多特征融合的文档图像篡改区域分割方法

doi:10.11996/JG.j.2095-302X.2026050984

摘要/Abstract

摘要：

随着文档图像在政务、金融等关键领域的广泛应用，其被篡改后不仅会造成经济损失还可能危及社会公信力，保障其信息安全至关重要。与此同时篡改技术的快速发展增加了真伪鉴别的难度，因此文档图像篡改检测研究日益受到关注。为了提升文档图像篡改区域的定位能力，提出了一种多特征融合的文档图像篡改区域分割方法(MF-Net)。首先，设计了由3个并行分支组成的多特征编码器。其中，空域分支提取丰富的空域信息，频域分支结合多级小波变换与卷积操作充分提取频域中与篡改有关的特征，噪声分支采用SRM滤波器组通过分析噪声特征为模型提供新的视角。3个分支协同作用，实现多视角特征的充分提取；其次，设计了由通道注意力、空间注意力和无参数注意力串联构成的混合注意力特征融合模块，利用双特征引导的通道加权、空间聚焦与无参数抑制的多层次细化，实现多特征的深度融合；最后，借鉴人类视觉对篡改异常区域的重点关注机制，在多视图迭代解码器中加入注意力模块，使其能够自适应地聚焦于异常区域，以实现篡改区域的精准定位。该方法在公开数据集TextTamper上进行实验，与当前最优方法相比，其在F1分数和IoU指标上分别提升了2.8%和5.2%。实验结果表明，该方法能有效提升文档图像篡改定位的综合性能。

关键词: 篡改检测, 文档图像, 多特征融合, 数字图像取证, 深度学习

Abstract:

With the widespread adoption of document images in critical fields such as government administration and finance, tampering with these images not only incurs economic losses but also undermines public trust, making information security essential. Meanwhile, rapid advancement in tampering techniques has escalated the challenge of authenticity verification, thereby fueling growing research interest in document image tampering detection. To enhance the localization accuracy of tampered regions in document images, a multi-feature fusion-based segmentation method for document image tampered regions was proposed, namely the Multi-Feature Fusion Network (MF-Net). Firstly, a multi-feature encoder was designed, comprising three parallel branches. The spatial branch extracted rich spatial-domain information, the frequency branch integrates multi-level wavelet transform and convolution operations to fully capture tampering-related features in the frequency domain; and the noise branch employed the Spatial Rich Model (SRM) filter bank to analyze noise characteristics, providing the model with an additional perspective. These three branches were used jointly enable comprehensive extraction of multi-view features. Secondly, a hybrid attention feature-fusion module was developed, which serially combined channel attention, spatial attention, and parameter-free attention. Through multi-level refinement involving dual-feature-guided channel weighting, spatial focusing, and parameter-free suppression, deep fusion of multi-source features was achieved. Finally, inspired by the human visual system’s mechanism of prioritizing attention to abnormal tampered regions, an attention module was incorporated into a multi-view iterative decoder. This allowed the decoder to adaptively focus on anomalous regions, thereby realizing precise localization of tampered areas. Experimental validation of the proposed method was conducted on the public TextTamper dataset. Compared with state-of-the-art methods, the proposed MF-Net achieved improvements of 2.8% and 5.2% in F1-score and IoU, respectively. The experimental results demonstrated that the proposed method effectively enhanced the overall performance of tampered region localization in document images.

Key words: tampering detection, document image, multi-feature fusion, digital image forensics, deep learning

中图分类号:

TP 391.41

赵洁, 杨若荃, 武斌. 多特征融合的文档图像篡改区域分割方法[J]. 图学学报, 2026, 47(5): 984-993.

ZHAO Jie, YANG Ruoquan, WU Bin. Document image tampering region segmentation method based on multi-feature fusion[J]. Journal of Graphics, 2026, 47(5): 984-993.

图/表 13

图1 MF-Net总体架构

Fig. 1 Overall architecture of MF-Net

图2 频域特征分支

Fig. 2 Frequency domain feature branches

图3 噪声特征分支

Fig. 3 Noise feature branch

图4 混合注意力特征融合模块

Fig. 4 Hybrid attention feature fusion module

图5 多视图迭代注意力解码器

Fig. 5 Multi-view iterative attention decoder

图6 TextTamper篡改图像示例

Fig. 6 Examples of TextTamper tampering with images

表1 消融实验结果

Table 1 Results of ablation experiment

网络结构	评价指标
网络结构	F1	IoU	AUC	FNR
Baseline	0.633	0.619	0.835	0.307
w/o B_fre	0.675	0.674	0.857	0.266
w/o B_noise	0.680	0.676	0.862	0.257
w/o HATF	0.682	0.673	0.864	0.253
w/o MAD	0.664	0.651	0.852	0.276
MF-Net	0.689	0.685	0.869	0.244

图7 消融实验可视化定位图

Fig. 7 Visualization positioning map of the ablation experiment

表2 对比实验结果

Table 2 Comparison of experimental results

对比方法	评价指标
对比方法	F1	IoU	AUC	FNR
DFCN^[27]	0.643	0.624	0.858	0.264
MVSS-Net^[28]	0.581	0.524	0.861	0.257
SE-Net^[29]	0.632	0.608	0.857	0.266
RRU-Net^[30]	0.651	0.600	0.866	0.245
CFL-Net^[31]	0.520	0.462	0.842	0.291
FTMA-Net^[8]	0.661	0.633	0.868	0.248
MF-Net (Ours)	0.689	0.685	0.869	0.244

图8 对比实验定位结果可视化

Fig. 8 Visualization of the positioning results of the comparison experiments

图9 困难案例梯度热力图与篡改掩码对比图

Fig. 9 Gradient heat map of difficult cases

表3 模型复杂度对比

Table 3 Comparison of model complexity

方法	参数量/M	FLOPs/G
DFCN^[27]	2.55	35.46
MVSS-Net^[28]	142.78	166.48
SE-Net^[29]	3.94	38.62
RRU-Net^[30]	4.10	69.57
CFL-Net^[31]	111.91	68.05
FTMA-Net^[8]	24.70	48.98
MF-Net (Ours)	9.27	21.53

图10 MF-Net在TextTamper数据集上的鲁棒性测试((a) 随压缩质量因子变化的F1分数；(b) 随高斯标准差变化的F1分数)

Fig. 10 Robustness test of MF-Net on the TextTamper dataset ((a) F1 score versus compression quality factor; (b) F1 score versus Gaussian standard deviation)

参考文献 32

[1]	WU H W, ZHOU J T, TIAN J Y, et al. Robust image forgery detection over online social network shared images[C]// 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2022: 13430-13439.
[2]	GUILLARO F, COZZOLINO D, SUD A, et al. TruFor: leveraging all-round clues for trustworthy image forgery detection and localization[C]// 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2023: 20606-20615.
[3]	GUO X, LIU X H, REN Z Y, et al. Hierarchical fine-grained image forgery detection and localization[C]// 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2023: 3155-3165.
[4]	CHEN J X, LIAO X, WANG W, et al. SNIS: a signal noise separation-based network for post-processed image forgery detection[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2023, 33(2): 935-951. DOI URL
[5]	ZHU J Y, LI D, FU X Y, et al. Learning discriminative noise guidance for image forgery detection and localization[C]// The 38th AAAI Conference on Artificial Intelligence. Washington: AAAI, 2024: 7739-7747.
[6]	王裕鑫, 张博强, 谢洪涛, 等. 基于空域与频域关系建模的篡改文本图像检测[J]. 网络与信息安全学报, 2022, 8(3): 29-40.
	WANG Y X, ZHANG B Q, XIE H T, et al. Tampered text detection via RGB and frequency relationship modeling[J]. Chinese Journal of Network and Information Security, 2022, 8(3): 29-40 (in Chinese).
[7]	QU C F, LIU C Y, LIU Y L, et al. Towards robust tampered text detection in document image: new dataset and new solution[C]// 2023 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2023: 5937-5946.
[8]	DONG L, LIANG W P, WANG R D. Robust text image tampering localization via forgery traces enhancement and multiscale attention[J]. IEEE Transactions on Consumer Electronics, 2024, 70(1): 3495-3507. DOI URL
[9]	FERRARA P, BIANCHI T, DE ROSA A, et al. Image forgery localization via fine-grained analysis of CFA artifacts[J]. IEEE Transactions on Information Forensics and Security, 2012, 7(5): 1566-1577. DOI URL
[10]	AMERINI I, URICCHIO T, BALLAN L, et al. Localization of JPEG double compression through multi-domain convolutional neural networks[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition Workshops. New York: IEEE Press, 2017: 1865-1871.
[11]	ZHOU P, HAN X T, MORARIU V I, et al. Learning rich features for image manipulation detection[C]// 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2018: 1053-1061.
[12]	FRIDRICH J, KODOVSKY J. Rich models for steganalysis of digital images[J]. IEEE Transactions on Information Forensics and Security, 2012, 7(3): 868-882. DOI URL
[13]	WU Y, ABDALMAGEED W, NATARAJAN P. ManTra-Net: manipulation tracing network for detection and localization of image forgeries with anomalous features[C]// 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE Press, 2019: 9535-9544.
[14]	KWON M J, YU I J, NAM S H, et al. CAT-Net: compression artifact tracing network for detection and localization of image splicing[C]// 2021 IEEE Winter Conference on Applications of Computer Vision. New York: IEEE Press, 2021: 375-384.
[15]	ZENG N Y, WU P S, ZHANG Y Q, et al. DPMSN: a dual-pathway multiscale network for image forgery detection[J]. IEEE Transactions on Industrial Informatics, 2024, 20(5): 7665-7674. DOI URL
[16]	WU Y B, KONG X W, YOU X G, et al. Printer forensics based on page document's geometric distortion[C]// The 16th IEEE International Conference on Image Processing. New York: IEEE Press, 2009: 2909-2912.
[17]	SCHREYER M, SCHULZE C, STAHL A, et al. Intelligent printing technique recognition and photocopy detection for forensic document examination[EB/OL]. [2025-09-30]. https://dblp.uni-trier.de/db/conf/informatiktage/informatiktage2009.html#conf/informatiktage/Schreyer09.
[18]	VAN BEUSEKOM J, SHAFAIT F, BREUEL T M. Document inspection using text-line alignment[C]// The 9th IAPR International Workshop on Document Analysis Systems. New York: ACM, 2010: 263-270.
[19]	VAN BEUSEKOM J, SHAFAIT F, BREUEL T M. Text-line examination for document forgery detection[J]. International Journal on Document Analysis and Recognition, 2013, 16(2): 189-207. DOI URL
[20]	BERTRAND R, RAMOS TERRADES O, GOMEZ-KRÄMER P, et al. A conditional random field model for font forgery detection[C]// The 13th International Conference on Document Analysis and Recognition. New York: IEEE Press, 2015: 576-580.
[21]	ABRAMOVA S, BÖHME R. Detecting copy-move forgeries in scanned text documents[J]. Electronic Imaging, 2016, 2016(8): 1-9.
[22]	SIRAJUDEEN M, ANITHA R. Forgery document detection in information management system using cognitive techniques[J]. Journal of Intelligent & Fuzzy Systems: Applications in Engineering and Technology, 2020, 39(6): 8057-8068.
[23]	REN R Y, HAO Q X, GU F, et al. EMF-Net: an edge-guided multi-feature fusion network for text manipulation detection[J]. Expert Systems with Applications, 2024, 249: 123548. DOI URL
[24]	DUAN J X, SUN H, JI F, et al. RealDTT: towards a comprehensive real-world dataset for tampered text detection[J]. International Journal of Computer Vision, 2025, 133(10): 6993-7011. DOI
[25]	GUO M H, LU C Z, HOU Q B, et al. SegNeXt: rethinking convolutional attention design for semantic segmentation[C]// The 36th International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2022: 84.
[26]	MMSegmentation Contributors. MMSegmentation: OpenMMLab semantic segmentation toolbox and benchmark[EB/OL]. (2024-03-22) [2025-09-26]. https://github.com/open-mmlab/mmsegmentation
[27]	ZHUANG P Y, LI H D, TAN S Q, et al. Image tampering localization using a dense fully convolutional network[J]. IEEE Transactions on Information Forensics and Security, 2021, 16: 2986-2999. DOI URL
[28]	CHEN X R, DONG C B, JI J Q, et al. Image manipulation detection by multi-view multi-scale supervision[C]// 2021 IEEE/CVF International Conference on Computer Vision. New York: IEEE Press, 2021: 14165-14173.
[29]	ZHANG Y L, ZHU G P, WU L G, et al. Multi-task SE-network for image splicing localization[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2022, 32(7): 4828-4840. DOI URL
[30]	BI X L, WEI Y, XIAO B, et al. RRU-Net: the ringed residual U-net for image splicing forgery detection[C]// 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. New York: IEEE Press, 2019: 30-39.
[31]	NILOY F F, BHAUMIK K K, WOO S S. CFL-Net: image forgery localization using contrastive learning[C]//2023 IEEE/CVF Winter Conference on Applications of Computer Vision. New York: IEEE Press, 2023: 4631-4640.
[32]	阿里云天池, 阿里安全. 真实场景篡改图像检测挑战赛[EB/OL]. (2022-02-16) [2025-10-30]. https://tianchi.aliyun.com/competition/entrance/531945/introduction.
	Alibaba Cloud Tianchi, Alibaba Security. Real-world image forgery localization challenge[EB/OL]. (2022-02-16) [2025-10-30]. https://tianchi.aliyun.com/competition/entrance/531945/introduction(in Chinese).