Application of image blobs and Illumination component in image tamper detection techniques

Abstract

newline This thesis presents several techniques to detect, localize and estimate newlineparameters of two major types of tamerping digital images: copy-move newlineforgery and image splicing forgery. In the former, a portion of an newlineimage is duplicated, sometimes with geometric transformations, and newlinepasted into another part of the same image. In the latter, a part of a newlinesecond image is pasted into an image. Existing methods in literature newlinegenerally require a number of parameters that are tuned to various newlineimage characteristics for efficient and robust detection of forgeries. Our newlinemotivation is to find general techniques that minimize the number of newlinesuch tunable parameters while overcoming certain other of existing newlinecopy-move and image splicing detection methods. newlineThe problem of photograph manipulation has attracted many researchers newlinein the field of Digital Image Forensics (DIF) that aims at validating newlinethe authenticity of images by identifying the imaging device that captured newlinethe image or detecting the traces of forgeries. Image tampering, newlinealso called image forgery, is the act of image editing and manipulation newlinefor malicious purposes to modify the semantic meaning of the visual newlinemessage. The availability of powerful image editing tools has greatly newlinesimplified the process of malicious manipulation, editing, and creating newlinetampered images even by lay persons. newlineIn this thesis, we propose the use of image blobs in copy-move forgery newlinedetection algorithms. Image blobs are the regions that differ in properties newlinesuch as brightness or color compared to surrounding regions. The newlinegoal of blob detection is to identify and mark these regions. As blobs newlineare detected in scale-space, we demonstrate through our experiments newlineiv newlineon standard benchmark datasets that they present several advantages newlineover image blocks and segments in copy-move forgery detection. newlineImage splicing merges the portions of images from different sources newlineinto one composite image. This introduces various artifacts such as newlinesharp transitions around the pasted area and abnormal transient at newlinethe s