基于多尺度前馈融合结构的重采样因子估计算法

doi:10.11871/jfdc.issn.2096-742X.2023.06.007

数据与计算发展前沿 ›› 2023, Vol. 5 ›› Issue (6): 67-80.

CSTR: 32002.14.jfdc.CN10-1649/TP.2023.06.007

doi: 10.11871/jfdc.issn.2096-742X.2023.06.007

基于多尺度前馈融合结构的重采样因子估计算法

郭静(),张玉金^*(),江智呈,孙冉

上海工程技术大学，电子电气工程学院，上海 201620

收稿日期:2022-05-20 出版日期:2023-12-20 发布日期:2023-12-25
通讯作者: *张玉金（E-mail: yjzhang@sues.edu.cn）
作者简介:郭静，上海工程技术大学，电子电气工程学院，硕士研究生，主要研究方向为图像处理、图像取证以及计算机视觉。
本文中负责提出总体算法，并对算法的可行性进行理论分析和实验验证，撰写论文。
GUO Jing, is a graduate student studying in the School of Electronic and Electrical Engineering at Shanghai University of engineering and technology. Her main research interests are image processing, image forensics, and computer vision.
In this paper, she is responsible for proposing the overall algorithm, theoretically analyzing and experimentally verifying the feasibility of the algorithm, and writing the paper.
E-mail: bella_guojing@163.com|张玉金，上海工程技术大学，电子电气工程学院，博士，副教授，主要研究方向为多媒体取证、图像处理与模式识别。
本文中负责对算法提出相关建议，对实验过程以及结果进行监督。
ZHANG Yujin, Ph.D., is an associate professor at the School of Electronic and Electrical Engineering, Shanghai University of engineering and technology. His main research interests include multimedia forensics, image processing, and pattern recognition.
In this paper, he is responsible for putting forward relevant suggestions on the algorithm and supervising the experimental process and results.
E-mail: yjzhang@sues.edu.cn
基金资助:
上海市自然科学基金(17ZR1411900);上海市科委重点项目(18511101600);上海高校青年教师培养资助计划项目(ZZGCD 15090);上海市信息安全综合管理技术研究重点实验室项目(AGK2015006)

A Estimation Algorithm of Resampling Factor Based on Multi-Level Feedforward Fusion Structure

GUO Jing(),ZHANG Yujin^*(),JIANG Zhicheng,SUN Ran

School of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai 201620, China

Received:2022-05-20 Online:2023-12-20 Published:2023-12-25

摘要/Abstract

摘要：

【目的】 重采样是掩盖图像篡改痕迹的重要手段，为了更加精确地实现对重采样缩放参数的检测，验证图像信息的真实性，本文提出一种基于多尺度前馈融合结构的重采样因子估计算法。【方法】 在预处理层中，首先使用两个线性高通滤波器得到重采样图像的残差特征，抑制图像内容带来的影响，放大区域内像素之间的关联性，其次利用4个低阶高通滤波器在不同方向上强化像素的梯度特征，该算法的主体结构为卷积神经网络，在网络的不同层级处提取出多尺度重采样分类痕迹，结合注意力机制，形成多尺度残差融合模块（Multiscale Residual Fusion Module, MRFM），补偿卷积过程中重采样信息的丢失，标定特征信息传递过程中的有效性，同时去除信息冗余，加速网络收敛。【结果】 实验表明，本文所提算法的网络增益由预处理层和多尺度残差融合模块共同决定，准确性明显高于对比的其他算法，尤其在强噪声的干扰下，本文所提算法具有明显的优势。

关键词: 重采样因子估计, 高通滤波器, 卷积神经网络, 多尺度残差融合

Abstract:

[Objective] Resampling is an important measure to cover the traces of image tampering. In order to accurately detect resampling scaling parameters and verify the authenticity of image information, this paper proposes a resampling factor estimation algorithm based on the multi-scale feed-forward fusion structure. [Methods] In the preprocessing layer, the residual characteristics of the resampled image are obtained by using two linear high-pass filters, the impact of image content is suppressed, the correlation between pixels in the region is enlarged, and then the gradient characteristics of pixels are strengthened in different directions by using four low-order high-pass filters. The main structure of the algorithm is a convolutional neural network, and multi-scale resampling classification traces are extracted at different levels of the network, combined with the attention mechanism. The Multiscale Residual Fusion Module (MRFM) is formed to compensate for the loss of resampled information during convolution, achieve effective transmission of the calibration characteristic information, and remove information redundancy to accelerate network convergence.[Results] Experiments show that the network gain of the proposed algorithm is determined by the pretreatment layer and the multi-scale residual fusion module, and the accuracy is significantly higher than that of other algorithms in comparison, especially under the condition of strong noise interference of strong. The proposed algorithm is of obvious advantages.

Key words: resampling factor estimation, high-pass filter, convolution neural network, multi-scale residual fusion

郭静, 张玉金, 江智呈, 孙冉. 基于多尺度前馈融合结构的重采样因子估计算法[J]. 数据与计算发展前沿, 2023, 5(6): 67-80.

GUO Jing, ZHANG Yujin, JIANG Zhicheng, SUN Ran. A Estimation Algorithm of Resampling Factor Based on Multi-Level Feedforward Fusion Structure[J]. Frontiers of Data and Computing, 2023, 5(6): 67-80, https://cstr.cn/32002.14.jfdc.CN10-1649/TP.2023.06.007.

图/表 16

图1

图2

图3

图4

图5

图6

图7

表1

MRFM的参数设定表"

MRFM	Filter size	Number of filters	Number of output channels
MRFM1	$7 × 7$	16	32
	$5 × 5$	16
	$5 × 5$	32
MRFM2	$5 × 5$	32	64
	$3 × 3$	32
	$3 × 3$	64
MRFM3	$3 × 3$	64	128
	$1 × 1$	64
	$1 × 1$	128

表1

表2

网络参数表"

Input size	Filter	Filter size	Number of filters	activation function	Output	Output size
[64,256,256,18]	Convolution	$5 × 5$	24	ReLU	$F 1, F 2$	[64,256,256,20]
[64,256,256,20]	Max pooling	-	-	-	$F 3$	[64,128,128,20]
[64,128,128,20]	Conv	$5 × 5$	32	ReLU	$F 4, F 5$	[64,128,128,32]
[64,128,128,32]	Max pooling	-	-	-	$F 6$	[64,64,64,32]
[64,64,64,32]	Conv	$3 × 3$	64	ReLU	$F 7, F 8$	[64,64,64,64]
[64,64,64,64]	Max pooling	-	-	-	$F 9$	[64,32,32,64]
[64,32,32,64]	Conv	$3 × 3$	64	ReLU	$F 10, F 11$	[64,32,32,64]
[64,32,32,64]	Max pooling	-	-	-	$F 12$	[64,16,16,64]
[64,16,16,64]	Conv	$3 × 3$	128	ReLU	$F 13$	[64,16,16,128]
[64,16,16,128]	Conv	$3 × 3$	128	Tanh	$F 14$	[64,16,16,128]
[64,16,16,128]	Concat	-	-	-	$F 15$	[64,16,16,256]
[64,16,16,256]	Average pooling	-	-	-	$F 16$	[64,8,8,256]

表2

表3

不同结构下重采样因子估计的准确性(%)"

Algorithm	$λ$
Algorithm	0.5	0.6	0.7	0.8	0.9	1	1.1	1.2	1.3	1.4	1.5	1.6
CNN	46.7	50.4	54.2	59.7	61.1	71.5	71.8	81.5	82.4	85.1	77.7	81.8
No MRFM	66.4	69.8	78.3	86.7	84.3	90.6	92.4	90.3	91.1	94.5	92.1	92.0
Ours	76.2	76.7	87.0	96.2	99.6	99.7	99.9	98.4	99.1	99.6	99.5	99.1

表3

表4

不同算法的估计准确性(%)"

Algorithm	$λ$
Algorithm	0.5	0.6	0.7	0.8	0.9	1	1.1	1.2	1.3	1.4	1.5	1.6
S1	61.5	69.2	79.5	87.6	88.4	94.2	93.2	94.1	95.1	97.4	97.3	97.9
S2	67.4	77.8	88.6	90.7	95.3	98.6	95.4	96.3	99.2	98.5	99.1	97.6
S3	71.2	76.3	87.2	94.1	94.8	98.8	93.2	97.5	96.4	99.1	99.2	99.2
Ours	80.7	82.4	91.3	97.2	99.7	99.8	99.9	99.1	99.5	99.7	99.6	99.4

表4

表5

表6

表7

表8

高斯白噪声下对下采样因子估计的准确性(%)"

Algorithm	$λ$
Algorithm	0.5	0.55	0.6	0.65	0.7	0.75	0.8	0.85	0.9
S2	38.2	42.1	46.6	51.4	52.6	49.7	59.2	67.4	70.3
S3	47.4	45.2	54.4	57.4	59.1	61.4	67.4	68.5	72.4
Ours	67.8	67.4	72.3	76.4	73.6	76.8	77.3	82.4	86.7

表8

图8

参考文献 22

[1]	MAIER A, LORCH B, RIESS C. Toward reliable models for authenticating multimedia content: Detecting resampling artifacts with bayesian neural networks[C]// 2020 IEEE International Conference on Image Processing (ICIP), IEEE, 2020: 1251-1255.
[2]	FERREIRA W D, FERREIRA C B R, DA CRUZ JÚNIOR G, et al. A review of digital image forensics[J]. Computers & Electrical Engineering, 2020, 85: 106685.
[3]	LUO S, LIU J, XU W, et al. Upscaling factor estimation on pre-JPEG compressed images based on difference histogram of spectral peaks[J]. Signal Processing Image Communication, 2021, 94(7): 116223. doi: 10.1016/j.image.2021.116223
[4]	KUZNETSOV A V, MYASNIKOV V V. Copy-move image forensics detection[J]. Computer Optics, 2013, 37(2): 244-253. doi: 10.18287/0134-2452-2013-37-2-244-253
[5]	ZHU D, ZHOU Z. Resampling Tamper Detection Based on JPEG Double Compression[C]// 2014 Fourth International Conference on Communication Systems and Network Technologies. IEEE, 2014: 914-918.
[6]	FENG X, COX I J, DOËRR G. An energy-based method for the forensic detection of re-sampled images[C]// 2011 IEEE International Conference on Multimedia and Expo. IEEE, 2011: 1-6.
[7]	POPESCU A C, FARID H. Exposing digital forgeries by detecting traces of resampling[J]. IEEE Transactions on Signal Processing, 2005, 53(2): 758-767. doi: 10.1109/TSP.2004.839932
[8]	SONG G S, YUN Y I, LEE W H. A new estimation approach of resampling factors using threshold-based peak detection[C]// 2011 IEEE International Conference on Consumer Electronics (ICCE). IEEE, 2011: 731-732.
[9]	LIU X, LU W, ZHANG Q, et al. Downscaling factor estimation on pre-jpeg compressed images[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2019, 30(3): 618-631. doi: 10.1109/TCSVT.76
[10]	LIANG Y, FANG Y, LUO S, et al. Image Resampling Detection Based on Convolutional Neural Network[C]// 2019 15th International Conference on Computational Intelligence and Security (CIS), 2019: 257-261.
[11]	PENG L, X LIAO, CHEN M. Resampling parameter estimation via dual-filtering based convolutional neural network[J]. Multimedia Systems, 2021, 27(2): 1-8. doi: 10.1007/s00530-020-00688-z
[12]	LUO S, LUO J, LU W, et al. Resampling Factor Estimation via Dual-stream Convolutional Neural Network[J]. Computers, Materials and Continua, 2020, 66(1): 647-657. doi: 10.32604/cmc.2020.012869
[13]	王露笛, 周晓光. 基于残差神经网络的充血性心力衰竭识别方法[J]. 科研信息化技术与应用, 2018, 9(6): 3-9.
[14]	肖建平, 龙春, 赵静, 等. 基于深度学习的网络入侵检测研究综述[J]. 数据与计算发展前沿, 2021, 3(3): 59-74.
[15]	WANG Q, WU B, ZHU P, et al. ECA-Net: Efficient Channel Attention for Deep Convolutional Neural Networks[C]// 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), IEEE, 2020: 11531-11539.
[16]	HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]// Proceedings of the IEEE conference on computer vision and pattern recognition. 2018: 7132-7141.
[17]	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: 10.1109/TIFS.2012.2190402
[18]	BAS P, FILLER T, T PEVNÝ. “Break Our Steganographic System”: The Ins and Outs of Organizing BOSS[C]// Proceedings of the 13th international conference on Information hiding, Springer Berlin Heidelberg, 2011: 59-70.
[19]	SCHAEFER G, STICH M. UCID: an uncompressed color image database[C]// Storage and Retrieval Methods and Applications for Multimedia 2004, International Society for Optics and Photonics, 2003: 472-480.
[20]	GLOE T, HME R. The Dresden Image Database for Benchmarking Digital Image Forensics[J]. Journal of Digital Forensic Practice, 2010, 3(2-4): 150-159. doi: 10.1080/15567281.2010.531500
[21]	GAO H, YU S, ZHUANG L, et al. Deep Networks with Stochastic Depth[J]. Springer International Publishing, 2016(9908):646-661.
[22]	BAYAR B, STAMM M C. A Generic Approach Towards Image Manipulation Parameter Estimation Using Convolutional Neural Networks[C]// Acm Workshop on Information Hiding & Multimedia Security, ACM, 2017: 147-157.

基于多尺度前馈融合结构的重采样因子估计算法

A Estimation Algorithm of Resampling Factor Based on Multi-Level Feedforward Fusion Structure

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 16

参考文献 22

相关文章 7

编辑推荐

Metrics

本文评价

True/Det	0.5	0.6	0.7	0.8	0.9	1.0	1.1	1.2	1.3	1.4	1.5	1.6
0.5	85.4	11.5	3.1	0	0	0	0	0	0	0	0	0
0.6	8.2	87.2	4.1	0.2	0	0	0.1	0.1	0	0	0.1	0
0.7	0.6	2.3	94.4	2.5	0.1	0	0.1	0	0	0	0	0
0.8	0.2	0	0.4	98.1	0	0.7	0.5	0	0	0.1	0.1	0
0.9	0	0	0	0.1	99.8	0.1	0	0	0	0	0	0
1.0	0	0	0	0	0	99.9	0.1	0	0	0	0	0
1.1	0	0	0	0	0	0	100	0	0	0	0	0
1.2	0	0	0	0	0	0	0.2	99.8	0	0	0	0
1.3	0	0	0	0	0	0.1	0	0	99.8	0	0	0.1
1.4	0	0	0	0	0	0	0.2	0	0	99.8	0	0
1.5	0	0	0	0	0	0	0	0	0	0.2	99.7	0.1
1.6	0	0	0	0	0.1	0	0.2	0	0	0	0.1	99.6

True/Det	0.5	0.6	0.7	0.8	0.9	1.0	1.1	1.2	1.3	1.4	1.5	1.6
0.5	80.5	10.4	5.6	2.5	0.4	0.1	0.3	0.2	0	0	0	0
0.6	7.8	83.3	2.1	5.2	0.3	0	1.3	0	0	0	0	0
0.7	1.2	4.4	92.5	0	0	0	0.9	1.0	0	0	0	0
0.8	0	0.6	1.5	97.3	0.3	0.3	0	0	0	0	0	0
0.9	0	0	0	0.1	99.7	0	0.2	0	0	0	0	0
1.0	0	0	0	0.1	0	99.8	0	0.1	0	0	0	0
1.1	0	0	0	0	0	0	99.9	0	0	0.1	0	0
1.2	0	0	0.2	0	0.1	0	0.3	99.0	0.2	0	0	0.2
1.3	0	0	0	0	0	0	0.1	0.3	99.6	0	0	0
1.4	0	0	0	0	0	0	0	0.1	0.2	99.7	0	0
1.5	0	0	0	0	0	0.2	0.1	0	0.1	0	99.5	0.1
1.6	0	0	0	0	0	0	0	0.1	0.2	0	0.2	99.5

True/Det	0.5	0.6	0.7	0.8	0.9	1.0	1.1	1.2	1.3	1.4	1.5	1.6
0.5	76.2	10.5	7.7	2.3	1.7	0.8	0	0.2	0	0.5	0.1	0
0.6	14.2	76.7	4.6	2.3	1.7	0.3	0	0.1	0	0	0.1	0
0.7	4.1	7.1	87.0	0.6	1.1	0.1	0	0	0	0	0	0
0.8	0	0.7	1.4	96.2	1.2	0	0.2	0.1	0	0	0.1	0.1
0.9	0	0	0.1	0	99.6	0.1	0	0	0	0.1	0.1	0
1.0	0	0	0	0	0.1	99.7	0.2	0	0	0	0	0
1.1	0	0	0	0	0	0.1	99.9	0	0	0	0	0
1.2	0	0	0	0.1	0.2	0.8	0.3	98.4	0.2	0	0	0
1.3	0	0	0	0	0.1	0	0.3	0.4	99.1	0	0.1	0
1.4	0	0	0	0	0	0	0	0	0.4	99.6	0	0
1.5	0	0	0	0	0	0	0.4	0.1	0	0	99.5	0
1.6	0	0	0	0	0	0	0	0	0	0.3	0.6	99.1

[1]	陈栋, 李明, 陈淑文. 结合Transformer和多层特征聚合的高光谱图像分类算法[J]. 数据与计算发展前沿, 2023, 5(3): 138-151.
[2]	石雪梅,朱克亮,张祥民,张树涛,陈良锋. 基于生成对抗网络的有遮挡人脸修复方法[J]. 数据与计算发展前沿, 2022, 4(4): 123-131.
[3]	肖楠,周明珠,邢军,罗泽,李晓辉. 基于高分辨率网络和注意力机制的真伪卷烟包装鉴别[J]. 数据与计算发展前沿, 2021, 3(5): 118-129.
[4]	张猛,李健. 鸟类音频数据预处理方法[J]. 数据与计算发展前沿, 2021, 3(5): 130-140.
[5]	陈子健,李俊,岳兆娟,赵泽方. 基于自编码器与属性信息的混合推荐模型[J]. 数据与计算发展前沿, 2021, 3(3): 148-155.
[6]	刘晓东,倪浩然. 深度学习技术在学科融合研究中的应用[J]. 数据与计算发展前沿, 2020, 2(5): 99-109.
[7]	欧阳与点,谢鲲. 网络性能数据恢复算法[J]. 数据与计算发展前沿, 2020, 2(3): 55-65.