《计算机应用》唯一官方网站 ›› 2024, Vol. 44 ›› Issue (4): 1303-1309.DOI: 10.11772/j.issn.1001-9081.2023040493

• 多媒体计算与计算机仿真 • 上一篇    

用于篡改图像检测和定位的双通道渐进式特征过滤网络

付顺旺1, 陈茜1(), 李智2, 王国美2, 卢妤3   

  1. 1.贵州大学 大数据与信息工程学院,贵阳 550025
    2.贵州大学 计算机科学与技术学院,贵阳 550025
    3.贵州电网有限责任公司,贵阳 550002
  • 收稿日期:2023-04-28 修回日期:2023-07-26 接受日期:2023-07-31 发布日期:2023-12-04 出版日期:2024-04-10
  • 通讯作者: 陈茜
  • 作者简介:付顺旺(1996—),男,贵州遵义人,硕士研究生,CCF会员,主要研究方向:深度学习、图像篡改检测
    陈茜(1981—),女,贵州贵阳人,教授,博士,主要研究方向:半导体材料、机器学习 chenzhangqianer@163.com
    李智(1977—),女,贵州贵阳人,教授,博士,CCF会员,主要研究方向:信息隐藏、人工智能、医学图像分析
    王国美(1975—),女,贵州贵阳人,副教授,硕士,主要研究方向:水印算法、人工智能、医学图像分析
    卢妤(1976—),女,贵州贵阳人,硕士,主要研究方向:人工智能、电网图像异常检测。
  • 基金资助:
    国家自然科学基金资助项目(62062023)

Two-channel progressive feature filtering network for tampered image detection and localization

Shunwang FU1, Qian CHEN1(), Zhi LI2, Guomei WANG2, Yu LU3   

  1. 1.College of Big Data and Information Engineering,Guizhou University,Guiyang Guizhou 550025,China
    2.College of Computer Science and Technology,Guizhou University,Guiyang Guizhou 550025,China
    3.Guizhou Power Grid Company Limited,Guiyang Guizhou 550002,China
  • Received:2023-04-28 Revised:2023-07-26 Accepted:2023-07-31 Online:2023-12-04 Published:2024-04-10
  • Contact: Qian CHEN
  • About author:FU Shunwang, born in 1996, M.S. candidate. His research interests include deep learning, image tamper detection.
    CHEN Qian, born in 1981, Ph. D., professor. Her research interests include semiconductor material, machine learning.
    LI Zhi, born in 1977, Ph. D., professor. Her research interests include information hiding, artificial intelligence, medical image analysis.
    WANG Guomei, born in 1975, M. S., associate professor. Her research interests include watermark algorithm, artificial intelligence, medical image analysis.
    LU Yu, born in 1976, M. S. Her research interests include artificial intelligence, abnormal detection of power grid images.
  • Supported by:
    National Natural Science Foundation of China(62062023)

摘要:

针对现有基于深度学习的篡改图像检测网络通常存在检测精度不高、算法可迁移性弱等问题,提出一种双通道渐进式特征过滤网络。利用两个通道并行提取图像的双域特征,一个通道提取图像空间域的浅层和深层特征,另一个通道提取图像噪声域的特征分布;同时,使用渐进式细微特征筛选机制过滤冗余特征,逐步定位篡改区域;为了更准确地提取篡改掩码,提出一个双输入细微特征提取模块,结合空间域和噪声域的细微特征,生成更准确的篡改掩码;在解码过程中,通过融合不同尺度的过滤特征和网络的上下文信息,提高网络对篡改区域的定位能力。实验结果表明,在检测和定位方面,与现有先进的篡改检测网络ObjectFormer、MVSS-Net(Multi-View multi-Scale Supervision Network)和PSCC-Net(Progressive Spatio-Channel Correlation Network)相比,所提网络的F1分数在CASIA V2.0数据集上分别提高10.4、5.9和12.9个百分点;面对高斯低通滤波、高斯噪声和JPEG压缩攻击时,相较于ManTra-Net(Manipulation Tracing Network)、SPAN(Spatial Pyramid Attention Network),所提网络的曲线下面积(AUC)分别至少提高了10.0、5.4个百分点。验证了所提网络可以有效解决篡改检测算法存在的检测精度不高、迁移性差等问题。

关键词: 篡改图像检测, 多尺度融合, 全局相关性, 被动取证, 残差网络

Abstract:

The existing image tamper detection networks based on deep learning often have problems such as low detection accuracy and weak algorithm transferability. To address the above issues, a two-channel progressive feature filtering network was proposed. Two channels were used to extract the two-domain features of the image in parallel, one of which was used to extract the shallow and deep features of the image spatial domain, and the other channel was used to extract the feature distribution of the image noise domain. At the same time, a progressive subtle feature screening mechanism was used to filter redundant features and gradually locate the tampered regions; in order to extract the tamper mask more accurately, a two-channel subtle feature extraction module was proposed, which combined the subtle features of the spatial domain and the noise domain to generate a more accurate tamper mask. During the decoding process, the localization ability of the network to tampered regions was improved by fusing filtered features of different scales and the contextual information of the network. The experimental results show that in terms of detection and localization, compared with the existing advanced tamper detection networks ObjectFormer, Multi-View multi-Scale Supervision Network (MVSS-Net) and Progressive Spatio-Channel Correlation Network (PSCC-Net), the F1 score of the proposed network is increased by an 10.4, 5.9 and 12.9 percentage points on CASIA V2.0 dataset; when faced with Gaussian low-pass filtering, Gaussian noise, and JPEG compression attacks, compared with Manipulation Tracing Network (ManTra-Net) and Spatial Pyramid Attention Network (SPAN), the Area Under Curve (AUC) of the proposed network is increased by 10.0 and 5.4 percentage points at least. It is verified that the proposed network can effectively solve the problems of low detection accuracy and poor transferability in the tamper detection algorithm.

Key words: tampered image detection, multiscale fusion, global correlation, passive forensics, residual network

中图分类号: