorization Basis Images,” Image and Vision Computing, vol. 26, no.
11, pp. 1515-1523, 2008.
[57] Gundimada S. and Asari V., “Facial Recognition using Multi Sensor Images Based on
Localized Kernel Eigen Spaces,” IEEE Transactions on Image Processing, vol. 18, no. 6, pp.
1314-1325, 2009.
[58] Marsico M., Nappi M., and Riccio D., “FARO: Face Recognition Against Occlusions and
Expression Variations,” IEEE Transactions on System, Man, and Cybernetics-Part A: Systems
and Humans, vol. 40, no. 1, pp. 121-132, 2010.
[59] D. J. Jobson, Z. Rahman, and G. A. Woodell, “Properties and performance of face
Recognition”; IEEE Transactions on Image Processing, Vol.6,issue.3,pp.451-462, 1997.
[60] Abate A., Nappi M., Riccio D., and Tucci M., “Occluded Face Recognition by Means of the
IFS,” in Proceedings of the 2nd International Conference Image Analysis and Recognition,
Berlin, pp. 1073-1080, 2005.
[61] Sharif M., Younas J., and Sajjad M., “Face Recognition Based on Facial Features,”
Research Journal of Applied Sciences, Engineering and Technology, vol. 4, no. 17, pp. 28792886, 2012.
[62] Martinez A., “Recognition of Partially Occluded and/or Imprecisely Localized Faces using a
Probabilistic Approach,” in Proceedings of IEEE Conference on Computer Vision and
Pattern Recognition, Hilton Head Island, vol. 1, pp. 712-717, 2000.
[63] Farb G., Delmas P., Morris J., and Shorin A., “Robust Face Matching under Large
Occlusions,” in Proceedings of the 14th International Conference on Image Analysis and
Processing, Modena, pp. 448-453, 2007.
[64] Deng Y., Li D., Xie X., Lam K., and Dai Q., “Partially Occluded Face Completion and
Recognition,” in Proceedings of the 16th Image Processing IEEE International Conference on
Image Processing, Cairo, pp. 4145-4148, 2008.
[65] Yu X., Tian J., and Liu J., “Active Appearance Models Fitting with Occlusion,” in
Proceedings of the 6th International Conference on Energy Minimization Methods in Computer
Vision and Pattern Recognition, Berlin, pp. 137-144, 2007.
[66] Lee P., Wang Y., Yang M., Hsu J., and Hung Y., “Distinctive Personal Traits for Face
Recognition under Occlusion,” IEEE Conference on Systems, Man and Cybernetics, Taipei, vol.
5, pp. 4202- 4207, 2006.
[67] Hotta K., “Robust Face Recognition under Partial Occlusion Based on Support Vector
Machine with Local Gaussian Summation Kernel,” Image and Vision Computing, vol. 26, no.
11, pp. 1490- 1498, 2008.
[68] Wright J., Yang A., Ganesh A., Sastry S., and Ma Y., “Robust Face Recognition via Sparse
Representation,” IEEE Transaction on Pattern Analysis and Machine Intelligence, vol. 31, no. 2,
pp. 210-227, 2009.
[69] Sharif M., Kamran A., Mudassar R., and Sajjad M., “Data Reductionality Technique for
Face Recognition,” in Proceedings of the Pakistan Academy of Sciences, vol. 48, no. 4, pp. 229234, 2011.
[70] Lin J., Ming J., and Crookes D., “Robust Face Recognition with Partially Occluded Images
Based on A Single or a Small Number of Training Samples,” in Proceedings of IEEE
International Conference on Acoustics, Speech and Signal Processing, Taipei, pp. 881-884,
2009.
[71] Chen W., Yuen P., Huang J., and Dai D., “Kernel Machine-Based One-Parameter
Regularized Fisher Discriminant Method for Face Recognition,” IEEE Transactions on Systems,
Man, and Cybernetics, Part B: Cybernetics, vol. 35, no. 4, pp. 659-669, 2005.
[72] Kepenekci B., Tek F., and Akar G., “Occluded Face Recognition by using Gabor Features,”
in Proceedings of the 3rd COST 276 Workshop on Information and Knowledge Management for
Integrated Media Communication, Budapest, pp.1-6, 2002.
[73] Lawrence S., Giles C., Tsoi A., and Back A., “Face Recognition: A Convolutional NeuralNetwork Approach,” IEEE Transactions Neural Network, vol. 8, no. 1, pp. 98-113, 1997.
[74] McCloskey S., Langer M., and Siddiqi K., “Removal of Partial Occlusion from Single
Images,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 33, no. 3, pp.
647-654, 2011.
[75] Meng Y., Lei Z., Simon C., and David Z., “Gabor Feature Based Robust Representation and
Classification for Face Recognition with Gabor Occlusion Dictionary,” Pattern Recognition, vol.
46, no. 7, pp. 1865-1878, 2012.
[76] Sharif M., Sajjad M., Abdul H., Younas J., and Mudassar R., “3D Face Recognition using
Horizontal and Vertical Marked Strips,” SINDH University Research Journal, vol. 43, no. 1-A,
pp. 57-62, 2011.
[77] Sharif M., Adeel K., Mudassar R., and Sajjad M., “Face Recognition using Gabor Filters,”
Journal of Applied Computer Science & Mathematics, vol. 5, no. 11, pp. 53-57, 2011.
[78] Wang X., Han T., and Yan S., “An HOG-LBP Human Detector with Partial Occlusion
Handling,” in Proceedings of IEEE 12th International Conference on Computer Vision,
Kyoto, pp. 32-39, 2009.
[79] Richa Shukla, Rohit Raja.,” Grid Based Multiple Features Based On Image Mosaicing”
International Journal of Advanced Research in Computer Engineering & Technology
(IJARCET)Volume 2, Issue 2, February 2013
[80] Yamini Chouhan, Rohit Raja.,” Face Recognition with 3D Head Pose Estimation and Land
Mark extraction Using Gabor Wavelets” International Journal of Science, Engineering and
Technology Research (IJSETR) Volume 2, Issue 5, May 2013
[81] Tilendra Shishir Sinha, Devanshu Chakravarty,Rajkumar Patra and Rohit Raja.,” Modelling
and Simulation for the Recognition of Physiological and Behavioural Traits Through HumanGait
and Face Images” Computer and Information Science » Numerical Analysis and Scientific
Computing » "Discrete Wavelet Transforms - A Compendium of New Approaches and Recent
Applications",: February 6, 2013 Chapter 5
[82] ashutosh.bhensle@gmail.com; rohitraja4u@gmail.com.,” An Efficient Face Recognition
using PCA and Euclidean Distance Classification” IJCSMC, Vol. 3, Issue. 6, pg.407 – 413,
June 2014.
[83] Tilendra Shishir Sinha, Devanshu Chakravarty, Raj Kumar Patra, Rohit Raja.,” Soft-
Computing Based Modeling And SimulationFor Face Identification From Side-View”
International Conference on Systemics, Cybernetics and Informatics, JANUARY 2010