Abstarct: In this thesis we propose an efficient and fully automatic method to detect copy -move forgery detection in digital images. Nowadays wide use of the Internet and digital images in it, have made such images sources of information, in courts of law they may be used as testimony to proof or refuse evidences, and also a number of tampered images have been published by newspapers. So determining image originality is a fundamental task. There are several methods to make fake images, but the most common is copy-move forgery, that the forger copies a part(s) of image and pasted it into another part(s) of that image. Many researchers have done beneficial researches on it. Most of them can find only copy-moved forgery. In other words, they can find regions which are only copied and pasted without any changes and are weak against some types of manipulations. Many forgers make some changes on copied regions so that the image sounds more natural. So a good copy-move forgery detection thechnique should be robust to some types of changes such as rotation, scaling, JPEG compression, Gaussian noise addition and gamma correction. Most existing methods do not deal with all these manipulations and often have time complexity. Detection method baxsed on SIFT features is robust to all such modifications, but failed to find flat copied regions. Zernike moments are invariant against rotation, additive white Gaussian noise, JPEG compression, and blurring. They can find flat copied regions too, but sensitive to scaling. So it is clear that using these two features is very proper to detect all copied regions in an image. So our scheme is robust against scaling, rotation, additive white Gaussian noise, JPEG compression, gamma correction and affine transformation and the following improvements have been imposed on it:  Detecting more than one image duplicated region from a single region, while other methods can find only a single copy of a region.  It can distinguish between similar object and regions, from copied objects and regions. So by combining detection methods baxsed on SIFT features and Zernike moments, the proposed method is robust against flat copied region in addition to mentioned modifications. Further more In the case of fraud detection, geometric transformation parameters are determined. Our proposed method is precise and has a reduced computational complexity and time processing.