Search

In this paper, prominent keypoint based features are compared in order to analyze their reliability and efficiency against forgery detection. Four features specifically SURF, KAZE, Harris corner points and BRISK features are used individually on a set of images. The method includes four phases: Image pre-processing, keypoint detection, feature vector description and feature vector matching. In feature matching, MaxRatio has been chosen as a varying parameter for calculating values of false positives and false negatives for each feature. MaxRatio defines the ratio for rejecting ambiguous matches of feature descriptors in the images. The optimal threshold value for MaxRatio is calibrated with the help of trade-off between detection accuracy and false positive ratio. The changes in false negative ratio and false positive ratio are picturized in order to find out optimal threshold for detection accuracy. ROC curves are also plotted for each feature at different values of MaxRatio and area under the ROC curves are calculated. The experiments are performed on two benchmark datasets, namely CASIA version 2.0 and MICC-F600. It has been perceived from experimental outcomes that KAZE features gave best values for all the performance metrics namely accuracy, precision, area under the ROC curve and F1-score with little compromise in time complexity, whereas Harris corner points gave the worst results as compared to rest of the features. Further, in order to improve the execution time, the computation of non-linear scale space process in KAZE can be simplified and GPU programming for real-time performance may also be used.

Image manipulation has eroded our trust of digital images, with more subtle forgery methods posing an ever-increasing challenge to the integrity of images and their authenticity. Over recent years,...

Image forgery has become the main concern of the society over the past years due to an increase in the number of fraudulent image manipulations. Therefore, it has become a necessity to build an effective forgery detection method to check the integrity and authenticity of images. In this paper, an unsupervised method for classification of forged images and a supervised method for localization of forged regions is proposed. The methods use GLCM, a textural feature descriptor that uses pixel correlation in image to extract features from it. Further, for feature matching, Euclidean and Hellinger distance is used and finally, the localization of tampered region is performed. Euclidean and Hellinger distance is calculated separately on each image for comparison of the results. The experimental results showed that Hellinger outperformed Euclidean distance in feature matching and gave better localization results. In another approach, a bounding box is taken from the image to get the testing area and the training area. SVM classifier is used for classification of forged blocks and the authentic blocks in the image. The experimentation is performed on CASIA v1.0 dataset.

Image forgery has become the main concern of the society over the past many years due to an increase in the number of false cases of image manipulation leaving a severe impact on the life of a victim. Therefore, it has become necessary to solidly build forgery detection method to overcome image forgery problem. This necessity led to research in this field of image processing and provide detection techniques to check the integrity of images. In this paper, a method for localization of such tampered region is proposed. The proposed method uses unsupervised learning approach to localize the region of image and find whether the region is forged or not. It uses GLCM, a textural feature descriptor that uses pixel correlation in image to extract features from it. Further, for feature matching, Euclidean and Hellinger distance is used and finally, the localization of tampered region is performed. Euclidean and Hellinger distance is calculated separately on each image for comparison of the results. The experimental results showed that Hellinger outperformed Euclidean distance in feature matching and gave better localization results.

Forgery detection techniques are required to verify the authenticity of the digital images. The additional noise is the most general way to hide the traces of the tampering done to the image. Original images which do not undergo any alterations are supposed to have a consistency in noise variation. If the image is forged, the noise no longer remains consistent throughout the image. In this paper, a method is proposed to detect the forgery based upon noise estimation and hog feature extraction. The image is first converted to YIQ colorspace, and then the block segmentation is performed on Y component of the YIQ image. Noise is estimated using PCA and hog features are extracted from each block of the image. An unsupervised clustering method is used to cluster the blocks of the image. The experimental results show that the proposed technique detects forged images more effectively as compared to previous method based only on noise estimation.

Copy-move forgery is the most commonly performed type of forgery. For copy-move forgery detection, block based and keypoint based methods are available. In this paper, keypoint based features are chosen as they are computationally less complex as compared to block based features. Four different keypoint based feature extraction algorithms i.e. SURF, KAZE, Harris corner points and BRISK are analyzed in order to check their efficiency for copy-move forgery detection. The method used involves four basic stages: Image pre-processing, interest point detector, feature vector description and feature matching. The results are compared on the basis of accuracy, fl-score and precision which are calculated using a threshold parameter for matching algorithm. It has been concluded that KAZE features give best results in all performance metrics and Harris corner points turn out to be unsuitable for copy move forgery detection due to the fact that Harris corner points are not scale invariant and detect only corners instead of edges.

In today’s modern era, digital images have noteworthy significance because they have become a leading source of information dissemination. However, the images are being manipulated and tampered. The image manipulation is as old as images itself. The history of modifying images dates back to the 1860s’, though it has become very popular in recent times due to the availability of various open source software available freely over the internet. Such software is responsible for eroding our trust on the integrity of the visual imagery. In this paper, a comprehensive survey of various image forgeries, its types and the currently used techniques to detect such forgeries is presented. The review delivers the downsides of various controversial forgeries that have happened in the history. It provides the taxonomy of various forgeries in digital images and a redefined the classification of forgery detection methods. It also highlights the pros and cons of forgery detection methods currently in use and directs path towards challenges for further research.

Images are the basic media of various information delivering units like our social media, news channel etc. However images are losing credibility because of the improvement in image altering software, so it is very easy for a forger to forge an image. It creates confusion for the receiver to believe on the image. So there is a need for the development of new techniques to catch forgery. Researchers are working to get a fool proof technique to detect forgery. There are various techniques to detect forgery. Here we will analyse the light direction. Every photograph may have a different incident light direction for which objects illuminate and are visible in the image. Analysing the light source direction exactly from a 2D image is quite abnormal in theory. Hence this work proposes a promising technique that utilises the shape information to detect the light direction in 3D vector of some known shapes.