Survey On Technique For Illegal Modification Detection
Survey On Technique For Illegal Modification Detection
Survey On Technique For Illegal Modification Detection
ISSN 2278-6856
Head of Department
Department of Computer Engineering
K. K. Wagh Institute of Engineering Education & Research,Nashik
Savitribai Phule Pune University, Pune
Abstract
Digital images are easy to manipulate and edit due to
availability of powerful image processing and editing software
like picasa, Photoshop. Nowadays, it is possible to add or
remove important part from an image without leaving any
obvious traces of tampering. Verifying digital images and
validating their contents, and identifying forgeries is one of the
critical challenges for governmental and nongovernmental
organizations and departments.
The image integrity
verification as well as identifying the areas of tampering on
images without need of any expert support or manual process
or prior knowledge of original image contents is now days
becoming the challenging research problem. The method as in
[16] deals with authenticity of images and is based on concept
of using illumination color estimation. Recently a new method
is also reported for efficient forgery detection particularly for
faces in images , that estimates illuminant color using the
physics based method as well as statistical edge method which
make the use of inverse intensity-chromaticity color space. The
estimate of illuminant color is extracted independently from
the different mini regions. The method employs SVM for
classification. The technique is capable of handling images
containing two or more people and needs no expert interaction
for detection of tampering.
1. INTRODUCTION
This paper describes the strategy followed by various
techniques to tackle with Image Forensics Challenge on
image forgery detection. Several authors have been
working in recent years on the forgery detection problem,
focusing on techniques based on camera sensor noise, and
on techniques based on dense local descriptors and
machine learning. Therefore, For detection we decided to
follow both these approaches, on two separate lines of
development, with the aim of combining decisions at some
later time of the process. Indeed, it is well known that,
given the different types of forgery encountered in practice,
and the wide availability of powerful photo-editing tools,
several detection approaches should be used at the same
Page 160
2. ILLUMINATION INCONSISTENCIES
In blind image forgeries exposure, investigation of image
automatically is by its assessment of illuminant color
consistency. Methods for illumination color estimation are
machine-learning based. C. Riess and E. Angelopoulos in
[2] presented a different approach by employing a physicsbased color constancy algorithm that operates on partly
reflective pixels. during
this approach, the
automated detection
of extremely
reflective half is
unnoticed. The author implies to segment the image to
estimate the illuminant color per segment. Recoloring
every image region in step with its native illuminant
estimate yields a suspected illuminant map. Unlikely
illuminant color estimates point towards a influenced
region. Unfortunately, the authors do not provide a
statistical decision criterion for forgery detection. Thus, an
expert is left with the difficult task of visually examining
an illuminant map for evidence of tampering.
Inconsistencies in illumination distribution can be used to
identify original and doctored image. Color is generally
used in computer vision, but in a very fundamental,
primitive way. One reason for utilizing very basic color
primitives is that the color information of a pixel is always
a mixture of illumination, geometry and object material.
Consider, for example, changes in illumination, which are
likely the spectrum of sunlight varies over the daytime,
shadows can fall on the object, or fake light is switched on.
Figure 1 shows two examples for different color
appearances. The pictures are element of the dataset. The
picture is once exposed to comparatively neutral (white)
light, and once to illuminants that approximate the
surroundings light at night. Thus, for robustness,
methodologies that make use of color be supposed to
openly address such emergence variations. Two separate
static methods to obtain a color illuminant: the statistical
generalized gray world estimates and the physics-based
inverse-intensity chromaticity space are as given below.
Both schemes do not require training data and are applied
to any image.
ISSN 2278-6856
Page 161
ISSN 2278-6856
ISSN 2278-6856
5.ACKNOWLEDGEMENT
I am thankful to my guide Prof.Dr.S.S.Sane,for his
guidance
and
encouragement
.
Their
expert
suggestionsand scholarly feedback had greatly enhanced
the effectiveness of this work.I would like to express the
deepest appreciation to authors Tiago Jos de Carvalho,
Christian Riess, Elli Angelopoulou, Hlio edrini and
Anderson de Rezende Rocha for their beneficial
information and knowledge
Page 163
ISSN 2278-6856
AUTHOR
Ms.Swapnali Pangre received
the
Diploma degree from K.K.Wagh
Polytechnique MSBTE in 2007 and
B.E. degrees in Computer Engineering
from K. K. Wagh Institute of
Engineering Education & Research
Savitribai Phule Pune University in 2010. Currently,she is
working toward the M.E. degree at the . Savitribai Phule
Pune University,Pune. Her main interests include digital
forensics, pattern analysis, data mining, machine learning
,Information Security and image processing.
Prof.Dr.S.S.Sane Vice Principal, Professor & Head of Dept. of
Computer Engineering, K K Wagh Institute of Engineering
Education & Research, Nashik
Page 164