Detecting median filtering via two-dimensional AR models of multiple filtered residuals

Yang, Jianquan; Ren, Honglei; Zhu, Guopu; Huang, Jiwu; Shi, Yun Q.

doi:10.1007/s11042-017-4691-0

Cited by 40 publications

(67 citation statements)

References 28 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To fairly compare F M F R with the state-of-the-art, including the AR method (10-D), [1] GLF method (56-D) [5] and the 2DAR method (81-D), [9] the same experimental setups are employed. The feature set obtained from transition probability F 1 M (n = 4, T = 1, 70-D) is added to verify complementary effect between transition probability and AR model.…”

Section: Comparison Proposed Methods With Prior Artsmentioning

confidence: 99%

“…We further explain it by Yule-Walker equations, which can be used to estimate AR coefficients. In the formula (8), γ(m), 1 ≤ m ≤ p is the auto-correlation and its unbiased estimator is shown in the formula (9). It can be inferred from (9) that large residual elements cause significantly changes on γ(m).…”

Section: Feature Extractionmentioning

confidence: 99%

“…When calculating AR coefficients, no truncation is executed on MFR, thus AR coefficients keep information about MFR elements with large absolute value. On the other hand, we can also suppose that AR coefficient is not sensitive to small value elements from (9) and (10). The flaw of AR model at small value element can be compensated by transition probability.…”

Section: Feature Extractionmentioning

confidence: 99%

“…Yang et al employed two-dimensional autoregressive (2D-AR) model to extract a composite feature set from median filtered residual, average filtered residua and Gaussian filtered residual. [9] Their method outperforms 1-D AR model a lot for JPEG scenario. Chen et al employed CNN model to automatically learn feature for median filtering forensics and achieved great improvements for JPEG compressed image of low resolution.…”

Section: Introductionmentioning

confidence: 96%

See 3 more Smart Citations

Blind Median Filtering Detection Using Auto-Regressive Model and Markov Chain

Peng

Gao

Zeng

2018

IJRC

View full text Add to dashboard Cite

Establishing the processing history of an image is important for robot vision. In this paper, an improved method for median filtering detection is proposed. That is, detect whether an image has been processed by median filtering. First, we analyze the statistical properties of median filtering residual and find that it is suitable for exposing fingerprints of median filtering. Then, the new feature set on median filtering residual is constructed by incorporating transition probability matrices of Markov chain with coefficients of auto-regressive model. A dimensionality reduction method is developed to lower the feature dimensionality. The final feature set is fed into support vector machines to construct a detector. Due to the distinction property of median filtering residual as well as compensated effect between transition probability and auto-regressive model, experimental results on large image database demonstrate that the proposed method is effectively in median filtering detection, even for images with heavy JPEG compression or at a low resolution. The performance of proposed detector outperforms prior arts. Additionally, the proposed method demonstrates good generalization ability.

show abstract

Section: Comparison Proposed Methods With Prior Artsmentioning

confidence: 99%

Section: Feature Extractionmentioning

confidence: 99%

Section: Feature Extractionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 96%

See 2 more Smart Citations

Blind Median Filtering Detection Using Auto-Regressive Model and Markov Chain

Peng

Gao

Zeng

2018

IJRC

View full text Add to dashboard Cite

show abstract

“…In the typical image forensics detection (IFD), the relationship between image pixels is statistically processed, or the frequency and spatial domains of an image are analyzed. So whether the extracted feature vector is based on numerical or frequency spatial, a median filter residual (MFR) [7][8][9][10][11] was used a lot. The MFR is mainly used as a pre-processing step of the suspicion image.…”

Section: Introductionmentioning

confidence: 99%

Composition of Visual Feature Vector Pattern for Deep Learning in Image Forensics

Rhee

2020

IEEE Access

View full text Add to dashboard Cite

In image forensics, to determine whether the image is impurely transformed, it extracts and examines the features included in the suspicious image. In general, the features extracted for the detection of forgery images are based on numerical values, so it is somewhat unreasonable to use in the CNN structure for image classification. In this paper, the extraction method of a feature vector is using a least-squares solution. Treat a suspicious image like a matrix and its solution to be coefficients as the feature vector. Get two solutions from two images of the original and its median filter residual (MFR). Subsequently, the two features were formed into a visualized pattern and then fed into CNN deep learning to classify the various transformed images. A new structure of the CNN net layer was also designed by hybrid with the inception module and the residual block to classify visualized feature vector patterns. The performance of the proposed image forensics detection (IFD) scheme was measured with the seven transformed types of image: average filtered (window size: 3×3), gaussian filtered (window size: 3×3), JPEG compressed (quality factor: {90, 70}), median filtered (window size: {3×3, 5×5}), and unaltered. The visualized patterns are fed into the image input layer of the designed CNN hybrid model. Throughout the experiment, the accuracy of median filtering detection was 98% over. Also, the area under the curve (AUC) by sensitivity (TP: true positive rate) and 1specificity (FP: false positive rate) results of the proposed IFD scheme approached to '1' on the designed CNN hybrid model. Experimental results show high efficiency and performance to classify the various transformed images. Therefore, the grade evaluation of the proposed scheme is "Excellent (A)."

show abstract

A Deep Learning Approach to Detection of Warping Forgery in Images

Yang

Feng

et al. 2020

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Detecting median filtering via two-dimensional AR models of multiple filtered residuals

Cited by 40 publications

References 28 publications

Blind Median Filtering Detection Using Auto-Regressive Model and Markov Chain

Blind Median Filtering Detection Using Auto-Regressive Model and Markov Chain

Composition of Visual Feature Vector Pattern for Deep Learning in Image Forensics

A Deep Learning Approach to Detection of Warping Forgery in Images

Contact Info

Product

Resources

About