D-Net: A dual-encoder network for image splicing forgery detection and localization

Zonglin Yang, Bo Liu, Xiuli Bi, Bin Xiao, Weisheng Li, Guoyin Wang, Xinbo Gao

Abstract: Highlights•The spatial pyramid global-feature extraction module expands the global in-sight.•The unfixed encoder autonomously learns the image fingerprints.•The fixed encoder intentionally provides the structural information of the pixels.•Not require a large number of training samples and a pre-training step.