An Image Copy-Move Forgery Detection based on SURF and Fourier-Mellin Transforms

Image forgery is widespread nowadays on social media. The problem worsened with advanced editing software, making forgery very hard to detect. A natural image consists of different features. During forgery detection, these features are extracted to find any manipulation in the image. Two main approaches under copy-move forgery detection are block-based and key-based techniques. The paper proposes exploiting a combined approach based on block-based and key-point techniques such as speed-up robust feature (SURF) and Fourier-Mellin transform (FMT). The image is first categorized into smooth and textured parts. Surf is applied to textural areas of the image, while FMT coefficients are exploited from smooth regions. Dense linear fitting (DLF) and random sampling consensus (RANSAC) are used separately to eliminate the false matching points and outliers. Finally, mathematical morphology is adapted to generate the binary map for both parts of the image to locate the forgery area. Experimental results prove that the suggested model is robust against blurring, scaling, and compression attacks.