A Technique to Detect Copy-Move Forgery using Enhanced SURF

Abstract

The field of digital imaging has advanced in recent years with increase of various digital gadgets and applications associated to it. With easily availability of image editing softwares which are either free of cost or budget friendly, image can be effortlessly tampered by the means of making forgery. This has led to increase in crime related to various image processing and computer vision applications. To combat with such forgeries digital forensic provide scientific techniques to identify whether image is original or forged. The proposed work implemented a image forgery check system based on SURF features. This is a pixel based technique where after preprocessing the images, relevant features are extracted and compared with a defined estimated threshold value. Based on the demonstrated results it is decided whether the image has been forged or not and if it is, then the area where tampering has been done is displayed as a forged part. The proposed algorithm is tested using open source CASIA image dataset. Also, the presented result shows that SURF feature based authentication provide forgery detection accuracy of 97%. The results are compared with other techniques in similar domain to prove the novelty of the work.

Convolutional neural network-based techniques and error level analysis for image tamper detection

“…The proposed model with augmented input achieved an accuracy of 98%. Srivastava et al [4] proposed a pixel-based technique for detecting image tampering using speed up robust feature (SURF) features. After preprocessing, significant features are compared against a threshold value, and if altered, the tampered portion is identified.…”