Please use this identifier to cite or link to this item:
http://hdl.handle.net/10266/4150
Title: | Feature Development based on CFA Artifacts for Image Forgery Detection |
Authors: | Amneet, Singh |
Supervisor: | Singh, Kulbir |
Keywords: | CFA atrifacts;Forgery Detection |
Issue Date: | 24-Aug-2016 |
Abstract: | Nowadays, a lot of information is being shared in the form of digital images. Along with the actual images, there are some cases when misleading information is being shared in the form of tampered images. Thus, there is a need to curb these incidents so that the reliable information can reach the masses. These issues can be handled through ‘Digital Image Forensics’. The basic ideology behind Digital Image Forensics is to reconstruct the image history, also known as ‘Digital Image Life Cycle’. It includes all the steps that are involved to make the real scene a digital image. It starts from the real scene acquisition and all the in-camera and post-camera processing techniques which leads an image to its final fruition. This has been observed that each process in the digital image life cycle, leaves a distinct trace or fingerprint that can be analyzed while reconstructing the image history. The inconsistencies in the fingerprint can be sufficient to expose the forgery that has been introduced in the image. The motivation of this report is to analyze the fingerprints left during the interpolation process. The image interpolation process is to assign the values to the pixels that were not gathered when the light was filtered by the Color Filter Array (CFA). The interpolation technique employs a demosaicing algorithm that assign the values to the missing pixels. In doing so, the algorithm leaves a certain trace, known as CFA artifacts, whose presence can lead to the identification of the demosaicing algorithm. The inclusion of forgery in the image would conclusively remove or alter the structure of the present CFA artifacts. A scheme is proposed which finds out the probability of the presence or absence of the CFA artifacts in a 2x2 blocks of image. The probability is based on the standard deviation as well as skewness of the prediction error that is calculated between the acquired and interpolated pixels. This probability is then used to create a forgery map that depicts the region(s) that has been tampered with. |
Description: | Master of Engineering-Wireless Communication |
URI: | http://hdl.handle.net/10266/4150 |
Appears in Collections: | Masters Theses@ECED |
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.