An Effective Detection Approach for Copy-Move Image Forgery Using Feature Matching by Adaptive Threshold-Aided Reserved G2NN

The ability to manipulate and change content in images is simple with the introduction of image editing tools. Important aspects are being added, altered, or removed from an image without leaving any perceptual evidence of manipulation behind. The development of media editing software helps the non-experts in image processing to adjust their digital images. There are numerous ways to counterfeit digital images as well as “image splicing, copy-move forging, and image retouching.” In forgery recognition, a portion of a picture is copied and used to replace a new portion of a similar image at a different location, which is the most popular technique for altering digital images. The photometric characteristic is large since the replicated image patch originates from the same image. Therefore, an efficient method to detect such artifacts is proposed. At first, the necessary image is collected from standard data resources. The composed image is then given to the pre-processing phase. Here, the Weiner filtering and contrast stretching process is performing the preprocessing. Further, the preprocessed image is subjected to the feature extraction process. The Multi-Head Cross-Attention-based Vision Transformer (MCA-VIT) extracts the required feature. Further, the resultant feature is sent as input to the Adaptive Threshold-based Reserved G2NN Feature Matching (AT-RG2NN-FM). The developed method performs the optimal feature matching to achieve the maximum detection rate. For further enhancement, parameters used in the developed model are tuned by using the Revised Hermit Crab Optimizer (RHCO). The overall process of the developed model is enhanced from the simulation result by comparing it with multiple standard approaches.