Downsampled image forensic network based on image recovery and spatial channel attention

doi:10.11772/j.issn.1001-9081.2024050672

Journal of Computer Applications ›› 2025, Vol. 45 ›› Issue (5): 1582-1588.DOI: 10.11772/j.issn.1001-9081.2024050672

• Cyber security • Previous Articles

Downsampled image forensic network based on image recovery and spatial channel attention

Aoling LIU¹, Wuyang SHAN¹(), Junying QIU², Mao TIAN³, Jun LI¹

^1.College of Computer Science and Cyber Security （Pilot Software College），Chengdu University of Technology，Chengdu Sichuan 610059，China
^2.College of Fashion and Design Arts，Sichuan Normal University，Chengdu Sichuan 610066，China
^3.School of Computer Science and Technology，Chongqing University of Posts and Telecommunications，Chongqing 400065，China

Received:2024-05-27 Revised:2024-09-30 Accepted:2024-10-25 Online:2024-11-05 Published:2025-05-10
Contact: Wuyang SHAN
About author:LIU Aoling， born in 1999， M. S. candidate. Her research interests include image forensics， deep learning.
SHAN Wuyang， born in 1988， Ph. D.， lecturer. His research interests include multimedia forensics， computer vision， data hiding.
QIU Junying， born in 1990， M. S.， lecturer. Her research interests include intelligent processing， analysis and security of digital art media.
TIAN Mao， born in 1988， Ph. D.， lecturer. His research interests include stereo matching， point cloud and image fusion， deep learning.
LI Jun， born in 1973， Ph. D.， professor. His research interests include high-performance computing， computer vision， artificial intelligence.
Supported by:
National Natural Science Foundation of China(42001417)

基于图像恢复和空间通道注意力的下采样图像取证网络

刘澳龄¹, 单武扬¹(), 邱骏颖², 田茂³, 李军¹

^1.成都理工大学计算机与网络安全学院（示范性软件学院），成都 610059
^2.四川师范大学服装与设计艺术学院，成都 610066
^3.重庆邮电大学计算机科学与技术学院，重庆 400065

通讯作者: 单武扬
作者简介:刘澳龄（1999—），女，四川内江人，硕士研究生，CCF会员，主要研究方向：图像取证、深度学习
单武扬（1988—），男，湖南岳阳人，讲师，博士，主要研究方向：多媒体取证、计算机视觉、数据隐藏
邱骏颖（1990—），女，四川成都人，讲师，硕士，主要研究方向：数字艺术媒体智能处理、分析和安全
田茂（1988—），男，四川德阳人，讲师，博士，CCF会员，主要研究方向：立体匹配、点云和影像融合、深度学习
李军（1973—），男，湖北潜江人，教授，博士，主要研究方向：高性能计算、计算机视觉、人工智能。
基金资助:
国家自然科学基金资助项目(42001417)

Abstract

Abstract:

Downsampling operation will make images lose high-frequency forensic traces and detail information， increasing the difficulty of image forensics. Existing deep learning-based image forensic networks cannot effectively detect the images tampered by downsampling operation， making the enhancement of robustness in downsampling image forensics methods becomes a bottleneck in image forensics. To solve these problems， a downsampling image forensic network named HirrNet （Hierarchical Ringed Residual U-Net） was proposed， which consists of an image recovery module and a tampering detection module. In the image recovery module， the idea of Hierarchical Conditional Flow （HCF） was used to reduce the loss of high-frequency information by recovering forensic traces and details in tampered images， so as to improve the performance of tampering detection. In the tampering detection module， an end-to-end image segmentation network RRU-Net （Ringed Residual U-Net） was employed for tampering detection. Besides， by combining the Spatial and Channel Squeeze & Excitation （SCSE） mechanism， the extraction of tampering-related features in the downsampled image was effectively enhanced. Experimental results show that HirrNet outperforms comparative networks in terms of Area Under the receiver operating characteristic Curve （AUC）， F1-score and Intersection and Union （IoU） on DSO， Columbia， CASIA， and NIST16 datasets. Compared with the comparative methods， HirrNet improves the AUC by 25 and 30 percentage points on average for the tampered images scaled down to 1/2 and 1/4 of their original sizes on CASIA dataset. These findings indicate that HirrNet can effectively resolve the poor robustness of existing downsampled image forensic methods.

Key words: image forensics, image recovery, spatial channel attention, downsampling

摘要：

下采样操作会使图像丢失高频取证痕迹和细节信息，增加图像取证的难度，而现有的基于深度学习的图像取证网络不能有效检测经过下采样操作篡改的图像，导致提高下采样图像取证方法的鲁棒性成为图像取证的瓶颈。为解决这个问题，提出一个下采样图像取证网络HirrNet（Hierarchical RRU-Net）。HirrNet主要包括图像恢复模块和篡改检测模块：图像恢复模块使用分层条件流（HCF）的思想，通过恢复篡改图像取证痕迹和细节信息减少高频信息的丢失，从而提高篡改检测性能；篡改检测模块则使用端到端图像分割网络RRU-Net（Ringed Residual U-Net）进行篡改检测。此外，通过结合空间和通道压缩与激励（SCSE）机制，可有效增强下采样图像中与篡改相关的特征的提取。实验结果表明，HirrNet在DSO、Columbia、CASIA和NIST16数据集上的受试者特征工作曲线下面积（AUC）、F1分数和交并比（IoU）优于对比网络。其中，在CASIA数据集上，对于尺寸缩小至原图1/2和1/4的篡改图像，HirrNet的AUC指标相较于对比方法平均提升25和30个百分点。可见，HirrNet可以有效解决现有的下采样图像取证方法鲁棒性差的问题。

关键词: 图像取证, 图像恢复, 空间通道注意力, 下采样

CLC Number:

TP751

Aoling LIU, Wuyang SHAN, Junying QIU, Mao TIAN, Jun LI. Downsampled image forensic network based on image recovery and spatial channel attention[J]. Journal of Computer Applications, 2025, 45(5): 1582-1588.

刘澳龄, 单武扬, 邱骏颖, 田茂, 李军. 基于图像恢复和空间通道注意力的下采样图像取证网络[J]. 《计算机应用》唯一官方网站, 2025, 45(5): 1582-1588.

Figures/Tables 10

References 33

1	HUH M， LIU A， OWENS A， et al. Fighting fake news： Image splice detection via learned self-consistency［C］// Proceedings of the 2018 European Conference on Computer Vision， LNCS 11215. Cham： Springer， 2018： 106-124.
2	ASNANI V， YIN X， HASSNER T， et al. Proactive image manipulation detection［C］// Proceedings of the 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2022： 15365-15374.
3	付顺旺，陈茜，李智，等. 用于篡改图像检测和定位的双通道渐进式特征过滤网络［J］. 计算机应用， 2024， 44（4）： 1303-1309.
	FU S W， CHEN Q， LI Z， et al. Dual-channel progressive feature filtering network for tampered image detection and localization［J］. Journal of Computer Applications， 2024， 44（4）： 1303-1309.
4	陈俊韬，朱子奇. 基于多尺度特征提取与融合的图像复制-粘贴伪造检测［J］. 计算机应用， 2023， 43（9）： 2919-2924.
	CHEN J T， ZHU Z Q. Image copy-paste forgery detection based on multi-scale feature extraction and fusion［J］. Journal of Computer Applications， 2023， 43（9）： 2919-2924.
5	ABECIDAN R， ITIER V， BOULANGER J， et al. Unsupervised JPEG domain adaptation for practical digital image forensics［C］// Proceedings of the 2021 IEEE International Workshop on Information Forensics and Security. Piscataway： IEEE， 2021： 1-6.
6	SHAN W， ZOU D， WANG P， et al. RIFD-Net： a robust image forgery detection network［J］. IEEE Access， 2024， 12： 20326-20340.
7	SHAN W， YI Y， QIU J， et al. Robust median filtering forensics using image deblocking and filtered residual fusion［J］. IEEE Access， 2019， 7： 17174-17183.
8	王朋博，单武扬，李军，等. 抗高强度椒盐噪声的鲁棒拼接取证算法［J］. 计算机应用， 2024， 44（10）： 3177-3184.
	WANG P B， SHAN W Y， LI J， et al. Robust splicing forensics algorithm against high-intensity salt-and-pepper noise［J］. Journal of Computer Applications， 2024， 44（10）： 3177-318.
9	MAYER O， STAMM M C. Forensic similarity for digital images［J］. IEEE Transactions on Information Forensics and Security， 2020， 15： 1331-1346.
10	LIU X， LIU Y， CHEN J， et al. PSCC-Net： progressive spatio-channel correlation network for image manipulation detection and localization［J］. IEEE Transactions on Circuits and Systems for Video Technology， 2022， 32（11）： 7505-7517.
11	ZHUANG P， LI H， TAN S， et al. Image tampering localization using a dense fully convolutional network［J］. IEEE Transactions on Information Forensics and Security， 2021， 16： 2986-2999.
12	LIANG J， LUGMAYR A， ZHANG K， et al. Hierarchical conditional flow： a unified framework for image super-resolution and image rescaling［C］// Proceedings of the 2021 IEEE/CVF International Conference on Computer Vision. Piscataway： IEEE， 2021： 4056-4065.
13	BI X， WEI Y， XIAO B， et al. RRU-Net： the ringed residual U-Net for image splicing forgery detection［C］// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Piscataway： IEEE， 2019： 30-39.
14	ROY A G， NAVAB N， WACHINGER C. Recalibrating fully convolutional networks with spatial and channel “squeeze and excitation” blocks［J］. IEEE Transactions on Medical Imaging， 2019， 38（2）： 540-549.
15	OEZTIRELI A C， GROSS M. Perceptually based downscaling of images［J］. ACM Transactions on Graphics， 2015， 34（4）： No.77.
16	MITCHELL D P， NETRAVALI A N. Reconstruction filters in computer-graphics［J］. ACM SIGGRAPH Computer Graphics， 1988， 22（4）： 221-228.
17	SUN W， CHEN Z. Learned image downscaling for upscaling using content adaptive resampler［J］. IEEE Transactions on Image Processing， 2020， 29： 4027-4040.
18	KIM H， CHOI M， LIM B， et al. Task-aware image downscaling［C］// Proceedings of the 2018 European Conference on Computer Vision， LNCS 11208. Cham： Springer， 2018： 419-434.
19	LI Y， LIU D， LI H， et al. Learning a convolutional neural network for image compact-resolution［J］. IEEE Transactions on Image Processing， 2019， 28（3）： 1092-1107.
20	WANG X， GIRSHICK R， GUPTA A， et al. Non-local neural networks［C］// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2018： 7794-7803.
21	ZHAO H， SHI J， QI X， et al. Pyramid scene parsing network［C］// Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2017： 6230-6239.
22	HU J， SHEN L， SUN G. Squeeze-and-excitation networks［C］// Proceedings of the 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway： IEEE， 2018： 7132-7141.
23	WOO S， PARK J， LEE J Y， et al. CBAM： convolutional block attention module［C］// Proceedings of the 2018 European Conference on Computer Vision， LNCS 11211. Cham： Springer， 2018： 3-19.
24	FU J, LIU J, TIAN H, et al. Dual attention network for scene segmentation[C]// Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Piscataway: IEEE,2019: 3141-3149.
25	DE CARVALHO T J， RIESS C， ANGELOPOULOU E， et al. Exposing digital image forgeries by illumination color classification［J］. IEEE Transactions on Information Forensics and Security， 2013， 8（7）： 1182-1194.
26	HSU Y F， CHANG S F. Detecting image splicing using geometry invariants and camera characteristics consistency［C］// Proceedings of the 2006 IEEE International Conference on Multimedia Expo. Piscataway： IEEE， 2006： 549-552.
27	NIST. Nimble challenge 2017 evaluation［EB/OL］. ［2022-01-23］..
28	DONG J， WANG W， TAN T. CASIA image tampering detection evaluation database［C］// Proceedings of the 2013 IEEE China Summit and International Conference on Signal and Information Processing. Piscataway： IEEE， 2013： 422-426.
29	WEN B H， ZHU Y， SUBRAMANIAN R， et al. COVERAGE - a novel database for copy-move forgery detection［C］// Proceedings of the 2016 IEEE International Conference on Image Processing. Piscataway： IEEE， 2016： 161-165.
30	DONG C， LOY C C， HE K， et al. Image super-resolution using deep convolutional networks［J］. IEEE Transactions on Pattern Analysis and Machine Intelligence， 2015， 38（2）： 295-307.
31	ZHANG Y， LI K， LI K， et al. Image super-resolution using very deep residual channel attention networks［C］// Proceedings of the 2018 European Conference on Computer Vision， LNCS 11211. Cham： Springer， 2018： 294-310.
32	XIAO M， ZHENG S， LIU C， et al. Invertible image rescaling［C］// Proceedings of the 2020 European Conference on Computer Vision， LNCS 12346. Cham： Springer， 2020： 126-144.
33	XIAO M， ZHENG S， LIU C， et al. Invertible rescaling network and its extensions［J］. International Journal of Computer Vision， 2023， 131（1）： 134-159.

下采样&上采样	缩小比例	DSO		Columbia		NIST16		CASIA
下采样&上采样	缩小比例	PSNR/dB	SSIM	PSNR/dB	SSIM	PSNR/dB	SSIM	PSNR/dB	SSIM
Bicubic & SRCNN^［30］	2	80.52	0.921 3	73.23	0.934 2	73.42	0.912 0	72.72	0.931 8
Bicubic & RCAN^［31］		84.16	0.926 5	75.25	0.945 5	75.29	0.921 6	74.22	0.941 6
Bicubic & IRN^［32］		90.06	0.975 6	80.32	0.953 5	80.02	0.944 6	81.54	0.964 8
Bicubic & IRN+^［33］		91.25	0.968 3	81.37	0.955 8	81.57	0.946 9	82.48	0.952 9
本文恢复方法		90.12	0.967 3	82.52	0.961 5	80.09	0.952 1	83.27	0.948 3
Bicubic & SRCNN^［30］	4	68.35	0.857 6	64.68	0.914 5	65.25	0.821 9	63.63	0.749 8
Bicubic & RCAN^［31］		70.28	0.875 3	66.29	0.925 5	69.29	0.841 6	66.26	0.821 6
Bicubic & IRN^［32］		78.36	0.916 8	74.46	0.926 5	74.46	0.868 8	74.35	0.941 7
Bicubic & IRN+^［33］		79.51	0.912 4	75.38	0.928 8	75.22	0.874 9	75.84	0.953 0
本文恢复方法		80.23	0.912 7	76.23	0.934 6	75.37	0.880 4	76.26	0.958 2

下采样&上采样	缩小比例	DSO		Columbia		NIST16		CASIA
下采样&上采样	缩小比例	PSNR/dB	SSIM	PSNR/dB	SSIM	PSNR/dB	SSIM	PSNR/dB	SSIM
Bicubic & SRCNN^［30］	2	80.52	0.921 3	73.23	0.934 2	73.42	0.912 0	72.72	0.931 8
Bicubic & RCAN^［31］		84.16	0.926 5	75.25	0.945 5	75.29	0.921 6	74.22	0.941 6
Bicubic & IRN^［32］		90.06	0.975 6	80.32	0.953 5	80.02	0.944 6	81.54	0.964 8
Bicubic & IRN+^［33］		91.25	0.968 3	81.37	0.955 8	81.57	0.946 9	82.48	0.952 9
本文恢复方法		90.12	0.967 3	82.52	0.961 5	80.09	0.952 1	83.27	0.948 3
Bicubic & SRCNN^［30］	4	68.35	0.857 6	64.68	0.914 5	65.25	0.821 9	63.63	0.749 8
Bicubic & RCAN^［31］		70.28	0.875 3	66.29	0.925 5	69.29	0.841 6	66.26	0.821 6
Bicubic & IRN^［32］		78.36	0.916 8	74.46	0.926 5	74.46	0.868 8	74.35	0.941 7
Bicubic & IRN+^［33］		79.51	0.912 4	75.38	0.928 8	75.22	0.874 9	75.84	0.953 0
本文恢复方法		80.23	0.912 7	76.23	0.934 6	75.37	0.880 4	76.26	0.958 2

方法	缩小比例	DSO			Columbia			NIST16			CASIA			COVERAGE
方法	缩小比例	AUC	F1	IoU	AUC	F1	IoU	AUC	F1	IoU	AUC	F1	IoU	AUC	F1	IoU
ForSim^［9］	无操作	0.78	0.45	0.35	0.73	0.60	0.42	0.63	0.16	0.12	0.54	0.15	0.10	0.60	0.18	0.14
PSCC-Net^［10］		0.72	0.30	0.21	0.86	0.52	0.38	0.74	0.24	0.20	0.85	0.18	0.11	0.78	0.26	0.26
DFCN^［11］		0.74	0.30	0.18	0.75	0.60	0.50	0.66	0.21	0.11	0.75	0.35	0.32	0.70	0.24	0.22
HirrNet		0.83	0.42	0.30	0.84	0.71	0.60	0.76	0.31	0.25	0.86	0.46	0.40	0.78	0.34	0.28
ForSim^［9］	2	0.71	0.31	0.32	0.71	0.43	0.38	0.60	0.14	0.10	0.41	0.12	0.08	0.60	0.16	0.14
PSCC-Net^［10］		0.68	0.26	0.19	0.74	0.41	0.34	0.71	0.21	0.18	0.59	0.15	0.10	0.73	0.24	0.21
DFCN^［11］		0.72	0.27	0.16	0.72	0.49	0.47	0.64	0.19	0.10	0.71	0.32	0.29	0.65	0.21	0.15
HirrNet		0.81	0.37	0.29	0.82	0.68	0.58	0.74	0.29	0.23	0.82	0.43	0.36	0.72	0.30	0.27
ForSim^［9］	4	0.62	0.24	0.25	0.64	0.36	0.27	0.54	0.10	0.07	0.35	0.06	0.06	0.56	0.12	0.10
PSCC-Net^［10］		0.58	0.21	0.16	0.62	0.31	0.24	0.62	0.14	0.12	0.49	0.11	0.07	0.63	0.18	0.15
DFCN^［11］		0.64	0.22	0.13	0.61	0.34	0.38	0.59	0.15	0.08	0.65	0.24	0.15	0.61	0.17	0.12
HirrNet		0.80	0.36	0.26	0.78	0.56	0.54	0.70	0.27	0.21	0.80	0.41	0.34	0.72	0.29	0.26

方法	缩小比例	DSO			Columbia			NIST16			CASIA			COVERAGE
方法	缩小比例	AUC	F1	IoU	AUC	F1	IoU	AUC	F1	IoU	AUC	F1	IoU	AUC	F1	IoU
ForSim^［9］	无操作	0.78	0.45	0.35	0.73	0.60	0.42	0.63	0.16	0.12	0.54	0.15	0.10	0.60	0.18	0.14
PSCC-Net^［10］		0.72	0.30	0.21	0.86	0.52	0.38	0.74	0.24	0.20	0.85	0.18	0.11	0.78	0.26	0.26
DFCN^［11］		0.74	0.30	0.18	0.75	0.60	0.50	0.66	0.21	0.11	0.75	0.35	0.32	0.70	0.24	0.22
HirrNet		0.83	0.42	0.30	0.84	0.71	0.60	0.76	0.31	0.25	0.86	0.46	0.40	0.78	0.34	0.28
ForSim^［9］	2	0.71	0.31	0.32	0.71	0.43	0.38	0.60	0.14	0.10	0.41	0.12	0.08	0.60	0.16	0.14
PSCC-Net^［10］		0.68	0.26	0.19	0.74	0.41	0.34	0.71	0.21	0.18	0.59	0.15	0.10	0.73	0.24	0.21
DFCN^［11］		0.72	0.27	0.16	0.72	0.49	0.47	0.64	0.19	0.10	0.71	0.32	0.29	0.65	0.21	0.15
HirrNet		0.81	0.37	0.29	0.82	0.68	0.58	0.74	0.29	0.23	0.82	0.43	0.36	0.72	0.30	0.27
ForSim^［9］	4	0.62	0.24	0.25	0.64	0.36	0.27	0.54	0.10	0.07	0.35	0.06	0.06	0.56	0.12	0.10
PSCC-Net^［10］		0.58	0.21	0.16	0.62	0.31	0.24	0.62	0.14	0.12	0.49	0.11	0.07	0.63	0.18	0.15
DFCN^［11］		0.64	0.22	0.13	0.61	0.34	0.38	0.59	0.15	0.08	0.65	0.24	0.15	0.61	0.17	0.12
HirrNet		0.80	0.36	0.26	0.78	0.56	0.54	0.70	0.27	0.21	0.80	0.41	0.34	0.72	0.29	0.26

编号	检测网络	无操作			缩小比例为2			缩小比例为4
编号	检测网络	AUC	F1	IoU	AUC	F1	IoU	AUC	F1	IoU
#1	Base+R	0.684	0.245	0.216	0.620	0.208	0.184	0.587	0.159	0.163
#2	Base+R+CAM	0.698	0.257	0.224	0.639	0.235	0.206	0.609	0.178	0.188
#3	Base+R+SAM	0.712	0.262	0.240	0.651	0.258	0.221	0.627	0.203	0.212
#4	Base+R+SE	0.754	0.403	0.347	0.736	0.384	0.321	0.659	0.275	0.237
#5	Base+R+CBAM	0.782	0.435	0.348	0.752	0.420	0.346	0.715	0.317	0.253
#6	Base+R+SCSE	0.826	0.477	0.390	0.803	0.447	0.371	0.774	0.403	0.342

Downsampled image forensic network based on image recovery and spatial channel attention

基于图像恢复和空间通道注意力的下采样图像取证网络

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 10

References 33

Related Articles 13

Recommended Articles

Metrics

[1]	Wanli SHEN, Yujin ZHANG, Wan HU. U-shaped feature pyramid network for image inpainting forensics [J]. Journal of Computer Applications, 2023, 43(2): 545-551.
[2]	Cong XU, Xingtian WANG, Yongpeng TAO. Reversible data hiding method based on high-order bit-plane redundancy [J]. Journal of Computer Applications, 2022, 42(1): 171-177.
[3]	CAO Yanjue, AN Bowen, LI Qiming. Real-time simulation and application of depth-of-field based on post-processing [J]. Journal of Computer Applications, 2015, 35(5): 1439-1443.
[4]	LIU Yi LIU Yongben. Image resampling tampering detection based on further resampling [J]. Journal of Computer Applications, 2014, 34(3): 815-819.
[5]	DU Hong-ye YAO Wang-shu. Method of blind image forensics based on Lambert-Phong model [J]. Journal of Computer Applications, 2012, 32(11): 3171-3173.
[6]	CHEN Zong-ming ZHOU Zhi-ping. Source camera identification using Noise variance and texture complexity analysis [J]. Journal of Computer Applications, 2012, 32(06): 1563-1566.
[7]	CHEN Zong-ming ZHOU Zhi-ping. Source camera identification based on nonsubsampled Contourlet transform denoising filter design [J]. Journal of Computer Applications, 2012, 32(02): 507-513.
[8]	OU Jiajia CAI Biye XIONG Bin LI Feng. Detection of image region-duplication forgery based on gray level co-occurrence matrix [J]. Journal of Computer Applications, 2011, 31(06): 1628-1630.
[9]	Chang-hui ZHOU Yong-jian HU Li-ling TAN. Study on performance of typical source camera classification algorithms [J]. Journal of Computer Applications, 2011, 31(04): 1133-1137.
[10]	. Image copy detection based on SVD in wavelet domain [J]. Journal of Computer Applications, 2010, 30(4): 917-920.
[11]	. Source camera identification schemes with color image information [J]. Journal of Computer Applications, 2010, 30(10): 2694-2697.
[12]	. Source camera identification technique using large components of imaging sensor pattern noise [J]. Journal of Computer Applications, 2010, 30(1): 31-35.
[13]	Zhen ZHANG Jiquan Kang Xijian Ping Yuan Ren. Blind detection of image splicing based on image quality metrics and moment features [J]. Journal of Computer Applications, 2008, 28(12): 3108-3111.