International Journal of Engineering and Technical Research (IJETR)

ISSN: 2321-0869, Volume-2, Issue-12, December 2014

Forgery Detection Based on Blur Inconsistencies:

A Review
Shehnaz, Mandeep Kaur
Passive techniques have been classified on the basis of
Abstract In this Digital era,due to the availability of (1)Tampering operations
easy-to-use photo-manipulation software tools, image security (2)Intrinsic regularities and inconsistencies
has become a prominent question as people are using them for (3)Natural and computer graphic image
their malicious purposes in various fields for example in
journalism, social networking, politics, criminalism etc.
Whenever tampering is done, it affects or changes the basic
Clues for Passive forgery detection methods
characteristics of images such as blurriness, sharpness ,noise, Based on Based on Natural and
luminance intensity etc.There are various methods which detects Tampering Intrinsic computer graphic
divergence in different characteristics. Now days, blur operation operations regularities and image
has become a common tool to hide visual blueprints such as inconsistencies
discontinuity in intensities among different regions. This tool is JPEG Ilumination Color features
being used to commit copy& move and splicing forgeries. But compression
this operation changes blur consistencies pattern of original properties
image. If these blur inconsistencies get detected, we can conclude
that image has been tampered. Therefore, in this paper, we Copy and Move Lens Radial Image Quality
provide a review of tampering detection methods based on blur metrics
inconsistencies. splicing Distortion Natural image
statistics
Index Terms Blur estimation, Blur inconsistencies Resampling chromatic
,perceptual , non-perceptual aberration
Brightness and
I. INTRODUCTION Contrast
Tampering methods/forgeries can be Cropping and
Copy and Move/Region Duplication Recompression
Splicing/Image Composites Sharpening and
Cloning Blurriness
Although Digital image tampering has become easy Yet it can
be detected using different techniques. These techniques are Here our aim is to present how blurring can be used to
classified into two categories: 1) Active 2)Passive detect image forgery. Blurring is very common method used
often to reduce the degree of discontinuity or to hide splicing
Active Techniques: They all require some prior /extra effects or to give real situation effect using artificial blur.
knowledge about original image that should be embedded at Blurring is being used in splicing and copy and move
the time of image production as mark of originality. These forgeries. An image can have different types of blurs such as
techniques are good only if embedded code/information Defocus blur, Motion Blur, Out-of-focus blur. But,
remains intact in the forged image also. Watermarking and identifying blur inconsistencies in whole forged image can
Digital signatures are common examples of Active aid to detect forgeries. Blur can be introduced in a region or at
techniques. But Their drawbacks are the edges. Therefore, different techniques detecting blur
They cannot work if embedded code get damaged. inconsistencies are categorised as Edge based ,Region based
They require special expensive hardware and software ,Edge-cum-Region based.
requirements to detect forgery.
II. PREVIOUS WORK
Passive techniques: Unlike active techniques, they do not
require additional information such as digital watermarking A. Edge based blur inconsistencies
and digital signatures to detect forgery. They are based on the Zhang and Zhang[1] proposed method to detect presence of
fact that tampering makes changes in statistical characteristics feather operation that is useful to create a smooth transition
of original image. They detect forgery based on between the selected region and its surroundings. Xin Wang
inconsistencies in different characteristics of different parts of et.al[2] proposed a method to detect forgery using defocus
the forged image. model. It estimates blurring at edge pixels to check defocus
blur inconsistency through local blur estimation using
Manuscript received December 22, 2014. Elder-Zucker method.It is based on the fact that image
Shehnaz, Information Technology, University Institute of Engineering patches with similar distances to lens have similar blur kernel
and Technology, Chandigarh,01733253286, sizes.This method is robust to digital photofinishing and
Mandeep kaur, Information Technology, University Institute of
Engineeringand Technology, Chandigarh, scanning. Junwen Wang et. al[3] developed a method to
. detect manual edges from a tampered image. It uses

198 www.erpublication.org
 Forgery Detection Based on Blur Inconsistencies: A Review

subsampled contourlet transform to analyse image edges.It C. Blur Estimation

extracts the difference between the normal edge and blur edge Most of the Forgery Detection method based on blur
by analysing phase congruency and prediction error image. inconsistencies use blur estimation metric to measure the
And these features are used to train the SVM to differentiate amount of blurriness in the tampered image. To assess image
blurred edges. This method detects blurring and locates the quality by human himself is quite time consuming process,
tampering boundary with a relative high accuracy. Patchara inconvenient and expensive method. Therefore, there is need
et.al[4] proposed a method to detect color image forgery to automate image quality assessment methods.They all
based on image luminance and image chroma. It extracts calculates blur scores for blurriness. There are two types of
some image features from image luminance using image quality assessment techniques as
rake-transform and from image chroma by using edge objective(non-perceptual) method and subjective(perceptual)
statistics. method. Objective image quality assessment methods predict
the perceived image quality automatically. Image quality
B. Region Based Blur inconsistencies assessment methods are classified according to the
D.Y.Hsiao and S.C.Pei[5] proposed a method to detect availability of a reference (original)asFull
blurred region using image DCT coefficients and optimal reference(FR),reduced-reference(RR) and no-reference(NR).
morphological operations. Y.Sutcu,B.Coskun [6] developed a Full reference based metrics requires both distorted and
method to detect copy and move forgery based on the idea original image to estimate the quality of distorted image.
that sharpness/blurriness of the tampered region is affected Whereas reduced reference approaches requires some
due to copy &move of particular region on the original information about the original image apart from the distorted
image.It is based on the regularity property of wavelet image. But no-reference approaches do not require any
coefficients that measure the decay of wavelet coefficients original image to assess the distorted image quality. As there
across scales. R.M Bora et.al [7] proposed a method to detect is usually problem of availability of original images as
splicing forgery. It is gradient based PBM method which is reference, therefore, no-reference based quality metrics are
more effective than Cepstral method.It uses blur estimate more useful and purposeful. There are many non-perceptual
measure to help in segmentation of inconsistent regions of (objective)and perceptual(subjective)blur metrics for image
images that has small amounts of motion blur and quality assessment. Each type of blur metric has its
no-reference Perceptual blur Metric(PBM) to detect advantages and limitations.
directional motion blur in forged image.It uses K-means
Type of blur Advantages Limitations
algorithm to achieve its effectiveness.Tao Wang et.al.[8]
metric
proposed a method to detect copy and move forgery. It is
Non Ensures More time
based on merging blur and affine moments invariants.It
-Perceptual accuracy complexity
detects duplicated and distorted regions.But it cannot detect
forgery when degradation scales get increased. Pravin kakar Perceptual Less Time Estimates only
et.al[9] provides a novel method based on discrepancies in complexity human perceptible
motion blur.It is based on estimation of motion blur through image degradation in
image gradients to detect motion inconsistencies to help in quality but doesnot
splicing detection.Fei Peng [10] proposed a method to detect assess low level
the presence of artificial blur which is considered as digital degradation
image forgery.It is based upon the fact that blur operation Perceptual/subjective metrics have been developed because
destroys joint consistency of color channel in the image. of the reason to automate the process to assess image quality
Firstly, it obtains blur region through blur estimation measure in terms of blurriness amount. Now we introduce latest
and then locates the artificial blur region by detecting developed perceptual no-reference blur image quality metric
abnormal hue in the blur region .It achieves better for image quality assessment. Yongfeng Wang et.al [14]
performance in detecting defocus blur and artificial blur. proposed no-reference perceptual-based blur metrics that
Pravin et. al[11] proposed method to detect motion blur detects blurred edges in cost effective way. It is based on
inconsistencies using spectral matting to help splicing perceptual-based edge analysis. Zhirong et. al[15] also
detection. They also developed a new measure to do developed a new no-reference perceptual blur metric that is
inconsistent region segmentation in images that contain small based on the analysis of the spread of edge and the study of
amounts of blur.It has been proved an effective method than human blur perception for varying contrast values. This
other existing blur based techniques.Zhipeng[12] et.al metric yields high accuracy.. Fatma Kerouh et.al [16]
provides another method to detect image tampering. It detects proposed another perceptual blind blur image quality metric
global or local blur manipulation using no-reference image that was developed in wavelet domain. It combines objective
quality metric.It extracts image features from MSCN measure based on edge analysis through wavelet transform
(meansubtractedcontrastnormalized)coefficients of different resolutions and Just Noticeable Concept(JNB).
regions to quantify tampered regions. But it does not work
good for images with poor resolution. Khosrobahrami III. CONCLUSION
et.al[13] developed a method to detect multi-type blurred
images. It clusters the blocks having similarity of local blur In this paper, we have presented a review of approaches based
kernels using K-means clustering algorithm. And Then it on blur inconsistencies to detect different forgeries.To
classifies the blur types of the clustered region into analyse these blur inconsistencies, firstly we need to estimate
out-of-focus or motion blur using a minimum distance blur amount in different regions of the forged image.For this
classifier. purpose we also presented different blur estimation metrics in

199 www.erpublication.org
 International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869, Volume-2, Issue-12, December 2014
[14],[15],[16]that measures the blur amount. These metrics work to compare these metrics.
have been tested over different datasets. So, in future, we will

Overview of the Previous work on forgery detection based on blur inconsistencies.

Year Author Proposed Model/ Advantages Limitations
Method
2005 Hsia and Pei DCT coefficients and Detects blurred regions Some detected inconsistent
optimal regions lead to false
morphological interpretations
operations.
2007 Y.Sutcu,B.C Regularity of wavelet Detects copy and move forgery
oskun coefficients
2008 XinWang Defocus Robust to photofinishing and scanning Detects defocus blur
et.al inconsistencies only
2010 Junwen et.al subsampled Locates tampering boundary with high Tested images are from
contourlet transform accuracy PEWNTAX K100D and canon
A710 digital camera.
2010 Patchara Based on Natural Better for natural images Cannot be used for internet
et.al image model and images .
Rake transform and Its computation time is relatively low
edge statistics
2010 based on Detects splicing forgery, Effective Good only for motion blur
discrepancies in method
motion.
2010 Pravin et. al matting the estimates the motion blur Slower than DCT based
components of the effectively. technique
image. detects splicing forgery
Better inconsistent region
interpretation for user
2010 Fei Peng abnormal hue in the detects the presence of artificial blur Detects only motion blur
et.al blur region . and defocus blur efficiently. inconsistencies
Good detection rate
No any influence of the texture of the
background and image size
2011 Pravin kakar Based on spectral Simple and speedy segmentation to More number of steps are
et.al matting detect inconsistent regions efficiently involved
estimates the motion blur
effectively.
detects splicing forgery
Better inconsistent region
interpretation for user
2013 Tao Wang Merge blur and affine detects copy and move forgery cannot detect forgery large
et.al moments invariants. degradation scales

2013 Zhipeng et.al MSCN coefficients Good for detecting guassian blur and Not good for images with poor
blur operation by mean filter resolution.
Cannot detect guassian blur
with sigma 0.1
2014 Khosrobahra similarity of local Detects multi-type blurs ( Cannot detect other complex
mi et.al blur kernels using Symmetric out-of-focus or uniform forms of kernel
K-means clustering motion blur)
algorithm and
minimum distance
classifier.

REFERENCES Conference on Intelligent Information Hiding and Multimedia Signal

Processing,2008.
[1] ] C. Zhang, H. Zhang, Detecting digital image forgeries [3]Junwen Wang , Guangjie , Bo Xu Hongyuan Li,Yuewei Dai, Zhiquan
throughweighted local entropy , in IEEE International Symposium on Wang, Image forgery Forensics based on manual blurred edge
Signal Processing and Information Technology, Giza, Egypt,December detection in IEEE International conference on Multimedia Information
2007, pp. 6267. Networking and Security,
[2]Xin Wang , Bo Xuan ,Si-long Peng , Digital Image Forgery detection [4]Patchara Sutthiwan,Yun Q.Shi,Wei Su,Tian-Tang Ng, Rake transform
based on the consistency of defocus blur in IEEE International and Edge statistics for image forgery detection in IEEE,2010

200 www.erpublication.org
 Forgery Detection Based on Blur Inconsistencies: A Review

[5]D.Y.Hsiao and S.C.Pei,Detecting digital tampering by blur estimation,

in Proceedings of the Ist International Workshop on Systematic
Approaches to Digital Forensic Engineering
.Taipei,china:IEEE,2005,pp.264-278
[6]Y.Sutcu,B.Coskun,Tamper detection based on regularity of wavelet
transform coefficients in IEEE, 2007.
[7] R.M Bora, N.M Shahane, Image forgery Detection through Motion Blur
Estimates in IEEE 2012.
[8] Tao Wang , Jin Tang,Bin Luo, Blind Detection of Region Duplication
Forgery by Merging Blur and Affine Moment invariants in IEEE 2013
Seventh International Conference on Image and Graphics.
[9] Pravin Kakar,Sudha Natarajan, Wee Ser, Detecting digital image
forgeries through inconsistent motion blur, in IEEE,2010
[10]FeiPeng,Xi-lanWang,DigitalImage Forgery Forensics by using Blur
estimation and Abnormal Hue detection in IEEE,2010.
[11]Pravin Kakar , Sudha Natrajan, Wee Ser,Exposing Digital Image
Forgeries by detecting Through discrepancies in Motion Blur in IEEE
Transactions On Multimedia ,Vol.13,No.3,June 2011
[12] Zhipeng,Chen,Yao and Rongrong Ni, Forensics blurred images based
on no-reference image quality assessment in IEEE, 2013.
[13] Khosrobahrami and Alex C.Kot,Image Tampering Detection By
Exposing Blur Type Inconsistency in 2014 IEEE International
Conference on Acoustic ,Speech and Signal Processing(ICASSP)
[14]Yongfeng Wang, Haiquing Du, Jingtao Xu,Yong Liu, A no-reference
perceptual blur metric based on complex edge analysis in IEEE
Proceedings of IC-NIDC2012.
[15]Zhirong LI,Yong LIU, Jingtao Xu, Haiquing Du, A no-reference
perceptual blur metric based on blur ratio of detected edges in
Proceedings of IEEE IC-BNMT2013.
[16]Fatma Kerouh and Amina Serir,A perceptual Blind Blur Image Quality
Metricin 2014 IEEE International Conference on Acoustic ,Speech and
Signal Processing(ICASSP)

Shehnaz received her B.Tech degree in Computer

Science and Engineering from UIET, Kurukshetra
University ,Kurukshetra (India) in 2011.Now, she is
pursuing Masters in Engineering (Information
Technology) from UIET, Panjab University,
Chandigarh(India).

Mandeep Kaur received her B.Tech degree in

Computer Science and Engineering from BCET,
Punjab TechnicalUniversity, Punjab (India) in 1999
and received her Masters in Engineering (Information
Technology) from PEC, Panjab University, Chandigarh
(India) in 2004. She is associated with University
Institute of Engineering and Technology, Panjab
University, Chandigarh (India) since 2005. Her area of
interests include Digital image processing and Image
forensics.

201 www.erpublication.org