IJSRD - International Journal for Scientific Research & Development| Vol.

4, Issue 12, 2017 | ISSN (online): 2321-0613

Survey on Fake Image Detection Using Image Processing

Ms. Shalet Xavier1 Ms.Steffy Francis2 Ms.Vidhu Valsan A3 Ms.Sheethal MS4
1,2,3
Student 4Assistant Professor
1,2,3,4
Department of Computer Science & Engineering
1,2,3,4
Sahrdaya College of Engineering and Technology, Kodakara, Thrissur, Kerala, India, Pin: 680684
Abstract— To ensure the actual presence of a real legitimate easier, hence there is essential to detect the forgeries in the
trait in contrast to a fake self-manufactured synthetic or images. There are many types of forgeries such as
reconstructed image is a significant problem in morphing, copy move, compositing, retouching, etc. The
authentication, which requires the development of new and image compositing is more popular one. There are some
efficient protection measures. The paper mainly deals with possible methodologies for identifying whether the image is
the detection of morphed images. In this digital world we manipulated or not. Image can be authenticated by Digital
come across many image processing software that produce watermarking. Motivated by the fact that 2/3 color samples
doctored Images with high sophistication, which are of a common photo are reconstructed by a demosaicing
manipulated in such a way that the tampering is not easily technique consisting of only a few formulas and this large
visible to naked eye. There is a need for developing population of reconstructed samples provides a good basis
techniques to distinguish the original images from the of reliable statistical characterization of the applied
manipulated ones, the genuine ones from the doctored ones. demosaicing technique. In this paper, we aim to accurately
There are number of ways of tampering an Image, such as estimate the demosaicing formulas for diversified
splicing two different images together, removal of objects demosaicing algorithms. Our proposed framework employs
from the image, addition of objects in the image, change of a partial second-order image derivative correlation model,
appearance of objects in the image or resizing the image. In which detects both the intra-color channel and the cross-
this paper, we present a software-based fake detection channel demosaicing correlation. A reverse classification
method that can be used in systems to detect different types scheme is incorporated to precisely classify the demosaiced
of fraudulent activities. This Image Morphing detection samples into small categories, which best reveal the original
technique detects traces of digital tampering and demosaicing grouping. Compared with an existing method,
implemented using image processing with Demosaicing our estimated demosaicing formulas regenerate the
Algorithms. The objective of the proposed system is to demosaiced samples from the sensor samples with
enhance the security of image frameworks, by adding significantly improved accuracy. Our reduced set of
liveness assessment in a fast, user-friendly, and non- demosaicing features also perform significantly better in
intrusive manner, through the use of image quality identification of 16 demosaicing algorithms in the presence
assessment. And also it helps to differentiate the original of common camera post-demosaicing processes. For real
and manipulated images and to trace the area morphed. applications including camera and RAW-tool identification,
Key words: Image Detection, demosaicing technique large-scale tests show that our proposed features achieve
nearly perfect identification performances based on the
I. INTRODUCTION cropped image blocks.
In this digital world we come across many image processing
software that produce doctored Images with high II. LITERATURE REVIEW
sophistication, which are manipulated in such a way that the A camera is a trustworthy device and photos traditionally
tampering is not easily visible to naked eye.Image morphing imply truth. Nowadays, this has been severely challenged.
has been the subject of much attention in recent years. There The advances of digital technology have given birth to
are various powerful tools available in market for Image numerous low-cost yet powerful tools which enable easy
morphing. Morphing technique continues to advance and image creation, modification and distribution. Consequently,
many programs can automatically morph images that image forgery becomes commonplace in Internet and other
correspond closely enough with relatively little instruction mass media. This has brought up new challenges concerning
from the user. This has led to the use of morphing authenticity and integrity of digital images. In recent years,
techniques to create slow-motion effects where none existed passive image forensics has become a booming research
in the original film or video footage by morphing between area to mainly address photo-authentication related
each individual frame by the technology called optical flow challenges such as image source identification, tampering
technology. Morphing has also seen as a transition technique discovery and steganalysis. Compared to active image
between one scene to another in television shows, even if forensic methods which require information hiding (e.g.
the contents of the two images are entirely different watermarking and steganography), passive image forensics
Digital images are playing every important role in are based on detection of intrinsic image regularities or tell-
our daily life; the digital images are omnipresent right from tale artifacts leftover due to specific tampering operations.
the cover pages of journals, newspapers, magazines etc. to Since available digital images usually do not carry a
evidences in court rooms, teaching aids etc. Images are used watermark, much wider applications can be expected for
everywhere either as a personal memory evidences or for passive image forensic approach . The different regularities
official purposes .Recently, the low cost camera, are associated with different origins and their detections are
sophisticated high end image processing, computer graphics useful in various forensic tasks. For instance, the work in
software, made editing and manipulated images become extracts photo-response non-uniformity (PRNU) sensor

All rights reserved by www.ijsrd.com 266

 Survey on Fake Image Detection Using Image Processing
(IJSRD/Vol. 4/Issue 12/2017/074)

noises for individual camera identification and tampering In this digital world we come across many image
detection. The work in estimates the demosaicing processing software that produce doctored Images with high
parameters for non-intrusive component forensic analysis on sophistication, which are manipulated in such a way that the
different camera models. Our proposed framework employs tampering is not easily visible to naked eye. The authenticity
a partial second-order image derivative correlation model, of a digital image has become a challenging task due to the
which detects both the intra-color channel and the cross- various tools present in the photo editing software packages.
channel demosaicing correlation. A reverse classification There are number of ways of tampering an Image, such as
scheme is incorporated to precisely classify the demosaiced splicing two different images together, removal of objects
samples into small categories, which best reveal the original from the image, addition of objects in the image, change of
demosaicing grouping. The simulation results show that our appearance of objects in the image or resizing the image.
proposed demosaicing features confidently outperform 2 This Image Morphing detection technique detects traces of
existing demosaicing detection methods in identifying 16 digital tampering in the complete absence of any form of
demosaicing algorithms in the presence of various common digital watermark or signature and is therefore referred as
post-demosaicing processes. Our proposed features are also passive. So there is a need for developing techniques to
highly effective in distinguishing different post-processes distinguish the original images from the manipulated ones,
and are more sensitive to small scenery variations.[1] the genuine ones from the doctored ones. In this paper we
Image morphing techniques can generate describe a novel approach for detecting Image morphing.
compelling 2D transitions between images. However, The new scheme is designed to detect any changes to a
differences in object pose or viewpoint often cause signal. We recognize that images from digital cameras
unnatural distortions in image morphs that are difficult to contain traces of re-sampling as a result of using a color
correct manually. Demosaicing of Color Filter Array filter array with demosaicing algorithms. Our results show
Captured Images Using Gradient Edge Detection Masks and that the proposed scheme has a good accuracy in locating
Adaptive Heterogeneity-Projection based on spectral-spatial tampered pixels.[4]
correlation, a novel adaptive heterogeneity-projection with The multimedia applications are rapidly increasing.
proper mask size for each pixel is presented. Combining the It is essential to ensure the authenticity of multimedia
extracted gradient/edge information and the adaptive components. The image is one of the integrated components
heterogeneity-projection values, a new edge-sensing of the multimedia. In this paper ,we designing a model
demosaicing algorithm is presented. Based on 24 popular based on customized filter mask to ensure the authenticity of
testing images, experimental results demonstrated that our image that means the image forgery detection based on
proposed high-quality demosaicing algorithm has the best customized filter mask. We have satisfactory results for our
image quality performance when compared with several dataset. Image compositing is most popular image forgery.
recently published algorithms. It presents a new approach to The figure 1 shows the creation of image compositing. The
extract more accurate gradient information on mosaic photo compositing is the result of cutting and joining a two
images directly. In what follows, the luminance estimation or more photographs with seamless transition without
technique for mosaic images is first introduced. Then, leaving any visual clues about the joining from other
combining the luminance estimation technique and Sobel photographs. The image compositing is also known as
operator , our proposed new approach to extract more photomontage and image splicing. The image compositing
gradient information on mosaic images is presented. [2] detection assumes that the image scene authenticity
A novel manipulation detection framework for properties and conditions such as illuminations, object
image patches using a fusion procedure, called FusionBoost, surface properties, shadow, noise, inter-reflections.
in conjunction with accurately detected derivative Perspective and projective views etc. are rare and difficult
correlation features. By first dividing all demosaiced match in a composite image. The image composite detection
samples of a color image into a number of categories, we techniques are able to detect the above inconsistent
estimate their underlying demosaicing formulas based on properties in different parts of the same images, the image
partial derivative correlation models and extract several edges, boundaries and colors, and image qualities may be
types of derivative correlation features. The features are affected by image compositing. The image forgery can be
organized into small subsets according to both the identified by the specific patterns relating to image attributes
demosaicing category and the feature type. For each subset, which disturbed by the forgery operations. Particularly
we train a lightweight manipulation detector using image composting is created by the two or more images
probabilistic support vector machines. FusionBoost is then sources, naturally all the different images are taken from the
proposed to learn the weights of an ensemble detector for different devices and at different world view conditions. The
achieving the minimum error rate. By applying the ensemble host image conditions are expected to reflect in image
detector on cropped photo patches from different image portions of the altered images. The abrupt and unnatural
sources, large-scale experiments show that our proposed luminance levels, colors and edges are able to detect the
method achieves low average detection error rates of 2.0% image forgeries.[5]
to 4.3% in simultaneously detecting a large variety of Image processing is often not necessary for image
common manipulation attempts for image patches from manipulation detection. For instance, a picture supposed to
several different source models. Our framework shows good be taken in India that shows the China monument in the
learning efficiency for highly imbalanced tasks. In several background will be suspect by inspection. Detection of
patch-based detection examples, we demonstrate the incongruous textural features, however, may require
efficacy of the proposed method in detecting image substantial image processing. The manipulation are
manipulations on local patches.[3] sometimes not noticeable by human eye, they do affect the

All rights reserved by www.ijsrd.com 267

 Survey on Fake Image Detection Using Image Processing
(IJSRD/Vol. 4/Issue 12/2017/074)

statistics of the image, because of detection of tampering is forgeries, and then save the result in the NTSC format on an
possible. Thus it becomes very important to develop ordinary videotape. Digital watermarks have been proposed
efficient techniques which may detect these forgeries which as a means for fragile authentication, content authentication,
are addition of an object in image, removal of object from detection of tampering, localization of changes, and
image and change of appearance of the object in image. The recovery of original content. While digital watermarks can
process of Image morphing detection can involve several provide useful information about the image integrity and its
works. These work include, but are not limited to, processing history, the watermark must be present in the
evaluation of image structure issues include discovery of image before the tampering occurs.In a Copy-Move forgery,
artifacts consistent with image manipulation or degradation, a part of the image itself is copied and pasted into another
metadata analysis, and indications of provenance and Image part of the same image. Textured areas, such as grass,
content issues include continuity issues, evidence of foliage, or fabric with irregular patterns, are ideal for this
manipulation, evidence of staging, and misplacing. There purpose because the copied areas will likely blend with the
are several possible techniques for detecting manipulation in background and the human eye cannot easily understand any
the source of a digital image. Image can be authenticated by suspicious artifacts. Because the copied parts come from
Digital watermarking. Digital watermarking has two classes the same image, its noise component, color palette, dynamic
of watermarks, fragile and robust. Robust watermarks range, and most other important properties will be
techniques are designed to be detected even after attempts compatible with the rest of the image and thus will not be
are made to remove them. Fragile watermarks techniques detectable using methods that look for incompatibilities in
are used for authentication purposes and are capable of statistical measures in different parts of the image.[7]
detecting even minute changes of the watermarked content.
But, neither type of watermark is ideal when considering
"information preserving transformations (such as
compression) which keep the meaning or expression of the
content and "information altering" transformations (such as
feature replacement) which modify the expression of the
content.[6]

Fig. 2: Distinguish between images containing noise with

large energy across the frequency spectrum and true
demosaicing signals generated by our algorithm. Bottom
Left: The signal , which represents an calculation of the
variance along each image diagonal. Bottom Right: The
spectrum of , represents the characteristic peak at.
Digital Photo images are everywhere, on the covers
Fig. 1: When demosaicing is performed with linear of magazines, in newspapers ,in courtrooms ,and all over the
interpolation, the original green pixels have higher variance internet. In this paper we propose methodologies to identify
than the interpolated green pixels. The spatial pattern of such unbelievable photo images and also identify forgerd
variances is the basis for detecting the presence of region by given only the forgerd image. Formats are
demosaicing. The green photosites pixel values in the Bayer additive tag for every file system and contents are relatively
array are IID with variance σ2, the above image shows the expressed with extension based on most popular digital
variance from which each pixel value is drawn. camera uses JPEG and other image formats like png,
Nowadays Digital images are manipulated due to bmp,etc. We have designed algorithm to find the abnormal
the availability of the image processing and editing anomalies and identify the forgerd regions. Different
softwares. Using these softwares we can add or remove keywords are using this purpose such as Digital image,
important features from an image without leaving any Forgery region, Copy-move Copy-create. Composition is
obvious traces of tampering. In this paper we discuss on the experimented by the Photographers, i.e., combining multiple
detection of a special type of digital forgery that is copy- images into one. Digital images offer many attributes for
move attack in which part of the image is copied and pasted tamper detection algorithm to take advantage of specifically
and also investigate the problem and describing an efficient the color and brightness of individual pixels as well as an
and reliable detection method . The method may image’s resolution and format. This paper focuses on images
successfully detect the forgerd part even when the copied saved in the JPEG format. Other fundamental properties of
area is enhanced to merge it with the background and when any digital forgery are used to develop additional detection
the forgerd image is saved in lossy format, such as JPEG. technique such as direction filter, which is used to detect the
To tamper an analogue video, one can easily digitize the forgery region . Photo image forgery is classified in to two
analog video stream, upload it into a computer, perform the

All rights reserved by www.ijsrd.com 268

 Survey on Fake Image Detection Using Image Processing
(IJSRD/Vol. 4/Issue 12/2017/074)

categories. The first class of image forgeries includes

images tampered by copying one area in an image and
pasting it onto another area. It is called as Copy-Move
Forgery or Cloning. The second class of forgeries is copying
and pasting areas from one or more images and pasting on to
an image being forged. [8]

Fig. 4: Flow diagram for Photo Morphing Detection. First

photographs from digital cameras or computer generated
Fig. 3: PRCG Image example.performance curves are
images are given to high pass (HP) filter. Then HP Filter is
generated.
applied, and then the Positional Variance of each diagonal is
Image Forgery is defined as adding or removing
calculated. Then analyzed, indicating the presence of
important features from an image without leaving any
demosaicing in the image.
obvious traces of tampering. Further, it can either be
intrusive (active) or non-intrusive (blind or passive). In
III. IMAGE PROCESSING
active approach, the digital image requires some kind of
pre-processing such as watermark embedded or signatures In imaging science, Image Processing is processing of
are generated at the time of creating the image. images using mathematical operations by using any form
Watermarking is such a method of active tampering of signal processing for which the input is an image, a series
detection, as a security structure is embedded into the image, of images, or a video, such as a photograph or video frame;
but most present imaging devices do not contain any the output of image processing may be either an image or a
watermarking or signature module and that are similar to the set of characteristics or parameters related to the image.
application of active protection. Signature is such a method Most image-processing techniques involve treating the
of active tempering detection, in which signature is image as a two-dimensional signal and applying standard
embedded into the image as a security means. Passive image signal-processing techniques to it. Images are also processed
forensics is usually a great challenge in image processing as three-dimensional signals where the third-dimension
techniques. It includes the concept of Copy-Move Forgery, being time or the z-axis. Image processing usually refers
Retouching and Image Splicing. Copy-Move is a special to digital image processing, but optical and analog image
type of image manipulation technique in which a part of the processing also are possible. This article is about general
image itself is copied and pasted into another part of the techniques that apply to all of them. The acquisition of
same image.Retouching is defined as hanging the image on images is referred to as imaging. Closely related to image
a whole. For example by adding onto brightness, creating processing are computer graphics and computer vision. In
noise, creating clarity onto the base image etc.Image- computer graphics, images are manually made from
splicing is defined as a paste-up produced by sticking physical models of objects, environments, and lighting,
together photographic images. Image splicing is a common instead of being acquired from natural scenes, as in most
type to create a tampered image where a region from one animated movies. Computer vision, on the other hand, is
image is copied and pasted into another image which often considered high-level image processing out of which a
produces composite. Image is called spliced image, cut and software intends to decipher the physical contents of an
join two or more snaps of pictures. The complicated forgery image or a sequence of images
may include some post-processing like blurring, JPEG
compression, etc. that performs the forgery detection very IV. CONCLUSIONS
hard.[9] In this survey article, we have reviewed over nine papers in
the literature of image demosaicing and forgery. The
majority of existing demosaicing algorithms exploit the
spectral correlation by sequential strategies - i.e., the
luminance channel is recovered first and then chrominance
channels are reconstructed based on the full-resolution
luminance image. Spatial adaptation based on local
deterministic or statistical has shown critical to the
performance of various demosaicing techniques. Our
comparative studies with very best demosaicing algorithms

All rights reserved by www.ijsrd.com 269

 Survey on Fake Image Detection Using Image Processing
(IJSRD/Vol. 4/Issue 12/2017/074)

have demonstrated the importance of jointly exploiting International Journal on Computational Sciences &
spatial and spectral correlations especially for images with Applications (IJCSA) Vo2, No.6, December 2012
high-saturation and varying-hue. We have also observed that [9] Hany Farid, “Image Forgery Detection”, IEEE Signal
even ad-hoc fusion by averaging different demosaiced Processing Magazine [16] MARCH 2009
images could lead to further improvement. There are three
directions along which further studies are definitely needed.
First, demosaicing of images with weak spectral correlation
remains a challenging task. Our understanding about the
tradeoff between spatial and spectral correlation is still
primitive though some attention has been paid to this issue
by one group of contributing authors to this session. Last but
not the least, the performance evaluation of demosaicing
algorithms needs more careful investigation.
Complementing the current proposed approach to identify
different image software processing pipelines, additional
features such as sensor noise pattern are needed in
identification of individual cameras of the same model.
Further investigation effort also includes verifying the
effectiveness of our proposed demosaicing features on the
camera models of non-Bayer CFAs.

ACKNOWLEDGMENT
The authors would like to thank the referees for their
valuable comments which helped improve the quality of the
paper greatly.

REFERENCES
[1] Hong CAO, student member, IEEE, and Alex C. KOT,
Fellow, IEEE, “Accurate Detection of Demosaicing
Regularity for Digital Image Forensics”,
[2] Kuo-Liang Chung, Senior Member, IEEE, Wei-Jen
Yang, Wen-Ming Yan, and Chung-Chou Wang,
“Demosaicing of Color Filter Array Captured Images
Using Gradient Edge Detection Masks and Adaptive
Heterogeneity-Projection”, IEEE Transactions On
Image Processing, VOL. 17, NO. 12, DECEMBER
2008
[3] Hong Cao, Member, IEEE, and Alex C. Kot, Fellow,
IEEE, “Manipulation Detection on Image Patches
Using FusionBoost”, IEEE Transactions On
Information Forensics And Security, VOL. 7, NO. 3,
JUNE 2012
[4] Nilesh P Ghatol, Rahul Paigude, Aniket Shirke “Image
Morphing Detection by Locating Tampered Pixels with
Demosaicing Algorithms”, International Journal of
Computer Applications (0975 – 8887) Volume 66–
No.8, March 2013
[5] Shrishail Math , R.C.Tripathi “Images Image composite
detection using customized”, International Journal of
Computer Graphics & Animation (IJCGA) Vol.1, No.3,
October 2011
[6] Mr. Vinayak K. Shingote, Prof. Ruhi Kabara, “Image
Authentication by Detecting Demosaicing ”, IJARCCE
Vol. 4, Issue 5, May 2015
[7] Jessica Fridrich, David Soukal, and Jan Lukáš
Detection of Copy-Move Forgery in Digital Images
“”,SUNY Binghamton, Binghamton, NY 13902-6000
[8] S.Murali1 , Govindraj B. Chittapur2 , Prabhakara H. S 1
and Basavaraj S. Anami3, “Comparison and analysis of
photo image forgery detection techniques” ,

All rights reserved by www.ijsrd.com 270