Copy-Rotate-Move Forgery Detection using Complex Wavelet Transform and Local Binary Pattern

Image region duplication is a type of modification attack on digital images, where a region of the digital image is copied and pasted onto another location within itself. In the recent decade, various techniques have been presented for image region duplication detection based on investigation of matching pixel blocks. However, such forgery detection becomes a challenge when accompanied with geometrical operation such as rescaling, rotation on the forged portions of the image. In this work, we provide a solution of image region duplication detection in a rotation-invariant way. The proposed method is based on Complex Wavelet Transform, utilizes Local Binary Pattern features, and is capable of detecting arbitrary degrees of rotation, unlike existing techniques. Mean-Variance method is used in order to optimize false matches resulting from naturally similar blocks in an image. According to our experimental analysis forgery detection efficiency of the presented methodology is higher, and its false positive rate is lower as compared to the existing methods.