Copy-Move Forgery Detection using Random Opposition based Learning-Artificial Hummingbird Image Processing Approach

Copy-Move Forgery Detection (CMFD) is particular type of tampering image where an image part is copied and pasted in another image to determine a significant feature. Therefore, an aim in detecting copy-move forgeries is for detecting an area of image which are enormously same. However, accurately detecting these forgeries is challenging particularly in noise and subtle manipulation which results in suboptimal performance. In this research, the Random Opposition Based Learning-Artificial Hummingbird Approach (ROBL-AHA) is proposed for CMFD. In AHA, ROBL is incorporated to balance the exploitation which helps to minimize the convergence speed and avoids local optima issue. Initially, the image is taken from the standard dataset using MICC-F2000 to analyse the model performance. ScaleInvariant Feature Transform (SIFT) is applied for extracting a distinct keypoints from an image which remain invariant to variations based on rotation, noise, and scaling. Convolutional Neural Network (CNN) is used for accurately detecting the forgeries. The proposed ROBL-AHA achieves a better accuracy of $\mathbf{9 9. 7 6 \%}$ compared to existing methods like Mask Regionbased CNN and Grasshopper Optimization Approach (GOA) with Spotted Hyena Optimizer (SOA) using MICC-F2000 dataset.