| Summary: In the scope of the thesis, the problem of image video forgery detection has been examined and the new image/video forgery detection algorithms has been proposed in order to solve these problems. The use of Colour Moments together with Colour Layout Descriptor and Discrete Cosinus Transformation together with Local Binary Pattern in feature extraction from image blocks for detection of image forgery has greatly improved detection accuracy compared to other methods in the literature. The use of image compression history with Benford s generalized law to dynamically determine the threshold value independently of the image video content in the block similarity research have been credited with the literature as a new approach. Keypoint based passive image authentication methods have been improved by detecting forgery in smooth regions through texture extraction prior to the key point extraction in the image. In detection the forgery intra frames, the higher success rate and the low running time were obtained in comparison with the other methods in the literature by using the Local Difference Binary and Bag of Words approaches in feature extraction from the frames. Apart from this, not only Frame Duplication forgery but also Frame Mirroring forgery which is not considered in the literature has also been detected by extracting binary features from the frames. In the thesis, besides Frame Duplication forgery on video, detection of blurred edges in video frames and region based forgery have also been identified.
Keywords: Passive methods, Copy-move forgery, Frame duplication forgery, Region based forgery. |
