Image Forgery Detection Using Feature Point Matching and Adaptive Over Segmentation

IDL - International Digital Library Of
Technology & Research
Volume 1, Issue 4, Apr 2017 Available at: www.dbpublications.org
International e-Journal For Technology And Research-2017
IDL - International Digital Library 1 | P a g e Copyright@IDL-2017
Image Forgery Detection Using Feature
Point Matching and Adaptive Over
Segmentation
Seema Sultana, Sunanda Dixit
Department of Information Sciences, Dayananda Sagar College of Engineering,
Bangalore, India
Seemasultna17@gmil.com
Abstract:A duplicate move fabrication
location conspire utilizing highlight point
coordinating and versatile over-division is
worked here. This plan coordinates both
square based and key point-based falsification
location strategies. To start with, the proposed
Adaptive Over-Segmentation calculation
fragments the host picture into non-covering
and sporadic pieces adaptively. At that point,
the component focuses are removed from each
piece as square elements, and the square
elements are coordinated with each other to
find the named include focuses, this strategy
can around show the presumed fraud districts.
To recognize the phony locales all the more
precisely, the Forgery Region Extraction
calculation is exhibited, which replaces the
component focuses with little superpixels as
highlight pieces. At that point combines the
neighboring hinders that have comparable
nearby shading highlights into the component
squares to create the blended areas. At last, it
applies the morphological operation to the
combined locales to create the identified
fabrication areas. The trial comes about show
that the proposed duplicate move falsification
identification plan can accomplish much better
location comes about even under different
testing conditions contrasted and the current
cutting edge duplicate move fabrication
discovery strategies.
Key words: Adaptive-Over Segmentation,
Forgery Region Extraction, Fabrication areas
1 Introduction
The presentation and quick spread of
computerized control to even now and moving
pictures raises moral issues of truth,
misdirection, and advanced picture
uprightness. With experts testing the moral
limits of truth, it makes a potential loss of
open trust in computerized media [1]. This
spurs the requirement for location devices that
are straightforward to altering and can tell

whether a picture has been altered just by
investigating the altered picture. Picture
altering is a computerized craftsmanship
which needs comprehension of picture
properties and great visual imagination [2].
One alters pictures for different reasons either
to appreciate fun of computerized works
making mind boggling photographs or to
deliver false confirmation. Regardless of
whatever the reason for act may be, the
falsifier ought to utilize a solitary or a blend
arrangement of picture handling operations.
Image Segmentation
Division is the way toward apportioning an
advanced picture into different portions (sets
of pixels) [3]. The objective of division is to
streamline and additionally change the
portrayal of a picture into something that is
more important and simpler to break down.
Picture division is normally used to find
articles and limits (lines, bends, and so on.) in
pictures. All the more correctly, picture
division is the way toward doling out a mark
to each pixel in a picture to such an extent that
pixels with a similar name share certain visual
qualities.
Image Forgery Detection
Computerized picture fraud has been
progressively simple to perform, because of
the advancement of PC innovation and picture
preparing programming. Nonetheless,
advanced pictures are a well-known wellspring
of data, and the dependability of computerized
pictures is along these lines turning into an
essential issue [4]. As of late, an ever
increasing number of scientists have started to
concentrate on the issue of computerized
picture altering. Of the current sorts of picture
altering, a typical control of an advanced
picture is duplicate move falsification, which
is to glue one or a few replicated region(s) of a
picture into different part(s) of a similar
picture.
Copy-Move Operation
Amid the duplicate and move operations, some
picture preparing techniques, for example,
revolution, scaling, obscuring, pressure, and
clamor expansion are once in a while
connected to make persuading falsifications.
Since the duplicate and move parts are
replicated from a similar picture, the
commotion segment, shading character and
other essential properties are perfect with the
rest of the picture; a portion of the imitation
recognition strategies that depend on the
related picture properties are not pertinent for
this situation. In earlier years, numerous
fabrication location techniques have been
proposed for duplicate move imitation
identification.
2 Detection Methods
The copy-move forgery detection methods fall
under two main categories:

 Block-based algorithms
 Feature keypoint-based algorithms.
Block-Based Methods:
In the existing block-based methods, the image
is segmented into overlapping and regular
image blocks. Then the forgery regions are
identified by matching blocks of image pixels
or transform coefficients. Some of the bloc-
based methods are PCA, DCT, DWT, and
SVD [5].
Key-point based algorithms:
An alternative to the block-based methods,
keypoint-based forgery detection methods are
proposed, where image keypoints are extracted
and matched over the whole image to resist
some image transformations while identifying
duplicated regions.
Scale-Invariant Feature Transform (SIFT) is
applied to host images to extract feature
points, which are then match to one another.
When the value of the shift vector exceeded
the threshold, the sets of corresponding SIFT
feature points are defined as the forgery
region. Speeded Up Robust Features (SURF)
is also applied to extract image feature instead
of SIFT [6].
3 Proposed System
This plan incorporates both the conventional
square based phony location strategies and key
point-based imitation recognition techniques.
Like square based phony identification
strategies, it is a picture blocking strategy
brought the Adaptive Over-Segmentation
calculation to separate the host picture into
non-covering and sporadic pieces adaptively.
At that point, like the key point-based phony
recognition techniques, the element focuses
are extricated from each picture hinder as
square elements as opposed to being removed
from the entire host picture as in the
conventional key point-base strategies [7].
Fig 1: Framework for copy-move forgery detection
scheme
The Adaptive Over-Segmentation calculation,
which is like when the extent of the host
pictures builds, the coordinating calculation of
the covering pieces will be considerably more
costly. To address these issues, the Adaptive
Over-division technique was proposed, which
can portion the host picture into non-covering
areas of sporadic shape as picture pieces a
short time later, the imitation locales can be

recognized by coordinating those non-
covering and unpredictabledistricts.
Divisiontechnique, the non-covering division
can diminish the computational costs
contrasted and the covering blocking;
moreover, as a rule, the sporadic and
significant areas can speak to the fraud district
superior to the consistent squares.
Fig 2: Flowchart of Adaptive-Over-Segmentation
Fig 3: Flowchart of Block Feature Matching
Algorithm
After the host picture is sectioned into picture
squares, piece components are separated from
the picture squares (IB). The conventional
piece based falsification identification
techniques extricated elements of an
indistinguishable length from the square
elements or specifically utilized the pixels of
the picture obstruct as the square components.
Notwithstanding, these components reflect
mostly the substance of the picture pieces,
forgetting the area data. Additionally, these
components are not impervious to different
picture changes. Along these lines, in this
venture, the component focuses are extricated
from each picture hinder as square elements
and the element focuses ought to be powerful
to different twists, for example, picture
scaling, revolution, and JPEG pressure.
4 Results
In addition to the plain copy-move forgery,
have tested proposed scheme when the copied
regions are distorted by various attacks [8]. In
this case, the forged images are generated by
using each of the 48 images in the dataset, and
the copied regions are attacked by geometric
distortions that include scaling and rotation
and common signal processing such as JPEG
compression.

Fig 4: Input Image Fig 5: SIFT Feature Detected Points
Fig 6: SURF Feature Detected Points

Fig 7: Matched Points
Fig 8: Forgery Detected
5 Conclusion
Computerized falsification pictures made with
duplicate move operations are trying to
identify. A novel duplicate move phony
recognition conspire utilizing versatile over-
division and highlight point coordinating is
proposed. The Adaptive Over-Segmentation
calculation is proposed to portion the host
picture into non-covering and sporadic squares
adaptively as indicated by the given host
pictures, utilizing this approach, for each
picture, a fitting piece introductory size to
improve the precision of the fabrication
discovery results can be resolved and, in the
meantime, lessen the computational costs.
Future work could focus on applying the
proposed forgery detection scheme based on
adaptive over-segmentation and feature-point
matching on other types of forgery, such as
splicing or other types of media, for example,
video and audio.
References
[1] J. Fridrich, D. Soukal, and J. Lukáš,
“Detection of copy–move forgery in digital
images,” in Proc. Digit. Forensic Res.
Workshop, Cleveland, OH, Aug. 2003.
[2] X. B. Kang and S. M. Wei, “Identifying
tampered regions using singular value
decomposition in digital image forensics,” in
Proc. Int. pp. 926–930, Dec. 2008.
[3] S. Bayram, H. T. Sencar,“An efficient and
robust method for copy–move forgery,” in
Proc. IEEE Int. Conf. Acoust., Speech, Signal
Process (ICASSP), pp. 1053–1056, Apr. 2009.
[4] S. J. Ryu, M. J. Lee, and H. K. Lee,
“Detection of copy-rotate-move forgery using
Zernike moments,” in Information Hiding.
Berlin, Germany: Springer-Verlag, pp. 51–65,
2010.
[5] S. Bravo-Solorio and A. K. Nandi,
"Exposing duplicated regions affected by
reflection, rotation and scaling," in Acoustics,
Speech and Signal Processing (ICASSP), 2011
IEEE International Conference on pp. 1880-
1883, 2011.

[6] X. Y. Pan and S. Lyu, "Region Duplication
Detection Using Image Feature Matching,"
Ieee Transactions on Information Forensics
and Security, vol. 5, pp. 857-867, Dec 2010.
[7] P. Kakar and N. Sudha, "Exposing Post
processed Copy–Paste Forgeries through
Transform-Invariant Features," Information
Forensics and Security, IEEE Transactions on,
vol. 7, pp. 1018-1028, 2012.
[8] B. Shiva Kumar and L. D. S. S. Baboo,
"Detection of region duplication forgery in
digital images using SURF," IJCSI
International Journal of Computer Science
Issues, vol. 8, 2011.

Image Forgery Detection Using Feature Point Matching and Adaptive Over Segmentation

More Related Content

What's hot (17)

Similar to Image Forgery Detection Using Feature Point Matching and Adaptive Over Segmentation (20)

Recently uploaded (20)

Image Forgery Detection Using Feature Point Matching and Adaptive Over Segmentation