Noise and Edge Based Dual Branch Image Manipulation Detection

TL;DR

This paper introduces a dual-branch neural network that uses noise and edge information to improve the detection of subtle image manipulations, achieving better results on multiple datasets.

Contribution

It proposes a novel dual-branch network with a noise-based input and a manipulation edge detection module for enhanced manipulation detection.

Findings

Effective in detecting manipulation artifacts on multiple datasets

Improves subtle manipulation trace extraction using noise and edge features

Incorporates distance-aware self-attention for better pixel correlation modeling

Abstract

Unlike ordinary computer vision tasks that focus more on the semantic content of images, the image manipulation detection task pays more attention to the subtle information of image manipulation. In this paper, the noise image extracted by the improved constrained convolution is used as the input of the model instead of the original image to obtain more subtle traces of manipulation. Meanwhile, the dual-branch network, consisting of a high-resolution branch and a context branch, is used to capture the traces of artifacts as much as possible. In general, most manipulation leaves manipulation artifacts on the manipulation edge. A specially designed manipulation edge detection module is constructed based on the dual-branch network to identify these artifacts better. The correlation between pixels in an image is closely related to their distance. The farther the two pixels are, the weaker the correlation. We add a distance factor to the self-attention module to better describe the correlation between pixels. Experimental results on four publicly available image manipulation datasets demonstrate the effectiveness of our model.