Student:	Eda Sena ERDÖL
Supervisor:	Doç. Dr. Güzin ULUTAŞ
Department:	Bilgisayar Mühendisliği
Institution:	Graduate School of Natural and Applied Sciences
University:	Karadeniz Technical University Turkey

Title of the Thesis:	PATTERN BASED IMAGE FORGERY DETECTION BY THE AID OF KEYPOINT AND PATCHMATCH METHODS
Level:	M.Sc.
Acceptance Date:	13/6/2019
Number of Pages:	82
Registration Number:	i3565

Summary:

Processing and editing operations on digital images become fast and common with the

aid of technological improvements. In addition to the good use of technology, it is also possible

to use maliciously, such as editing reality of images. The most common method of vicious

editing is the copy-move forgery. Copy-move forgery can be used to cover objects (i.e covering

the image of the crime device or the perpetrator of the crime scene in the crime scene images)

or increasing the number of objects desired (i.e increasing the number of missiles and missile

launchers in the air attack image). In this thesis, two studies which are proposed for the detection

of copy-paste forgery in important areas such as military, medical, public and law will be

explained. Moreover, Copy-move forgery detection methods are used in hyperspectral satellite

image data sets with dense and important information which are not found in literature. The

studies carried out within the scope of the thesis were compared with the existing studies in the

literature and the resistance to attacks was tested. It has been reported with the experimental

results that the accuracy of the proposed methods is higher than conventional methods and

resistant to attacks.

Key Words: Digital Image Processing, Copy-Move Forgery, Forged Image Detection,

LIOP, PatchMatch, RINBP, LBP, SIFT, RANSAC, Hyperspectral Image.