A Survey on Photo Forgery Detection Methods

. In recent years, digital image forgery detection has become one of the hardest studying area for researchers investigations in the field of information security and image processing. Image forgery detection methods can be divided into two extensive groups such as Active methods and Passive (Blind) methods. Active methods have been used data hiding techniques like watermarking and digital signatures. Passive forensic methods (or Blind) use image statistics or they investigate the attributes of the image to determine the forgeries. Passive detection techniques are also split into three branches; image splicing, image retouching, copy-move. Such image forgery detection methods are focus of this paper.


Introduction
Widespread of the digital cameras and the image editing tools like Adobe Photoshop, Microsoft Paint that gives for people doctored images for the bad aims.Increasing in image forensics has given boosting to special techniques for the detection of changing image.Image Forgery Detection is a new studying area which goals confirm the authenticity of image by collected information their feature.Diverse methods have been evolved to handle with changing image and forgery because of provide the origin of the image.We will evaluate different groups of algorithm.Unfortunately, we have not been able to review some important study here.
Digital image forgery methods can be classified two main categories.One of them active methods and the other passive(blind) methods [2].Active methods must be have preembedded information, such as digital watermarking and steganography.Digital watermarking is a method embedding secret information in the data, can be divided two categories, visible and invisible [1].Steganography's example of digital signatures.
Passive methods can be divided five group.
• Brute Forces [5] • Block-Base Techniques • Keypoint Based Techniques Brute force techniques is based exhaustive search and auto correlation methods [6].Block based techniques use like this algorithms; DCT(Discrete Cosine Transform) [7], PCA(Principle Component Analysis) [8], SVD(Singular Value Decomposition) [8], DWT (Discrete Wavelet Transform) [10].Keypoint based techniques use like this algorithms; SIFT(Scale Invariant feature transform), SURF(Speeded-up Robust features).An Digital forgery example of a newspaper report on Saddam and Bill Clinton photographs is shown in Figure 1.

Basic Steps of the Methods
Although there are a large number of proposed copy-move fraud detection methods, all the methods applies the general flow chart given in Figure 2. As seen in the flow chart, there are two alternatives techniques after pre-processing: block based methods and keypoint based methods.In both methods pre-processing steps are applied.For example; many methods studying on gray level images and so color channels must be combined.
Step 1. Image Pre-Processing; is the first step.Methods change some specific details on the image like image filtering, DCT coefficients, RGB to gray level before the feature extraction step [9].Step 2. Feature Extraction; in the block based method the image is divided into sub-blocks in the form of a rectangle.For each sub-block, the feature vectors are computed.Then similar attribute vectors are matched.On the other hand, in feature based methods, only image regions with high entropy value are determined without any image subdivision.These regions are called "key points" and the feature vectors of these regions subtracted [10].
Step 3. Feature Matching; similarity between two feature is marked for a duplicated image.Some methods uses lexicographic sorting and others best-bin-first search(kd-tree algorithm) in determing the similar features vectors.
Step 4. Filtering; have been committed for decrease the possibility of the false positive matches.There are many distance algorithm with the near intensities.Some research proposed Euclidean distance [11], and someone correlation coefficient [12].
Step 5. Pro-Processing; this is last step of all the methods and its optional.The goal of this last step is to protect only those blocks(or keypoint) that share common feature.When considering a set of mappings for the region, it is expected that the source and target blocks(or keypoint) of these mappings are close to each other.

Fig.2. Image Forgery Detection Flow-Chart 3 Comparison of the Related Work
In this subdivision of manuscript, we supply some experimental solutions to describe the demonstration of the analytic algorithm suggested.
Author(s) Techniques Benefits Drawbacks [13] Coefficient map and threshold.DWT and segmentation A good result for diverse condition, copy-move and transformation Execution time [14] Transform matrix.Segmentation Give preferable result using SIFT

Conclusion
In this paper, a brief survey of the image forgery detection(IFD) methods will be help the researcher in this studying area.We have mostly surveyed publication on IFD between 2013 and 2018.We provided a wide review of up-to-date copy move forgery methods.The methods are classifying two groups; like block base and keypoint based.Although both methods are superior to themselves, there have been more recent publications on block based methods last five years.