A fast copy-move image forgery detection approach on a reduced search space

This paper proposes an overlapping block-based passive forensic scheme for copy-move forgery detection in digital images that works on a reduced search space. It uses a Gaussian image pyramid to generate and analyze input images at different resolutions. The conventional overlapping block-based procedures produce satisfactory results but are highly compute-intensive for large and medium-sized images. An increase in image size leads to a rapid rise in the number of overlapping blocks in the image, making processes like feature extraction, matching, and shift-vector calculations very expensive. The proposed approach initially performs relative forgery detection through block-wise processing of lower resolution components of the original image. In this process, discrete cosine transform is used to extract significant coefficients from each block and further analyzed to identify forgeries relatively in the selected lower resolution components. This process aids in selecting a smaller search space for potentially forged areas in the original image. Finally, the actual forgery detection is performed on this reduced search space, decreasing the computational overhead while maintaining accuracy in the results. The proposed procedure also shows robustness against various attacks and post-processing operations.

Automated summary

This paper proposes a passive forensic scheme for copy-move forgery detection in digital images that reduces the search space by using a Gaussian image pyramid to generate and analyze input images at different resolutions. The proposed approach initially performs relative forgery detection through block-wise processing of lower resolution components of the original image, and then actual forgery detection is performed on the reduced search space. The procedure is computationally efficient and robust against various attacks and post-processing operations.