CN111127407A - Fourier transform-based style migration counterfeit image detection device and method - Google Patents
Fourier transform-based style migration counterfeit image detection device and method Download PDFInfo
- Publication number
- CN111127407A CN111127407A CN201911263941.3A CN201911263941A CN111127407A CN 111127407 A CN111127407 A CN 111127407A CN 201911263941 A CN201911263941 A CN 201911263941A CN 111127407 A CN111127407 A CN 111127407A
- Authority
- CN
- China
- Prior art keywords
- image
- frequency domain
- detection
- domain image
- fourier transform
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20048—Transform domain processing
- G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a style migration forged image detection device based on Fourier transform, which comprises a Fourier transform module, a frequency domain image set and a motion estimation module, wherein the Fourier transform module is used for constructing the frequency domain image set; a first part detection feature construction module for constructing a first part detection feature; a second part detection feature construction module for constructing a second part detection feature; the model training module is used for merging the two parts of detection features into a complete feature vector and performing model training by using the labels of the training image set to obtain a classifier; and the forged image detection module is used for merging the two detection features of the detected image, detecting by using the obtained classifier and giving a detection result. By researching the style migration counterfeiting detection in the Fourier frequency domain, the method can supplement the conventional counterfeit image detection method and improve the difficulty of image counterfeiting.
Description
Technical Field
The invention belongs to the technical field of image processing, and particularly relates to a style migration forged image detection device and method based on Fourier frequency domain statistical characteristics and gradient characteristics.
Background
Style migration is an image processing technique that can transform any given image to a specified style. Since the 90 s of the 20 th century, there were many researchers working on style migration algorithms and proposed many methods that were effective for specific textures. In recent years, with the rise of convolutional neural networks, a series of methods for performing style migration using convolutional neural networks have emerged, and remarkable effects are obtained. Images generated by a plurality of style migration technologies have higher similarity in style with real images, and reliable identification is difficult to perform by using a traditional identification means. The appearance of the style migration image enriches the processing mode of the image on one hand and reduces the threshold of image counterfeiting on the other hand.
Currently, a series of counterfeit image detection methods exist for conventional image counterfeit methods including copy movement, stitching, and image patching. However, these conventional forged image detection methods focus more on local modification of images, and do not have good applicability to global image modification such as style migration. The method for detecting forgery specially aiming at neural network style migration is a relatively new research field, most of the existing work is based on color characteristics of image space domain signals, and the neural network style migration forgery detection work based on Fourier frequency domain characteristics is not available.
Disclosure of Invention
The invention aims to provide a device and a method for detecting a style migration forged image based on Fourier frequency domain statistical characteristics and gradient characteristics, and provides a new idea for detecting the style migration forged image. Firstly, carrying out Fourier transformation on a training image set, then carrying out statistical distribution modeling on a frequency domain image set obtained by transformation, and calculating statistical characteristics corresponding to the statistical distribution; meanwhile, smoothing the frequency domain image set, and extracting gradient features of the smoothed frequency domain image; and finally, training a classifier model by using the obtained features and the training image set labels, thereby obtaining the detection device capable of detecting the style migration forged images.
The invention provides a style migration forged image detection device based on Fourier transform, which comprises:
the Fourier transform module is used for constructing a frequency domain image set corresponding to a training image set containing a real image and a style migration forged image and constructing a frequency domain image corresponding to an image to be detected;
the amplitude value sample set building module is used for obtaining a pixel value distribution diagram corresponding to the Fourier frequency domain image;
the first part detection feature construction module is used for extracting each statistic feature of the frequency domain image pixel distribution diagram so as to construct a first part detection feature;
the second part detection feature construction module is used for extracting gradient features of the frequency domain image so as to construct second part detection features;
the model training module is used for merging the first part of detection features and the second part of detection features corresponding to the training image set into a complete feature vector, and performing model training by using the labels of the training image set to obtain a classifier with smaller generalization error;
and the forged image detection module is used for merging the first part detection characteristic and the second part detection characteristic of the detection image, detecting by using the obtained classifier and giving a detection result.
The invention also provides a method for detecting the style migration forged image based on Fourier transform, which comprises the following steps:
s1, constructing a training image set containing real images and style migration forged images;
s2, carrying out Fourier transform on the images in the training image set to obtain a corresponding frequency domain image set;
s3, constructing an amplitude value sample set of each frequency domain image in the frequency domain image set;
s4, extracting the statistic characteristics of the amplitude distribution of each frequency domain image according to the amplitude value sample set of each frequency domain image in the frequency domain image set constructed in the step S3, and constructing a first part of detection characteristics;
s5, extracting the gradient characteristics of each frequency domain image in the frequency domain image set, and constructing a second part of detection characteristics;
s6, combining the detection characteristics of the step S4 and the step S5 to obtain complete detection characteristics, and training to obtain a classifier based on the characteristics and labels of the images in the training image set;
s7, carrying out Fourier transform on the image to be detected to obtain a frequency domain image to be detected;
s8, extracting the statistic characteristics of the frequency domain image to be detected;
s9, extracting the gradient characteristics of the frequency domain image to be detected;
and S10, combining the statistic characteristics of the image to be detected extracted in the step S8 and the gradient characteristics of the frequency domain image to be detected extracted in the step S9, and inputting the combined characteristics into the classifier obtained in the step S6 to obtain a detection result.
Further, the training image set constructed in step S1 includes a subset of real art images and a subset of style migration forged images, and the image to be detected in step S7 includes a real art image or a style migration forged image.
Further, for a three-channel color image, the color channel used for fourier transform in step S2 includes the luminance channel of the YUV color space.
Further, the fourier transform is a two-dimensional fast fourier transform in step S2.
Further, the statistical features in step S4 and step S8 include, but are not limited to, mean, variance, skewness coefficient, and kurtosis coefficient.
Further, the gradient feature extraction methods in step S5 and step S9 include, but are not limited to, a gradient direction histogram.
Further, in steps S5 and S9, before extracting the gradient feature, the input image is smoothed by, but not limited to, median filtering or gaussian filtering.
Further, the classifier obtained in step S6 includes, but is not limited to, a support vector machine or a neural network.
The invention has the beneficial effects that:
1) by researching the style migration counterfeiting detection in the Fourier frequency domain, the method can supplement the conventional counterfeit image detection method and improve the difficulty of image counterfeiting.
2) The style migration forgery detection method can effectively distinguish the real image from the style migration forgery image.
3) The method and the device can identify the images generated by the multiple style migration methods.
Drawings
Fig. 1 is a block configuration diagram of a fourier transform-based genre migration-counterfeit image detection apparatus according to the present invention.
Fig. 2 is a flowchart of the training phase of the fourier transform-based method for detecting a stylized migration-forged image of the present invention.
Fig. 3 is a flowchart of the detection stage of the fourier transform-based method for detecting a stylized migration-forged image according to the present invention.
Detailed Description
The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings and embodiments.
The style migration forged image detection device based on Fourier transform, as shown in FIG. 1, comprises a Fourier transform module for constructing a frequency domain image set corresponding to a training image set containing a real image and a style migration forged image, and constructing a frequency domain image corresponding to an image to be detected. The amplitude value sample set construction module is used for acquiring a pixel value distribution diagram corresponding to the Fourier frequency domain image; the first part detection feature construction module is used for extracting each statistic feature of the frequency domain image pixel distribution diagram so as to construct a first part detection feature; the second part detection feature construction module is used for extracting gradient features of the frequency domain image so as to construct second part detection features; the model training module is used for merging the first part of detection features and the second part of detection features corresponding to the training image set into a complete feature vector, and performing model training by using the labels of the training image set to obtain a classifier with smaller generalization error; the fake image detection module is used for merging the first part detection feature and the second part detection feature of the detection image, detecting by using the obtained classifier and giving a detection result.
Based on the detection device, the method for detecting the style migration forged image based on the Fourier transform comprises a training stage and a detection stage, as shown in FIG. 2, wherein the training stage comprises the following steps:
1) and constructing a training image set comprising the real image and the style migration forged image, and representing the label of the training image set by {0, 1}, wherein 1 represents the style migration forged image and 0 represents the real image.
2) And carrying out two-dimensional fast Fourier transform on the images in the training image set to obtain a corresponding frequency domain image set. The specific process of performing two-dimensional fast Fourier transform is as follows: and (3) performing two-dimensional fast Fourier transform on each image of the training image set, taking the amplitude spectrum obtained by the transform, performing absolute value taking and logarithm taking on the amplitude spectrum, translating the 0 frequency to the center of the amplitude spectrum, and normalizing the obtained amplitude spectrum to [0, 255 ].
3) And constructing a sample set of amplitude values of each image in the frequency domain image set.
4) Calculating and extracting statistic characteristics of amplitude value distribution of each image according to the amplitude value sample set of each image obtained in the step 3), and constructing a first part detection characteristic. The specific process for constructing the first part of detection features comprises the following steps: and respectively calculating statistics such as a mean value, a variance, a skewness coefficient, a kurtosis coefficient and the like of the frequency domain image according to the statistical distribution of the amplitude values of the frequency domain image.
5) And (3) extracting the gradient characteristics of the images in the frequency domain image set in the step 2) and constructing the second part of detection characteristics. The specific process for constructing the second part detection features comprises the following steps: and performing Gaussian filtering or median filtering on the frequency domain image, and extracting the gradient characteristics of the smoothed frequency domain image by constructing a directional gradient histogram.
6) And combining the two detection features of each image into a complete feature vector to obtain a complete feature set, carrying out Z-Score standardization on the feature vector, inputting the feature vector and a label of each training image into a classifier model (such as log probability regression, a support vector machine or a neural network), and training by using a corresponding optimization algorithm.
As shown in fig. 3, the detection phase comprises the following steps:
1) and carrying out Fourier transform on the image to be detected to obtain a frequency domain image to be detected.
2) And extracting the statistic features of the frequency domain image to be detected based on the statistic feature extraction mode in the training step 4).
3) And extracting gradient features from the frequency domain image to be detected based on the gradient feature extraction mode in the training step 5).
4) Combining the characteristics of the images to be detected obtained in the steps, and inputting the characteristics into the classifier obtained in the training step 6) to obtain a detection result.
The invention is further illustrated by the following specific examples.
1) 753 real images and 528 style migration forged images are taken to form a training image set, wherein the label of the real images is marked as 0, and the label of the style migration forged images is marked as 1.
2) For three-channel color images in the training image set, calculating a Y channel of the three-channel color images in a YUV space as an input image matrix of Fourier transform, wherein the formula is as follows:
Y＝0.299·R+0.587·G+0.114·B
r, G, B represents the pixel values corresponding to the R channel, G channel, and B channel, respectively.
And carrying out two-dimensional fast Fourier transform on each image in the training image set, wherein the formula is as follows:
wherein, (u, v) refers to the coordinates of the frequency domain plot; f (u, v) represents an amplitude value at the frequency domain image coordinate point (u, v); (x, y) refers to the coordinates of the original image; f (x, Y) represents the pixel value of the original image Y channel at (x, Y); m and N respectively refer to the row number and the column number of the frequency domain image matrix; i represents an imaginary unit.
And taking the amplitude spectrum obtained by transformation, carrying out absolute value taking and logarithm taking on the amplitude spectrum, translating the 0 frequency to the center of the amplitude spectrum, and normalizing the obtained amplitude spectrum to [0, 255 ].
3) And respectively constructing a corresponding amplitude value sample set P for each frequency domain graph, wherein the formula is as follows:
P＝{F(u,v)|u＝0,1,…,N-1；v＝0,1,…,M-1}
4) constructing a first part detection characteristic:
for the obtained amplitude value sample set P, the following statistics are calculated respectively:
the sample mean is used to describe the central trend of the distribution of amplitude values, and the formula is as follows:
where P represents an element in the sample set of magnitude values P.
Variance s2For describing the degree of dispersion of the amplitude value distribution, the formula is as follows:
skewness coefficient g1The distribution form for describing the distribution of the amplitude values is as follows:
wherein m is2Means second order sample center moment, m3Refers to the third order sample central moment.
Crest factor g2The distribution form for describing the distribution of the amplitude values is as follows:
wherein m is4Refers to the fourth order sample central moment.
5) Constructing a second part of detection characteristics:
performing smoothing operation on the frequency domain image by using Gaussian filtering or median filtering, then performing directional gradient histogram calculation based on the following steps, and extracting gradient features:
a) calculating gradient values of all pixel points;
b) calculating a direction histogram for each cell;
c) combining the unit cells into a larger block, and performing normalization operation in the block;
d) and combining the histogram features in all the blocks, namely obtaining the final gradient feature.
6) Model training:
inputting the feature vectors and the labels extracted from the training image set into a support vector machine model for training, selecting the model by using a cross-validation method, and obtaining the optimal hyper-parameters (including kernel function, cost parameter and gamma parameter) by using a grid search method.
7) Identifying the image to be detected:
fourier transform is carried out on the image to be detected to obtain a frequency domain image, and the statistic characteristics and gradient characteristics of the frequency domain image are respectively extracted based on the steps 4) and 5); splicing the obtained features to obtain a complete feature vector; inputting the obtained feature vector into the support vector machine trained in the step 6) for detection, and obtaining a detection result.
In order to verify the effectiveness and the practicability of the invention, 753 real images and 528 style migration forged images are used as training image sets, and classifiers are trained according to steps 1) to 6). And performing model evaluation by using a test image set which is intersected with the training image set to form an empty set, wherein the test image set comprises 322 real images and 226 style migration forged images, classifying the test image set by using a classifier obtained by training, and comparing the test image set with real labels of the training image set, so that the accuracy of the prediction result of the classifier is 79.04%, and the method is effective and feasible.
It will be apparent to those skilled in the art that various modifications and improvements can be made to the embodiments of the present invention without departing from the inventive concept thereof, and these modifications and improvements are intended to be within the scope of the invention.
Claims (10)
1. A Fourier transform-based style migration-based counterfeit image detection device is characterized by comprising:
the Fourier transform module is used for constructing a frequency domain image set corresponding to a training image set containing a real image and a style migration forged image and constructing a frequency domain image corresponding to an image to be detected;
the amplitude value sample set building module is used for obtaining a pixel value distribution diagram corresponding to the Fourier frequency domain image;
the first part detection feature construction module is used for extracting each statistic feature of the frequency domain image pixel distribution diagram so as to construct a first part detection feature;
the second part detection feature construction module is used for extracting gradient features of the frequency domain image so as to construct second part detection features;
the model training module is used for merging the first part of detection features and the second part of detection features corresponding to the training image set into a complete feature vector, and performing model training by using the labels of the training image set to obtain a classifier;
and the forged image detection module is used for merging the first part detection characteristic and the second part detection characteristic of the detection image, detecting by using the obtained classifier and giving a detection result.
2. A style migration forged image detection method based on Fourier transform is characterized by comprising the following steps:
s1, constructing a training image set containing real images and style migration forged images;
s2, carrying out Fourier transform on the images in the training image set to obtain a corresponding frequency domain image set;
s3, constructing an amplitude value sample set of each frequency domain image in the frequency domain image set;
s4, extracting the statistic characteristics of the amplitude distribution of each frequency domain image according to the amplitude value sample set of each frequency domain image in the frequency domain image set constructed in the step S3, and constructing a first part of detection characteristics;
s5, extracting the gradient characteristics of each frequency domain image in the frequency domain image set, and constructing a second part of detection characteristics;
s6, combining the detection characteristics of the step S4 and the step S5 to obtain complete detection characteristics, and training to obtain a classifier based on the characteristics and labels of the images in the training image set;
s7, carrying out Fourier transform on the image to be detected to obtain a frequency domain image to be detected;
s8, extracting the statistic characteristics of the frequency domain image to be detected;
s9, extracting the gradient characteristics of the frequency domain image to be detected;
and S10, combining the statistic characteristics of the image to be detected extracted in the step S8 and the gradient characteristics of the frequency domain image to be detected extracted in the step S9, and inputting the combined characteristics into the classifier obtained in the step S6 to obtain a detection result.
3. The method according to claim 2, wherein the training image set constructed in step S1 includes a subset of real art images and a subset of style migration forged images, and the images to be detected in step S7 include real art images or style migration forged images.
4. The method according to claim 2, wherein for a three-channel color image, the color channels used for fourier transform in step S2 comprise the luminance channels of the YUV color space.
5. The method of claim 2, wherein the fourier transform is a two-dimensional fast fourier transform in step S2.
6. The method according to claim 2, wherein the set of amplitude values of each frequency domain image in the set of frequency domain images constructed in step S3 is P:
P＝{F(u,v)|u＝0,1,…,N-1；v＝0,1,…,M-1}
wherein, (u, v) refers to coordinates of the frequency domain image; f (u, v) represents an amplitude value at the frequency domain image coordinate point (u, v); m and N refer to the number of rows and columns, respectively, of the frequency domain image matrix.
7. The method of claim 2, wherein the statistical features in step S4 and step S8 include a mean, a variance, a skewness coefficient, and a peak-skewness coefficient.
8. The method of claim 2, wherein the gradient feature extraction method in steps S5 and S9 comprises a histogram of gradient directions.
9. The method of claim 2, wherein in steps S5 and S9, before extracting the gradient feature, the input image is smoothed by a median filter or a gaussian filter.
10. The method of claim 2, wherein the classifier obtained in step S6 comprises a support vector machine or a neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911263941.3A CN111127407B (en) | 2019-12-11 | 2019-12-11 | Fourier transform-based style migration forged image detection device and method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911263941.3A CN111127407B (en) | 2019-12-11 | 2019-12-11 | Fourier transform-based style migration forged image detection device and method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111127407A true CN111127407A (en) | 2020-05-08 |
| CN111127407B CN111127407B (en) | 2022-06-28 |
Family
ID=70498496
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911263941.3A Active CN111127407B (en) | 2019-12-11 | 2019-12-11 | Fourier transform-based style migration forged image detection device and method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111127407B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114267089A (en) * | 2022-03-03 | 2022-04-01 | 中国科学院自动化研究所 | Method, device and equipment for identifying forged image |
| WO2024048891A1 (en) * | 2022-08-31 | 2024-03-07 | 숭실대학교 산학협력단 | Device and method for detecting adversarial example on basis of symbolic feature of image, and computer-readable recording medium in which program for executing same is recorded |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106373115A (en) * | 2016-08-22 | 2017-02-01 | 天津大学 | Smoothed filter operation type and parameter recognition method |
| CN107103327A (en) * | 2017-05-26 | 2017-08-29 | 中国科学院信息工程研究所 | Image detecting method is forged in a kind of dyeing based on Color Statistical difference |
| US20180232605A1 (en) * | 2017-02-16 | 2018-08-16 | Honeywell International Inc. | Determining image forensics using gradient statistics at edges |
| CN108830823A (en) * | 2018-03-14 | 2018-11-16 | 西安理工大学 | The full-reference image quality evaluating method of frequency-domain analysis is combined based on airspace |
| CN109214394A (en) * | 2018-08-02 | 2019-01-15 | 中国科学院信息工程研究所 | It is a kind of that image detecting method and device are forged based on the Style Transfer of color and texture features |
-
2019
- 2019-12-11 CN CN201911263941.3A patent/CN111127407B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106373115A (en) * | 2016-08-22 | 2017-02-01 | 天津大学 | Smoothed filter operation type and parameter recognition method |
| US20180232605A1 (en) * | 2017-02-16 | 2018-08-16 | Honeywell International Inc. | Determining image forensics using gradient statistics at edges |
| CN107103327A (en) * | 2017-05-26 | 2017-08-29 | 中国科学院信息工程研究所 | Image detecting method is forged in a kind of dyeing based on Color Statistical difference |
| CN108830823A (en) * | 2018-03-14 | 2018-11-16 | 西安理工大学 | The full-reference image quality evaluating method of frequency-domain analysis is combined based on airspace |
| CN109214394A (en) * | 2018-08-02 | 2019-01-15 | 中国科学院信息工程研究所 | It is a kind of that image detecting method and device are forged based on the Style Transfer of color and texture features |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114267089A (en) * | 2022-03-03 | 2022-04-01 | 中国科学院自动化研究所 | Method, device and equipment for identifying forged image |
| CN114267089B (en) * | 2022-03-03 | 2022-05-31 | 中国科学院自动化研究所 | Method, device and equipment for identifying forged image |
| WO2024048891A1 (en) * | 2022-08-31 | 2024-03-07 | 숭실대학교 산학협력단 | Device and method for detecting adversarial example on basis of symbolic feature of image, and computer-readable recording medium in which program for executing same is recorded |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111127407B (en) | 2022-06-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Babu et al. | Statistical features based optimized technique for copy move forgery detection | |
| Li et al. | Apple leaf disease identification and classification using resnet models | |
| Wang et al. | iVAT and aVAT: enhanced visual analysis for cluster tendency assessment | |
| CN102663391B (en) | Image multifeature extraction and fusion method and system | |
| CN104268593B (en) | The face identification method of many rarefaction representations under a kind of Small Sample Size | |
| CN104778481A (en) | Method and device for creating sample library for large-scale face mode analysis | |
| CN108197644A (en) | A kind of image-recognizing method and device | |
| CN102194114A (en) | Method for recognizing iris based on edge gradient direction pyramid histogram | |
| CN111340824A (en) | Image feature segmentation method based on data mining | |
| CN108682007A (en) | Jpeg image resampling automatic testing method based on depth random forest | |
| CN110991547A (en) | Image significance detection method based on multi-feature optimal fusion | |
| CN111127407B (en) | Fourier transform-based style migration forged image detection device and method | |
| CN108090513A (en) | Multi-biological characteristic blending algorithm based on particle cluster algorithm and typical correlation fractal dimension | |
| Ma et al. | Multiscale 2-D singular spectrum analysis and principal component analysis for spatial–spectral noise-robust feature extraction and classification of hyperspectral images | |
| CN112070116B (en) | Automatic artistic drawing classification system and method based on support vector machine | |
| Shire et al. | A review paper on: agricultural plant leaf disease detection using image processing | |
| CN111325158B (en) | CNN and RFC-based integrated learning polarized SAR image classification method | |
| Rajesh et al. | Intelligent Detection of Disease in Plant Leaf using SVM | |
| Zhang et al. | Saliency detection via image sparse representation and color features combination | |
| CN116310452B (en) | Multi-view clustering method and system | |
| CN117079272A (en) | Bullet bottom socket mark feature identification method combining manual features and learning features | |
| CN105701499A (en) | Image processing method used for brain MRI image classification | |
| CN112818779B (en) | Human behavior recognition method based on feature optimization and multiple feature fusion | |
| Zayed et al. | A New Refined-TLBO Aided Bi-Generative Adversarial Network for Finger Vein Recognition | |
| Li et al. | HEp-2 cells staining patterns classification via wavelet scattering network and random forest |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |