Development of Copy-Move Forgery Detection Scheme

Abstract

Digital image forgery is quite an old issue in image forensic. We tend to live in digital era and its easy for anyone to access the data, process it and at the end share that data. This poses a major security challenge for institutions working in the field of image forensic. It’s difficult for an individual to differentiate between the original and the forged picture because different types of algorithm and software are used to manipulate the picture; some of which include Photoshop, Corel Draw etc. Today with the rise of internet and image processing software packages, a digital image can be effortlessly forged. Its very difficult for human vision to recognize that whether the image is forged or original. This decreases the reliability of digital image analysis. It is therefore imperetive to create the better forgery detection algorithms, which will check the authenticity of digital images by keeping in mind that images can be used as an evidence as a part of medical record, in court of law, used as financial documents and as news items. Major objective of image forensic is to detect the forgery in an image. Image forgery technique is getting popularity with each passing day and have acceptance in different areas such as IT, medical image, forensic investigation, journalism and intelligence services etc. It is becoming a challenge to secure the authentic data in e-Government services and paperless places where the data stored in digital format because it’s easy to manipulate the files, document, and image and voice data. Such challenges triggered a good interest among researchers in developing new techniques for detecting forged pictures. Copy-move forgery is used with the intent of replicating an object or hiding an undesired image. It can be achieved when content is copied and pasted within the same image. For localization and detection of the digital forged image, different techniques have been proposed over the last few decades. Among them, passive techniques are widely used in literature. The passive technique does not require any previous information for authentication. It specifically detects those changes through which forgery can happen in an image. Copy-move forgery detection can be achieved by using block-based and key-point based method. The detection process involves pre-processing (to organize an image for further evaluation), keypoint or block-based method, feature extraction (through which suitable features of interest are detected from an image), feature matching (based on similarity between two features) and post-processing (used to reduce the false alarm and keeping those matches having familiar behavior). Key-point based methods work on pixels of small sets and are much faster than dense matching methods. In feature extraction phase, the key-point based method does not require to divide an image into blocks, rather it works on an entire image and the results can be achieved by applying different feature matching techniques such as SURF and SIFT. To analyze the performance of forgery detection, F-measures along with the precision and recall are used. The proposed method used the SURF for feature detection and extraction, used the kNN matching algorithm for matching purpose and at the end segmentation is used to produce the forged region. And the proposed method aims is to improve the robustness and accuracy than state-of-the-art techniques.