JISE


  [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16]


Journal of Information Science and Engineering, Vol. 39 No. 4, pp. 759-775


High-resolution Noise Artifact Tracking network for Image Splicing Forgery Detection


HAIBO YU1,2, TAO ZHANG1,2,+, WENHUAN ZHU3, LI ZHANG1 AND JIALI JIN1
1School of Artificial Intelligence and Computer Science
Jiangnan University
Wuxi, 214122 P.R. China
E-mail: taozhang@jiangnan.edu.cn
+
2China Ship Scientific Research Center
Wuxi, 214122 P.R. China

3School of Electronic Information and Electrical Engineering
Shanghai Jiao Tong University
Shanghai, 200240 P.R. China


Aiming at the problems that existing methods lose a lot of spatial information and are easily interfered by image content, this paper proposes an end-to-end trainable highresolution noise artifact tracking network (HRNAT-Net) for image splicing forgery detection. The network adopts HRNet as the network backbone, which can always retain highresolution features, thereby effectively reducing the loss of important spatial information such as edge features. Besides, the network utilizes dilated convolution, which further preserves rich tampering information. Noise information are used as auxiliary information to build a dual-stream network to exploit visual and noise features simultaneously. Through multiple multi-scale feature fusion, the expressive ability of features is further enhanced. Finally, the visual features and noise features of corresponding resolutions are concatenated, and through multi-scale feature fusion, visual features and noise features are combined to locate the tampered area. The proposed network effectively solves the loss of spatial information, and simultaneously seeks forensic clues from the visual and noise levels. We evaluate the network performance on the CASIA v2.0 dataset and the Columbia dataset using metrics such as pixel-level Precision, Recall, and F1, and the detection accuracy (F1) reaches 0.845 and 0.950, respectively. Experimental results show that the proposed network outperforms state-of-the-art methods and has good robustness to JPEG compression, Gaussian noise, and resizing attacks.


Keywords: image splicing forgery, noise inconsistency, HRNet, multi-scale, dilated convolution

  Retrieve PDF document (JISE_202304_04.pdf)