Please use this identifier to cite or link to this item:
|Title:||Performance Enhancement of Copy-Move Forgery Detection by Using Shi Tomasi-Surf Detector and Surf-PSO Algorithm|
Kansal, Ankush (Guide)
|Keywords:||FORGERY;SHI TOMASI DETECTOR;SURF DETECTOR;SURF-PSO|
|Abstract:||In this era of digital computing, digital images are one of the principal means of communication. With the tremendous utilization of digital images and the accessibility of capable image editing tools such as Adobe Photoshop, GNU Image Manipulation Program “GIMP”, it turns out to be very easy to control or alter the digital images and create forgeries without leaving any visual pieces of information. Accordingly, digital images have lost their trust and it has become important to check the originality of content when they are utilized as a part of some basic situations like criminal investigation etc. In this way, Digital Image Forensics emerged as a research field that plans to check the authenticity and integrity of digital images. Many block-based and key-points based techniques have been proposed so far to detect Copy-Move Forgery in digital images; where a part of an image is copied and pasted elsewhere within the same image. Among all the available forgery detection techniques, the most computationally effective and robust technique is Speeded-Up Robust Transform (SURF) framework. The major advantage of SURF over other prevailing forgery detection techniques is that SURF is fastest among other techniques, has less computational complexity, lesser length of descriptor vector etc. However, the key-points detected by SURF detector are not able to detect forgeries in regions with inconspicuous changes. Therefore, to overcome this issue and further enhance the detection accuracy, in this thesis, SURF detection technique is replaced by Shi-Tomasi detector. Finally, SURF descriptors are deployed to give unique identity to each key-point for matching process. The proposed algorithm has improved the precision and recall from 91.49% and 89.59% to 97.87% and 93.75% respectively as compared to conventional SURF based detection. Also, the detection results of conventional SURF based framework highly depends upon the value of parameters which are mostly determined with human perception. Henceforth; the predetermined parameter values limit the application of copy-move forgery detection since they are not applicable to all the images. Therefore, a novel approach by integrating SURF based detection with particle swarm optimization (SURF-PSO) has been proposed. This utilizes Particle swarm optimization for each image independently to generate optimized value of parameters for forgery detection under SURF framework. In this thesis, the proposed SURF-PSO is applied on five images separately and the results prove that SURF-PSO performs much better than the conventional SURF technique. The precision of the SURF-PSO is 1 for three images while it is 0 or 0.5 for conventional SURF and for a particular image, SURF-PSO detected 1064 matched points while conventional SURF detected only 132.|
|Description:||Master of Engineering -ECE|
|Appears in Collections:||Masters Theses@ECED|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.