Region duplication detection based on Harris corner points and step sector statistics

Region duplication is a simple and effective operation for digital image forgeries. The detection of region duplication is very important in digital image forensics. Most existing detection methods for region duplication are based on exhaustive block-matching of image pixels or transform coefficients. They may not be effective when the duplicate regions have gone through some geometrical transformations. In this paper, a novel region duplication detection method that is robust to general geometrical transformations is proposed. Firstly, the Harris corner interest points in an image are detected. Then, an image region description method based on step sector statistics is developed to represent the small circle image region around each Harris point with a feature vector. Finally, the small circle image regions are matched using the best-bin-first algorithm to reveal duplicate regions. Experimental results show that the proposed method can work effectively on the forged images from two image databases, and it is also robust to several geometrical transformations and image degradations.

► An image region duplication forgery detection scheme based on Harris corner points and step sector statistics is proposed.
► Harris corner points are detected as the reference points.
► Step sector statistics are utilized as the feature representation of each point.
► Best-bin-first algorithm is used for feature matching.
► Experimental results demonstrate that the approach is effective.