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

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

Abstract

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

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.

Figures/Tables 16

Fig.1

Fig.2

Fig.3

Fig.4

Fig.5

Fig.6

Fig.7

Table 1

Parameter setting table of MRFM"

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

Table 1

Table 2

Network parameter table"

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]

Table 2

Table 3

Accuracy of resampling factor estimation for different structures (%)"

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

Table 3

Table 4

The detection accuracy of different algorithms (%)"

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

Table 4

Table 5

Table 6

Table 7

Table 8

Accuracy of Downscaling Factor Estimation in Gaussian Noise (%)"

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

Table 8

Fig.8

References 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.