Abstract

The copy-move forgery refers to the copying and pasting of a region of the original image into the target region of the same image, which represents a typical tampering method with the characteristics of easy tampering and high-quality tampering. The existing single feature-based methods of forgery detection have certain shortcomings, such as high false alarm rate, low robustness, and low detection accuracy. To address these shortcomings, this paper proposes an improved two-stage detection method based on parallel feature fusion and an adaptive threshold generation algorithm. Firstly, the SLIC super-pixels segmentation algorithm is used for image preprocessing, and a similar region extraction algorithm without threshold is employed to obtain the suspected tampering regions with high similarity. Secondly, the parallel fusion feature is obtained based on the SIFT and HU features to express the characteristics of local regions. Then, the corresponding threshold value is generated based on the histogram of oriented gradient (HOG) to describe the texture characteristics of the obtained regions, which acts as a criterion to judge whether a region has been forged or not. The experimental results show that the proposed method outperforms the existing methods, achieving the accuracy of 99.01% and 98.5% on the MICC-F220 and MICC-F2000 datasets respectively. In addition, the proposed method has stronger robustness performance on COMOFOD dataset than the comparison methods.

Full Text
Published version (Free)

Talk to us

Join us for a 30 min session where you can share your feedback and ask us any queries you have

Schedule a call