MVSS-Net: Multi-View Multi-Scale Supervised Networks for Image Manipulation Detection

Abstract

The key research question for image manipulation detection is how to learn generalizable features that are sensitive to manipulations in novel data, whilst specific to prevent false alarms on authentic images. Current research emphasizes the sensitivity, with the specificity mostly ignored. In this paper we address both aspects by multi-view feature learning and multi-scale supervision. By exploiting noise distribution and boundary artifacts surrounding tampered regions, the former aims to learn semantic-agnostic and thus more generalizable features. The latter allows us to learn from authentic images which are nontrivial to be taken into account by the prior art that relies on a semantic segmentation loss. Our thoughts are realized by a new network which we term MVSS-Net and its enhanced version MVSS-Net++. Comprehensive experiments on six public benchmark datasets justify the viability of the MVSS-Net series for both pixel-level and image-level manipulation detection.

RWN: Robust Watermarking Network for Image Cropping Localization

“…We have conducted more embedding experiments over 1000 images from the test set, and the average PSNR between the protected images and the original images is 36.23dB, and the average SSIM [29] is 0.983. [2] 0.545 0.364 0.399 0.485 0.323 CAT-Net [19] 0.467 0.433 0.419 0.428 0.365 Accuracy. Fig.…”

“…Mantra-Net [1] uses fully convolutional networks for feature extraction and further uses long short-term memory (LSTM) cells for pixel-wise anomaly detection. In MVSS-Net [2], a system with multi-view feature learning and multi-scale supervision is developed to jointly exploit the noise view and the boundary artifact to learn manipulation detection features.…”

“…3 shows the tamper detection results on two test images. Note that for fair comparison, we add the same attacks on the original images for [1,19,2]. We observe that although the images are tampered by a variety of random attacks, we succeed in localizing the tampered areas.…”

“…Image manipulation leads to severe security threats, where misleading photojournalism, copyright violation, or fabricating stories can be a means for some writers or politicians to potentially influence public opinion. Researchers have developed a number of schemes to detect various kinds of digital attacks, e.g., tampering [1,2], Deep-Fake [3] and cropping [4,5,6]. Among them, cropping is an extremely cheap and effective way to manipulate the layout of an image and alter the underlying meaning.…”

“…The results demonstrate that our scheme can accurately locate the cropped region. We also show the effectiveness of tamper detection by comparison with some state-of-the-art image forgery detection schemes [1,19,2].…”