Enhanced dominated completed median ternary pattern for detection of copy-move image forgery against post-processing operations
Copy-move forgery which is an act of copying an object and pasting it on another location of the same image is one of the most common types of tampering techniques to manipulate image content. Besides, most of the images are also being tampered by post-processing operations such as JPEG compression,...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2020
|
| Subjects: | |
| Online Access: | http://eprints.utm.my/id/eprint/96437/1/RafidahMuhammadMFABU2020.pdf.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Copy-move forgery which is an act of copying an object and pasting it on another location of the same image is one of the most common types of tampering techniques to manipulate image content. Besides, most of the images are also being tampered by post-processing operations such as JPEG compression, contrast adjustment, brightness change, colour reduction, noise addition, and blurring before pasted, makes it more challenging to detect, as found in the standard dataset CoMoFoD. In multimedia forensics, many efforts have been undertaken to detect whether an image is pristine or manipulated by proposing various techniques to improve the robustness of these detection methods. However, researchers continue to face challenges in detecting tampered region with the presence of all these post-processing attacks in a copy-move image forgery and relatively few methods were attempted to address them. The main objective of this research is to design and develop an improved descriptor with features invariance to post-processing operations for copy-move forgery detection. Generally, image processing steps consist of four main steps which are pre-processing, feature extraction, block matching and evaluation of results. In this process, an improvement is made by employing sign operator for feature extraction with mean (robust) as threshold to extract invariance feature vectors against post-processing attacks from each block of image. An Euclidean distance is employed to filter out the weak features and obtain rough suspected matches. The results obtained were very encouraging with Correct Detection Rate (CDR) of more than 99% achieved for the normal tampered images, while the ones with post-processing operations fluctuated between 91% and 99.7%. The results have proven that the proposed copy-move forgery detection performed better than the existing techniques. |
|---|