Image splicing forgery detection combining coarse to refined convolutional neural network and adaptive clustering

Xiao, Bin; Yang, Wei; Bi, Xiuli; Li, Weisheng; Ma, Jianfeng

doi:10.1016/j.ins.2019.09.038

Cited by 99 publications

(35 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Some researchers [9][10][11][16][17][18][19] have regarded image splicing forgery detection as a binary classification of the image patches. As these detection networks focus only on the local patches and ignore the relationship between the image patches, they easily return false decisions.…”

Section: How To Learn Image Fingerprints Intentionalmentioning

confidence: 99%

“…Wei, et al [10] proposed a two-stage VGG network, and Xiao, et al [11] designed a two-stage network that locates the tampered regions on both coarse and refined scales, and generates the final detected tampered regions by an adaptive clustering approach. These three methods [9][10][11] are performed on image patches, which causes high computational complexity. Moreover, their detection results are either inaccurate or require complex postprocessing.…”

Section: Related Workmentioning

confidence: 99%

“…This subsection compares the performances of the proposed D-Unet and several existing detection methods: color filter array (CFA) [28], noise inconsistency (NOI) [29], aligned double quantization (ADQ) [1], error level analysis (ELA) [30], U-shaped network (U-Net) [24], RRU-Net [15], ManTra [22], LSTM [21], coarse-to-refined network (C2RNet) [11], and FusionNet [19]. The details of each detection method are given in Table IX.…”

Section: Comparative Experiments and Analysismentioning

confidence: 99%

“…C2RNet [11] A two-stage detection method, including a coarse convolutional neural network(C-CNN) and a refined CNN(R-CNN), which learns the differences of the image properties between un-tampered and tampered regions from image patches with different scales.…”

Section: Lstm [21]mentioning

confidence: 99%

See 3 more Smart Citations

RRU-Net: The Ringed Residual U-Net for Image Splicing Forgery Detection

Yang

Xiao

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

Self Cite

119

View full text Add to dashboard Cite

Recently, many detection methods based on convolutional neural networks (CNNs) have been proposed for image splicing forgery detection. Most of these detection methods focus on the local patches or local objects. In fact, image splicing forgery detection is a global binary classification task that distinguishes the tampered and non-tampered regions by image fingerprints. However, some specific image contents are hardly retained by CNN-based detection networks, but if included, would improve the detection accuracy of the networks. To resolve these issues, we propose a novel network called dual-encoder U-Net (D-Unet) for image splicing forgery detection, which employs an unfixed encoder and a fixed encoder. The unfixed encoder autonomously learns the image fingerprints that differentiate between the tampered and non-tampered regions, whereas the fixed encoder intentionally provides the direction information that assists the learning and detection of the network. This dualencoder is followed by a spatial pyramid global-feature extraction module that expands the global insight of D-Unet for classifying the tampered and non-tampered regions more accurately. In an experimental comparison study of D-Unet and state-of-theart methods, D-Unet outperformed the other methods in imagelevel and pixel-level detection, without requiring pre-training or training on a large number of forgery images. Moreover, it was stably robust to different attacks.

show abstract

Section: How To Learn Image Fingerprints Intentionalmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

Section: Comparative Experiments and Analysismentioning

confidence: 99%

Section: Lstm [21]mentioning

confidence: 99%

See 2 more Smart Citations

RRU-Net: The Ringed Residual U-Net for Image Splicing Forgery Detection

Yang

Xiao

et al. 2019

2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

Self Cite

119

View full text Add to dashboard Cite

show abstract

“…Still, they did not find a way to integrate different deep learning methods to fully mine the transformer fault characteristics from different perspectives and further improve the accuracy of fault diagnosis.Convolutional Neural Networks (CNN) and Recurrent Neural Network (RNN) are two of the most important frameworks in deep learning field [20][21][22]. CNN has been applied in mature fields such as visual recognition, image processing, and fault diagnosis [23][24][25]. As an improved model of RNN, long short-term memory (LSTM) was introduced to make up for the defects of long-term memory loss, gradient dissipation or explosion in the feedback process of the RNN model [26,27].…”

mentioning

confidence: 99%

A Deep Parallel Diagnostic Method for Transformer Dissolved Gas Analysis

Duan

2020

Applied Sciences

View full text Add to dashboard Cite

With the development of Industry 4.0, as a pivotal part of the power system, large-capacity power transformers are requiring fault diagnostic methods with higher intelligence, accuracy and anti-interference ability. Considering the powerful capability for extracting non-linear features and the sensitivity differences to features of deep learning methods, this paper proposes a deep parallel diagnostic method for transformer dissolved gas analysis (DGA). In view of the insufficient and imbalanced dataset of transformers, adaptive synthetic oversampling (ADASYN) was implemented to augment the fault dataset. Then, the newly constructed dataset was normalized and input into the LSTM-based diagnostic framework. Then, the dataset was converted into images as the input of the CNN-based diagnostic framework. At the same time, the problem of still insufficient data was compensated by the introduction of transfer learning technology. Finally, the diagnostic models were trained and tested respectively, and the Dempster-Shafer (DS) evidence theory was introduced to fuse the diagnostic confidence matrices of the two models to achieve deep parallel diagnosis. The results of the proposed deep parallel diagnostic method show that without complex feature extraction, the diagnostic accuracy rate could reach 96.9%. Even when the dataset was superimposed with 3% random noises, the rate only decreased by 0.62%.Appl. Sci. 2020, 10, 1329 2 of 18 their shortcomings in learning ability, processing efficiency, and feature extraction ability. For example, the learning ability of the fuzzy methods are not satisfying [7,8]. Neural networks (NN) tend to fall into local optimal solutions [9,10]. The K-nearest neighbor (KNN) method is inefficient in high-dimensional space [11]. The support vector machine (SVM) is essentially a two-classifier, which makes it more troublesome to deal with multi-classification problems [12,13]. There are also some relevant diagnosis studies on combining traditional ratio methods and intelligent methods, which are combinations of previous researches, and have achieved certain effects [14][15][16]. However, the features extracted by the ratio methods are still limited. And the anti-interference ability of the combined models need further testing. In recent years, due to the strong ability of deep learning to extract complex non-linear features, some papers have tried to introduce it into the field of transformer fault diagnosis. And the fault diagnostic accuracy of deep learning methods has been significantly improved compared to traditional machine learning diagnostic models'. The authors in [17] proposed a DBN-based fault diagnostic method that used the uncoded ratio of DGA data as the model input. Compared with the traditional methods', the accuracy was significantly improved. The authors in [18] introduced a method for identifying and locating winding faults based on CNN and transformer impulse tests. And the results showed that the method was effective. The authors in [19] studied the internal fault diagnosis of...

show abstract

Image splicing forgery detection by combining synthetic adversarial networks and hybrid dense U‐net based on multiple spaces

Yang

Wang

et al. 2022

Int J of Intelligent Sys

Self Cite

View full text Add to dashboard Cite

With the popularity of image editing tools, the originality and information security of images are facing serious threats. The most common threat is splicing forgery that copies a part of the area from one donor image to the acceptor one. Some research works were proposed to protect the image originality, whereas they are still difficult to apply in practice.There are two main reasons: (a) very limited data for learning models; (b) huge attribute differences between the donor and acceptor images. We propose two novel tasks to conquer the above challenges: Synthetic Adversarial Networks (SANs) and Hybrid Dense U-Net (HDU-Net). SAN finds the most secluded position for inserting tampered areas in an image by learning the association between scenes and objects, and can enlarge the original small data set by more than 40 times. We call the data set SF-Data generated by SAN. We combine the dense U-Net that detects the differences of the essential attributes of image with four spaces containing more available feature information to propose HDU-Net. Then, the synthetic data set SF-Data are used to train HDU-Net. We perform various attack experiments on several public data sets to demonstrate the effectiveness and robustness of our method.

show abstract

Image splicing forgery detection combining coarse to refined convolutional neural network and adaptive clustering

Cited by 99 publications

References 21 publications

RRU-Net: The Ringed Residual U-Net for Image Splicing Forgery Detection

RRU-Net: The Ringed Residual U-Net for Image Splicing Forgery Detection

A Deep Parallel Diagnostic Method for Transformer Dissolved Gas Analysis

Image splicing forgery detection by combining synthetic adversarial networks and hybrid dense U‐net based on multiple spaces

Contact Info

Product

Resources

About