Paper no Year Citation Title

1 2019 Pei Li, Loreto Prieto, Domingo Mery and Patrick J. On Low-Resolution
Flynn,"On Low-Resolution Face Recognition in the Face Recognition in
Wild: Comparisons and New Techniques", DOI the Wild: Comparisons
10.1109/TIFS.2018.2890812, IEEE and New Techniques
Transactions on Information Forensics and Security

2 2015 Huang, Zhiwu, Shiguang Shan, Ruiping Wang, A benchmark and

Haihong Zhang, Shihong Lao, Alifu Kuerban, and comparative study of
Xilin Chen. "A benchmark and comparative study video-based face
of video-based face recognition on COX face recognition on COX
database." IEEE Transactions on Image Processing face database
24, no. 12 (2015): 5967-5981.

3 2018 Ding, Changxing, and Dacheng Tao. "Trunk-branch Trunk-Branch

ensemble convolutional neural networks for video- Ensemble
based face recognition." IEEE transactions on Convolutional Neural
pattern analysis and machine intelligence 40, no. 4 Networks for Video-
(2018): 1002-1014. based Face
Recognition

4 2017 Dhamija, Jalendu, Tanupriya Choudhury, Praveen An Advancement

Kumar, and Yogesh Singh Rathore. "An towards Efficient Face
Advancement towards Efficient Face Recognition Recognition Using Live
Using Live Video Feed:" For the Future"." In Video Feed: For the
Computational Intelligence and Networks (CINE), Future
2017 3rd International Conference on, pp. 53-56.
IEEE, 2017.

5 2019 Mokhayeri, Fania, Eric Granger, and Guillaume- Domain-specific face

Alexandre Bilodeau. "Domain-specific face synthesis for video
synthesis for video face recognition from a single face recognition from a
sample per person." IEEE Transactions on single sample per
Information Forensics and Security 14, no. 3 person
(2019): 757-772.
6 2010 Zhao, Qianqian, and Hualong Cai. "The research The research and
and implementation of face detection and implementation of
recognition based on video sequences." In Future face detection and
Computer and Communication (ICFCC), 2010 2nd recognition based on
International Conference on, vol. 1, pp. V1-318. video sequences
IEEE, 2010.

7 2012 An, Le, Bir Bhanu, and Songfan Yang. "Face Face recognition in
recognition in multi-camera surveillance videos." multi-camera
In Pattern Recognition (ICPR), 2012 21st surveillance videos
International Conference on, pp. 2885-2888. IEEE,
2012.

8 2010 Lu, Ke, Zhengming Ding, Jidong Zhao, and Yue Wu. Video-based Face
"Video-based face recognition." In Image and Recognition
Signal Processing (CISP), 2010 3rd International
Congress on, vol. 1, pp. 232-235. IEEE, 2010.

9 2016 Hatimi, Hicham, Mohamed Fakir, and Mohamed Face recognition using
Chabi. "Face recognition using a fuzzy approach a fuzzy approach and a
and a multi-agent system from video sequences." multi-agent system
In Computer Graphics, Imaging and Visualization from video
(CGiV), 2016 13th International Conference on, pp. sequences
442-447. IEEE, 2016.

10 2009 Hu, Changbo, Josh Harguess, and J. K. Aggarwal. PATCH-BASED FACE

"Patch-based face recognition from video." In RECOGNITION FROM
Image Processing (ICIP), 2009 16th IEEE VIDEO
International Conference on, pp. 3321-3324. IEEE,
2009.
11 2017 Wang, Ya, Tianlong Bao, Chunhui Ding, and Ming Face Recognition in
Zhu. "Face recognition in real-world surveillance Real-world
videos with deep learning method." In Image, Surveillance Videos
Vision and Computing (ICIVC), 2017 2nd with Deep Learning
International Conference on, pp. 239-243. IEEE, Method
2017.

12 2013 Gong, Dihong, Kai Zhu, ZhiFeng Li, and Yu Qiao. "A A Semantic Model for
semantic model for video based face recognition." Video Based Face
In Information and Automation (ICIA), 2013 IEEE Recognition
International Conference on, pp. 1369-1374. IEEE,
2013.

13 2019 He, Lingxiao, Haiqing Li, Qi Zhang, and Zhenan Sun. Dynamic Feature
"Dynamic Feature Matching for Partial Face Matching for Partial
Recognition." IEEE Transactions on Image Face Recognition
Processing 28, no. 2 (2019): 791-802.

14 2018 Su, Ya. "Robust video face recognition under pose Robust Video Face
variation." Neural Processing Letters 47, no. 1 Recognition Under
(2018): 277-291. Pose Variation

15 2016 Yang, Shuo, Ping Luo, Chen-Change Loy, and WIDER FACE: A Face
Xiaoou Tang. "Wider face: A face detection Detection Benchmark
benchmark." In Proceedings of the IEEE
conference on computer vision and pattern
recognition, pp. 5525-5533. 2016.

16 2016 Pandit, Trupti M., P. M. Jadhav, and A. C. Phadke. Suspicious Object

"Suspicious object detection in surveillance videos Detection In
for security applications." In Inventive Surveillance Videos
Computation Technologies (ICICT), International For Security
Conference on, vol. 1, pp. 1-5. IEEE, 2016. Applications
17 2016 Cheng, De, Yihong Gong, Sanping Zhou, Jinjun Person Re-
Wang, and Nanning Zheng. "Person re- Identification by Multi-
identification by multi-channel parts-based cnn Channel Parts-Based
with improved triplet loss function." In CNN with Improved
Proceedings of the IEEE Conference on Computer Triplet Loss Function
Vision and Pattern Recognition, pp. 1335-1344.
2016.

18 2015 Parkhi, Omkar M., Andrea Vedaldi, and Andrew Deep face recognition
Zisserman. "Deep face recognition." In BMVC, vol.
1, no. 3, p. 6. 2015.
19 2015 Schroff, Florian, Dmitry Kalenichenko, and James Facenet: A unified
Philbin. "Facenet: A unified embedding for face embedding for face
recognition and clustering." In Proceedings of recognition and
the IEEE conference on computer vision and clustering
pattern recognition, pp. 815-823. 2015.

20 2002 Yang, Ming-Hsuan, David J. Kriegman, and Detecting faces in

Narendra Ahuja. "Detecting faces in images: A images: A survey
survey." IEEE Transactions on pattern analysis and
machine intelligence 24, no. 1 (2002): 34-58.

21 2004 Yang, Ming-Hsuan. "Recent advances in face Recent advances in

detection." IEEE ICPR 2004 Tutorial (2004). face detection
22 2013 Ohlyan, Sonia, Sunita Sangwan, and T. Ahuja. "A A Survey On Various
survey on various problems & chal-lenges in face Problems & Challenges
recognition." Int. J. Eng. Technol.(IJERT) 2, no. 6 In Face Recognition
(2013): 2278-0181.

23 2014 Taigman, Yaniv, Ming Yang, Marc'Aurelio Ranzato, Deepface: Closing the
and Lior Wolf. "Deepface: Closing the gap to gap to human-level
human-level performance in face verification." In performance in face
Proceedings of the IEEE conference on computer verification
vision and pattern recognition, pp. 1701-1708.
2014.
Description Conclusion
Low resolution FR and Person ReID on different deep convolutional generative adversarial
datasets. 4 methods- networks (DCGAN) pretraining on training
A) super-resolution techniques, dataset improves performance of LR FR
B) comparison between virtual and real LR
C) face identification
D) face re-identification.

proposes a novel Point-to-Set Correlation Learning Point to set corelation method to consider
(PSCL) method for Video to still, still to video or heterogenity between still image(euclidean
video to video face recognition. point) and video sequence (Rimannian
elements)and outperforms then state of the
art methods. However, still Video based face
recognition needs more effort.

Uses TBE-CNN with improved triplet loss function TBE-CNN incorporates one trunk network
to tackle occlusion , pose variation and blurr and several branch networks. The trunk
images for Video based FR network is trained to learn face
representations for holistic face images, and
each branch network is trained to learn face
representations for image patches cropped
from one facial component.

Uses two methods to derive proposedmethod. PCA+Fisherface+SVD method improved FR

1. PCA and Fisherface (Fisheface is used for feature accuracy to 99.5%
extraction)
2. SVD-based face recognition (Singular value
decomposition) This is like 2D PCA
Proposed method- Train these two models and
rank obtained results. This gives final result-
recognized face.

The performance of still-to-video FR systems can
decline significantly when individual has very few
or single reference image and that too with
different capture conditions.
Domain specific face synthesis algorithm exploits
intra class variation information to augment
reference set by caracterizing capture conditions
augmenting the reference gallery set of still-
to-video FR systems using the proposed
DSFS approach can provide a significantly
higher level of accuracy compared to state-
of-the-art approaches, with only a moderate
increase in its computational complexity.
this paper uses
1- the method of difference in
background images and the Kalman filter to track
and extract the region of human body
2-uses the AdaBoost algorithm to detect human
face in the region.
3- improved Hidden Markov Model which is
named as Pseudo two-dimensional Hidden Markov
Model was used for feature extraction and
recognition in face image

the poses of probe data are usually not frontal Face recognition performance is better than
view, contrary to the standard format of the gallery the result of any single camera
data.
This paper propose a multi-camera video based
face recognition framework using a novel image
representation called Unified Face Image (UFI),
which is synthesized from multiple camera feeds.
The probe frames from different cameras are
wraped towards a template frontal face and then
averaged. The generated UFI is a frontal view of
the subject. Used SIFT flow as a high level
alignment tool to warp the faces.

uses Locality Preserving Projections (LPP) to

recognize video-based face sequence, so it can
discover more space-time semantic information
hidden in video face sequence, simultaneously
make full use of the intrinsic nonlinear structure
information to extract discriminative manifold
features.

fuzzy logic is used for classification.

In the first step, faces are detected using texture
color and geometrical face.
In the second step, the multi-agent system and
fuzzy approach are used in the recognition process

First, face patches are cropped from the video

frame by frame. Then, face patches are matched to
an overall face model and stitched together. By
accumulating the patches, a reconstructed face
is built which is used in recognition.
In short it is reconstructing the full face from
patches.
In the first part, a dataset was constructed by
automatically collecting and labelling the data
from realworld surveillance video.
In the second part, fine-tune the VGG face
recognition model by our dataset.

training faces from the same person are clustered

semantically with the K-means algorithm, then
faces from different persons are further grouped
according to semantics. Finally, face features from
each group are sliced into n slices, and PCA+LDA
algorithm is applied yielding n*3 classifiers.
At the testing stage, the matching score of the
given pair of testing video sequences is computed
by semantically assigning each frame into one of
the groups. Final matching score is the sum of the
best matching scores from each group

FCN is applied to extract spatial feature maps of

given gallery and probe faces. Then sliding window
of the same size as the probe feature maps is set
to decompose the gallery feature maps into
several gallery subfeature maps. Finally, Sparse
Representation Classification (SRC) is used to
achieve alignment-free matching.

The key idea is performing alignment and

recognition based on sparse representation
simultaneously.
RVPose- By considering that multi-pose faces
possess the same texture and 3-dimensional
shape, RVPose aligns a sequence of faces with
pose variations simultaneously, which is reduced
to a 3-dimensional shape-constrained video
alignment problem.
Algorithms Dataset adv disadv
super-resolution Scface deep
algorithm UCCface

Point-to-Set Cox face

Correlation db
Learning (PSCL)
method

TBE-CNN PaSC, COX Good deep

Face, and paper
YouTube
Faces

PCA+Fisherface+S AT&T
VD Face
database
ORL
images

Domain specific Cox face advanced

face synthesis db (S2V)
algorithm
Object tracking , Own Object May fail HMM
haar features, tracking for image
AdaBoost, method people
Improved Hidden with
Markov model different
skin tone
like asian,
afrikan,
amerian

SIFT ChokePoi can use Handles

LBP for feature nt dataset sample pose
extraction and k- images
NN for from this
classification paper

Locality Preserving UCSD/Ho not of

Projections (LPP) nda Video much use
Database

Multi-agent ORL variations

system and Fuzzy Database illuminati
logic on and
speed

uses sparse Yaleface handles

representation to database noise and
recognize occlusion
reconstructed
faces.
CNN(VGG) Own

K-means, XM2VTS
PCA+LDA

Fully CAISA- good

Convolutional NIR- image
Networks (FCNs) Distance,
and Sparse CASIA-
Representation NIR-
Classification Mobile,
(SRC) and LFW

RVPose YTC, YTF

and
MobFaces
category

good image

good image
Too bright or too dark
could lead to error
results for the detection
of face region in
recognition.
speed of human
motion is one of the
influencing factors.Too
fast or too slow also
would lead to the failure
in extraction of the body
contour.

large changes in pose,

illumination and
expression cannot
currently be handled
PCA means igenfaces
LDA means fishefaces
 Literature
Year Technique Pros
source
2019 [1] DAGAN Handles Low Resolution
2018 [3] TBE-CNN with improved triplet loss Deals with Occlusion, Pose, Low Resolution
2017 [11] Transfer Learning with VGG face model Improves Performance
Improves performance
2019 [13] FCN+ SRC of partial face recognition
2018 [14] RVPose Handles Pose variation
2019 [5] DSFS + SRS FR based on a single facial image
2017 [4] LDA+SVD Higher efficiency and accuracy
effective extract of motive human
body in the video
2010 [6] AdaBoost + Hidden Markov Model and then detection and
recognition of face
handles noise and
2009 [10] Patch based method occlusion better
acquire more semantic
2010 [8] LPP information
inside the video face
2012 [7] SIFT LBP + K- Means Handles pose variations and multi views
great improvement
2013 [12] K- Means + PCA+ LDA over the traditional face
recognition approaches

outperforms most of the existing

2015 [2] PSCL methods
for V2S/S2V face recognition tasks

2016 [9] Multi Agent + Fuzzy handles variations illumination and speed
 Limitations Dataset accuracy

ose, Low Resolution

92.10%

moderate increase in computational

complexity
Changes in face affects efficiency 93.92%

Illumination and speed of human body in video leads to errorneous results

large changes in pose, illumination

and expression cannot be handled

s and multi views

95.33

mination and speed