Two-stream Encoder-Decoder Network for Localizing Image Forgeries

TL;DR

This paper introduces a two-stream encoder-decoder network that combines high-level and low-level features to accurately localize image forgeries, improving upon existing methods.

Contribution

The paper presents a novel two-stream network architecture that simultaneously learns high-level and low-level manipulation features for precise forgery localization.

Findings

Outperforms state-of-the-art methods on multiple datasets.

Effectively captures both noise residuals and high-level artifacts.

Demonstrates robustness across various manipulated images.

Abstract

This paper proposes a novel two-stream encoder-decoder network, which utilizes both the high-level and the low-level image features for precisely localizing forged regions in a manipulated image. This is motivated from the fact that the forgery creation process generally introduces both the high-level artefacts (e.g. unnatural contrast) and the low-level artefacts (e.g. noise inconsistency) to the forged images. In the proposed two-stream network, one stream learns the low-level manipulation-related features in the encoder side by extracting noise residuals through a set of high-pass filters in the first layer of the encoder network. In the second stream, the encoder learns the high-level image manipulation features from the input image RGB values. The coarse feature maps of both the encoders are upsampled by their corresponding decoder network to produce dense feature maps. The dense feature maps of the two streams are concatenated and fed to a final convolutional layer with sigmoidal activation to produce pixel-wise prediction. We have carried out experimental analysis on multiple standard forensics datasets to evaluate the performance of the proposed method. The experimental results show the efficacy of the proposed method with respect to the state-of-the-art.