Recently, tampering in digital images considered the main challenge in image forensic analyses. Hence, copying a part and pasting it in the same image became the most crucial action in image forgery. It threatens the integrity and authenticity of image ownership. The intruder utilizes the development tools of image processing programs to make the forged image the same as the authentic one and strict for detection. This work as copy-move forgery detection (CMFD) manipulates the problem of a few key points in small-size and homog- enous digital images by adopting a merging scheme to detects sufficient key points by using a linear scale-space detector based on Speedup Robust Feature(SURF) and curvature scale space detector based on Maximally Stable Extremal Region (MSER). Afterward, these key points are described distinctively by extract unique vectors, and matching these vectors to find duplicated regions. In the post-processing stage, we propose new filters, the first is called Parallel Filter and the other called Distance Ratio Filter. These Filters aim to remove false-positive results and boost true positive results thus improving the accuracy of the detection scheme. The experimental results on standard data sets (MICC 220, F8 Multi) show that CMFD is efficient and insensitive against simple and combination post-processing attacks like photometric and geometric transformations. Also, it is invariant against non- uniform transformation like (skew, wrap), and detects multi cloning efficiently with a high true-positive ratio (TPR=98.5) and low false-positive ratio (FPR=4).

Keywords