الفهرس 
Chapter 1 IntroductionOnly 14 pages are availabe for public view
 |  | from | 122 |  |  |
| | | from | 122 | | |
Abstract
Digital images and their applications gained a huge interest in several fields. Image forgeries are applied to give the digital images other meanings or to deceive the viewers. Image forgeries appear in many cases such as cybercrimes, military and intelligence deception, electronic signature forgery, the evidences in courts, electronic documents modifications, social media, or defamation of important characters. There are two types of digital image authentication. The first type is active authentication, which uses digital signature and image watermarks. These techniques have certain constraints such as knowing the content of the digital image. They need special equipment like cameras and development software. The second type is passive authentication, which is used to detect digital image forgeries represented in image cloning, image splicing, image resampling, image retouching, and image morphing. Passive authentication has an advantage that it does not need any previous knowledge of the image content to detect the forgery. A copy-move forgery is a type of image tampering that is created by copying a part of the image and pasting it on another part of the same image to perniciously hide or clone certain regions. Copy-move forgery is the most famous type, and it is widespread in all image forgeries. Copy-move forgery is easy to perform and the forged part has the same properties of the whole image that makes it difficult to detect. There are many algorithms used to detect copy-move forgery attacks depending on different techniques. The thesis covers different directions of Copy-Move Forgery Detection (CMFD) and gives a wide coverage of earlier CMFD algorithms and techniques. The proposed approaches in this thesis are aimed to enhance the existing CMFD algorithms in addition to build more efficient, fast, and robust algorithms.
Scale Invariant Features Transform (SIFT) algorithm is used strongly to detect copy move forgeries due to its efficiency in digital image analysis. SIFT algorithm extracts image features, which are invariant to geometrical transformations such as scaling, translation, and rotation. These features are used in performing the
IV
matching between different views of a scene or an object. The presented approach enhances the efficiency of using SIFT algorithm in detecting copy move forgery by two ways. Firstly, it enhances the image itself by applying different types of digital filters to reinforce the image features giving the ability to detect forgeries. Secondly, the matching strategy is adapted based on a new thresholding approach to increase the true positive rate and decrease the false positive rate. The proposed approach is assessed on the four popular and challenged datasets MICC-F220, MICC-F2000, MICC-F600, and SATs-130. Results show that the proposed approach gives better results compared with traditional copy-move detection approaches. In addition, it gives better stability and reliability to different copy-move forgery conditions.
Two stages object recognition based CMFD algorithm is a promising approach that represents a new methodology for CMFD in digital images. The proposed algorithm is performed in two successive stages; matching stage and refinement stage. In the matching stage, close morphological operation and Connected Component Labeling (CCL) are used to segment the target image into different objects. The Speeded Up Robust Features (SURF) are extracted from each object and used to build an object catalog. The objects in the catalog are compared to each other, and matched objects are determined. If matched objects exist, the image is categorized as forged image. Otherwise, it is categorized as original image. The refinement stage, on the other hand, is implemented to ensure the originality of the target image. Thus, the candidate image that is classified as original is fed into the refinement stage to certify its originality. In this stage, close and open morphological operations as well as CCL are utilized to obtain the various objects in the image. Afterward, the SURF features are extracted from each object and used to build a new object catalog. The match between the objects in this catalog is obtained. If similar objects are found, the candidate image is classified as forged. Otherwise, the image is categorized as original. Also, the proposed approach is tested under the well-known challenged datasets MICC-F220, MICC-F2000, MICC-F600, and SATs-130. The results demonstrate the capability and robustness of the proposed technique in detecting the copy-move
V
forgery under different geometrical attacks. Furthermore, the outcomes show that the suggested technique outperforms the previous CMFD methods in terms of accuracy and execution time. Another innovative CMFD technique for automatic detection of copy-move forgery based on deep learning approaches is proposed. A Convolutional Neural Network (CNN) is specifically designed for CMFD application. The CNN is exploited to learn hierarchical feature representations from input images, which are used for detecting the forged and original images. The extensive experiments demonstrate that the proposed deep CMFD algorithm outperforms the traditional CMFD systems by a considerable margin on the three publicly accessible datasets, MICC-F220, MICC-F2000, and MICC-F600. Furthermore, the three datasets are incorporated and joined to the SATs-130 dataset to form new datasets combination. An accuracy of 100% has been achieved for the four datasets. This proves the robustness of the proposed technique against diverse known attacks.