CN114444566A - Image counterfeiting detection method and device and computer storage medium - Google Patents
Image counterfeiting detection method and device and computer storage medium Download PDFInfo
- Publication number
- CN114444566A CN114444566A CN202111542176.6A CN202111542176A CN114444566A CN 114444566 A CN114444566 A CN 114444566A CN 202111542176 A CN202111542176 A CN 202111542176A CN 114444566 A CN114444566 A CN 114444566A
- Authority
- CN
- China
- Prior art keywords
- image
- forgery detection
- feature
- detection model
- gray
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 101
- 238000000034 method Methods 0.000 claims abstract description 26
- 238000000605 extraction Methods 0.000 claims abstract description 8
- 230000008569 process Effects 0.000 claims abstract description 8
- 230000004927 fusion Effects 0.000 claims abstract description 7
- 238000012549 training Methods 0.000 claims description 26
- 230000006870 function Effects 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 13
- 238000013528 artificial neural network Methods 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 9
- 238000013527 convolutional neural network Methods 0.000 claims description 8
- 238000012795 verification Methods 0.000 claims description 7
- 238000011176 pooling Methods 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 description 14
- 238000013135 deep learning Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 239000003086 colorant Substances 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000001815 facial effect Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000012952 Resampling Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 239000000470 constituent Substances 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 230000008921 facial expression Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012821 model calculation Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 229920006395 saturated elastomer Polymers 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- 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
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/90—Determination of colour characteristics
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to an image forgery detection method and system. The method comprises the steps of obtaining an original image and a gray image, carrying out feature extraction and feature fusion on the original image and the gray image through a double-flow image forgery detection model to obtain an image feature classification result, carrying out average voting according to the image feature classification result, and obtaining an image forgery detection probability value and an image authenticity result. The original image and the gray level image are subjected to feature extraction and feature information fusion and splitting respectively, and the expression of color information is inhibited when the image counterfeiting detection model processes the image information, so that the model can perform multi-dimensional counterfeiting detection in an image color space detection technology, and the image detection capability of the network can be enhanced by adding a gray level stream network structure in the model.
Description
Technical Field
The present invention relates to the field of image recognition, and in particular, to a method and an apparatus for detecting image forgery, and a computer storage medium.
Background
With the rapid development of the deep network technology, the image counterfeiting cost is lower and lower, and the quality of the generated counterfeit image is higher and higher, so that the counterfeit video and the image in the network are randomly visible. The image forgery detection method is mainly divided into a traditional detection method and a method based on deep learning. The traditional identification algorithm is an algorithm for extracting features according to the statistical characteristics of images, such as a CFA difference detection algorithm, a pattern noise estimation detection algorithm, a resampling trace detection algorithm and the like. The detection algorithm based on deep learning is further classified into a classification detection algorithm based on CNN, such as a network detection algorithm based on ResNet, Xception, and the like, and an algorithm based on a video time series difference, such as a forgery detection algorithm based on LSTM, and the like. The deep learning method mainly extracts features by a CNN method, and the features extracted by the CNN have stronger representation power compared with manual features, so that the image forgery detection algorithm is mainly a deep learning algorithm.
However, the existing deep learning algorithm mainly takes an image forgery detection algorithm as a binary classification problem, a backbone network is used for extracting feature information only aiming at the characteristics of an original image and using a binary classifier to perform image forgery detection, the model usually pays attention to color information along with continuous data fitting of the model, and the attention to the color information is little, or after the feature information is extracted through processing an obtained gray image, whether the image is forged or not is judged by identifying a fusion boundary generated in the image forgery process of the gray image. The existing deep counterfeiting method mainly classifies four types: identity exchange, facial attribute manipulation, facial expression exchange, full face synthesis, and the like. The methods are continuously iterated and updated, the generated images are more and more vivid and reach the degree that the images are difficult to identify by naked eyes, and the authenticity of the images cannot be effectively verified in the prior art by extracting and identifying the individual image characteristic information, for example, the authenticity of the images cannot be judged by extracting and identifying the image fusion boundary characteristics when the images are forged by a method of facial attribute operation. Therefore, an image forgery detection method is needed to obtain multi-dimensional feature information so as to accurately identify the authenticity of an image.
Disclosure of Invention
In order to solve the problem of inaccurate image detection in the prior art, the invention provides an image forgery detection method, which comprises the following steps,
s1, acquiring an original image, and carrying out graying processing on the original image to obtain a grayscale image;
s2, inputting the original image and the gray image into a pre-trained image forgery detection model to obtain an image feature classification result;
and S3, performing average voting according to the image feature classification result to obtain an image forgery detection probability value and an image authenticity result.
On the basis of the technical scheme, the invention can be improved as follows.
Further, a step of training an image falsification detection model is included before the step of S2,
the method for training the image forgery detection model specifically comprises the following steps,
acquiring an original image, carrying out graying processing to obtain an image data set comprising the original image and a grayscale image, and dividing the image data set into a training set and a verification set according to a preset proportion;
constructing an image forgery detection model based on a deep convolutional neural network;
and performing repeated iterative training on the image counterfeiting detection model by using the training set, verifying the image counterfeiting detection model after each training through a cross entropy loss function and the verification set, and selecting the image counterfeiting detection model with the highest prediction precision as a pre-trained image counterfeiting detection model.
Further, the image forgery detection model constructed based on the deep convolutional neural network specifically comprises the following steps: a plurality of depth convolution neural networks ResNet50 are constructed based on a bottleneck layer, at least two depth convolution neural networks ResNet50 in the plurality of depth convolution neural networks ResNet50 are respectively used for extracting feature information of RGB images and gray level images in image data sets, image feature classification results are obtained according to the feature information, and weight is not shared between the at least two depth convolution neural networks ResNet 50.
Further, in the process of training an image forgery detection model, the image data set can be generalized, specifically including randomly cropping the image and the grayscale image into pictures of different sizes, and performing horizontal flipping or/and vertical flipping or/and random cropping or/and random angle rotation or/and contrast adjustment or/and brightness adjustment to obtain the image data set.
Further, in S1, a graying expression of the grayscale image obtained by graying the original image is as follows:
Gray＝0.299*R+0.587*G+0.114*B；
wherein R, G, B represent the three channels in the RGB color system, respectively.
Further, the step S2 specifically includes the step S201 of inputting the original image and the grayscale image into a feature extraction layer in the image forgery detection model for feature extraction, so as to obtain RGB features F respectively1∈RH ×W×CAnd a gray scale feature F2∈RH×W×CWherein H, W, C denotes the height, width and channel of the image feature, respectively;
s202, after the RGB features and the gray features are fused, the fused features are subjected to batch normalization operation through a batch normalization layer and nonlinear mapping operation through an activation function, wherein the fusion expression of the RGB features and the gray features is as follows:
s203, according to the characteristic size pair F12Splitting, and taking the first half of the characteristic diagram as F3∈R(H/2)×(W)×(2C)The second half of the feature map is F4∈R(H/2)×(W)×2C；
S204, adding F1、F2、F3And F4And performing feature flattening through the global average pooling layer, and obtaining an image feature classification result through a classification layer in the image counterfeiting detection model.
Another object of the present invention is to provide an image-forgery-detection apparatus, including a memory and a processor, wherein the memory stores at least one program, and the at least one program is executed by the processor to implement the image-forgery-detection method.
It is a further object of the present invention to provide a computer-readable storage medium comprising a memory having stored therein a computer program which, when executed by a processor, implements the image-forgery-detection method described above.
The technical scheme of the invention has the beneficial effects that: the original image and the gray image after the gray processing are respectively subjected to feature extraction, the image forgery detection model is used for fusing and splitting feature information of the original image and the gray image after the gray processing, the significant difference information on the image information is obtained, the feature information is supplemented through the gray image, and the expression of color information is inhibited when the image forgery detection model processes the image information, so that the image recognition model can pay attention to information except for colors at the same time, and the RGB feature information and the gray information are fused to better express the features of the forgery image. The model can perform forgery detection in multiple dimensions by using an image color space detection technology, and the image detection capability of the network can be enhanced by adding a gray scale stream network structure in the model.
Drawings
FIG. 1 is a flow chart of an image forgery detection method according to the present invention;
FIG. 2 is a schematic diagram of an image forgery detection model according to the present invention;
FIG. 3 is a flowchart of a method for obtaining image features according to the present invention.
Detailed Description
The principles and features of this invention are described below in conjunction with the following drawings, which are set forth by way of illustration only and are not intended to limit the scope of the invention.
As shown in fig. 1, the present invention provides a technical solution to the above technical problem as follows:
s1, acquiring an original image, and carrying out graying processing on the original image to obtain a grayscale image;
s2, inputting the original image and the gray image into a pre-trained image forgery detection model to obtain an image feature classification result;
and S3, performing average voting according to the image feature classification result to obtain the image forgery detection probability and the image authenticity result.
The technical scheme of the invention is that the original image and the gray image after gray processing are respectively subjected to feature extraction, the image forgery detection model is used for fusing and splitting feature information of the original image and the gray image, the significant difference information on the image information is obtained, the feature information is supplemented through the gray image, and the expression of color information is inhibited when the image forgery detection model processes the image information, so that the image recognition model can pay attention to information except for colors at the same time, and the RGB feature information and the gray information are fused to better express the features of the forged image. The model can perform forgery detection in multiple dimensions by using an image color space detection technology, and the image detection capability of the network can be enhanced by adding a gray scale stream network structure in the model.
Specifically, in the present embodiment, the present invention performs forgery detection on an authenticity image by combining RGB image information and grayscale image information. An image forgery detection model is constructed based on a ResNet network, a network frame of the image forgery detection model is shown in FIG. 2, and the image forgery detection model adopted by the invention comprises two depth convolution neural networks ResNet50 based on Bottleneck (Bottleneck layer), which are respectively used for extracting the characteristic information of RGB images and gray level images and do not share weight. The deep neural network ResNet50 is constructed based on the bottleneck layer, and the bottleneck layer is of a one-dimensional structure, so that compared with a residual structure, the network depth is deepened while the parameters of the constructed image counterfeiting detection model are reduced, the model calculation amount is greatly reduced, and the image counterfeiting detection model is relatively easy to train. Acquiring an original image, carrying out graying processing on the original image to obtain a grayscale image, converting an RGB image into the grayscale image through graying processing, wherein the graying expression is as follows:
Gray＝0.299*R+0.587*G+0.114*B；
r, G, B represent the three channels of RGB, respectively, and 0.299, 0.587, and 0.114 are grayed weights, respectively. Respectively extracting the characteristics of the original image and the gray image through a multilayer convolution layer of an image forgery detection model to respectively obtain RGB characteristics F1∈RH×W×CAnd a gray scale feature F2∈RH×W×CWhere H, W, C represent the height, width, and channel of an image feature, respectively. Respectively carrying out average pooling on the RGB characteristics and the gray characteristics through a global average pooling layer, and predicting through a classification layer to obtain F1、F2The probability of different classes of sample mapping, that is, the true-false classification probability of the RGB image and the gray-scale image prediction result.
In order to enable the image forgery detection model to carry out multi-dimensional forgery detection in the image color space detection technology, the RGB characteristics and the gray level characteristics are fused, the fused characteristics are subjected to batch normalization operation through a batch normalization layer and nonlinear mapping operation is carried out through an activation function, the activation function is mainly used for carrying out nonlinear mapping operation on the convolution layer output result, at present, functions such as ReLU (rectified Linear Unit), sigmoid, tanh and the like are mainly used, the invention takes a ReLU (rectified Linear Unit) rectification Linear unit as a better embodiment, has the characteristics of nonlinearity and linearity in a specific range of a deep neural network, for all inputs greater than 0, the corresponding gradients have a constant derivative value, so that the information integration capability is greatly enhanced, the training of the network is accelerated and simplified, and the problems of gradient loss and overfitting of the CNN model can be effectively reduced. In addition, Batch Normalization (BN), which is an effective method for normalization layer by layer, changes the parameter distribution of each layer of the model during the forward propagation of the batch data in the model, which results in a reduction in training speed, and requires good parameter initialization, making it very difficult to train a model with saturated nonlinearity, therefore, batch normalization needs to be performed on the hidden layer in the neural network, and the input batch data is scaled and translated to a stable mean and standard deviation by transformation during each training of the model, so that each dimension of net input is normalized to a standard normal distribution, thereby improving optimization efficiency, without particularly intending parameter initialization. The BN layer can enhance the adaptability of the model to input different distributions, has a slight regularization effect, accelerates the convergence rate and the training speed of the model, effectively improves the convergence rate of the model, can relieve the gradient dispersion effect in deep network training to a certain extent, and plays a certain role in improving the network generalization performance, thereby ensuring that the deep network model is easier and more stable to train. In addition, the BN can be used for accelerating the training speed of the model and improving the precision of the model.
In this embodiment, as shown in fig. 3, the RGB feature and the gray scale feature are fused, and the fusion expression is:
F12∈R(H)×(W)×(2C)wherein
Representing a cascade of channels, according to the characteristic size pair F12Splitting, and taking the first half of the characteristic diagram as F3∈R(H/2)×(W)×(2C)The second half of the feature map is F4∈R(H/2)×(W)×2CThen F is mixed3And F4Respectively through a global average pooling layerPerforming average pooling to obtain F by classification layer prediction3、F4Maps the probabilities of the different classes. Preprocessing the image to obtain a gray level image, inputting the RGB image and the gray level image into a double-current ResNet50 network structure for model reasoning, and obtaining four classification results. Specifically, in this embodiment, an image forgery detection model is trained, an original image is obtained and subjected to graying processing, an image data set including the original image and a grayscale image is obtained, and the image data set is divided into a training set and a verification set according to a preset proportion; dividing an image data set of an original image and a gray image into a training set and a verification set according to a preset proportion; constructing an image forgery detection model based on the ResNet50 network; and performing repeated iterative training on the image counterfeiting detection model by using the training set, verifying the image counterfeiting detection model after each training through a cross entropy loss function and the verification set, and selecting the image counterfeiting detection model with the highest prediction precision as a pre-trained image counterfeiting detection model. And selecting a Cross-entropy Loss Function (Cross-entropy Loss Function) to define the Cross entropy between the real value and the predicted value. The cross entropy represents the distance between the actual output and the expected output, and the smaller the value of the cross entropy is, the closer the two are. Each feature is constrained by a corresponding cross entropy loss function. And iteratively updating the optimized learning parameters by minimizing the objective function, wherein the smaller the Loss is, the closer the predicted value is to the true value. Each feature is constrained by a corresponding cross-entropy loss function, as shown in the following equation,
Lossall＝0.25*Loss1+0.25*Loss2+0.25*Loss3+0.25*Loss4
the super-parameters before loss can be adjusted, and according to the previous experimental result, the super-parameters are all set to be 0.25, so that the corresponding constraints can be made on the network no matter whether the RGB characteristics or the gray characteristics or the fused characteristics, and the network cannot pay too much attention to certain characteristic information because of the influence of the size of the super-parameters.
And carrying out average voting according to the four classification results to obtain a final classification and a true and false detection probability value.
Prob0＝0.25*prob10+0.25*prob20+0.25*prob30+0.25*prob40,
Prob1＝0.25*prob11+0.25*prob21+0.25*prob31+0.25*prob41,
Wherein Prob0、Prob1Represents the probabilities, prob, of four classifier classes 0 and 1, respectively10、prob20、prob30、prob40Respectively representing the probability values of 0 of the four classifier categories. class denotes the final prediction class if Prob1>Prob0The prediction result is 1, indicating that the image is a counterfeit image and the probability is Prob1Otherwise, the image is represented as a real image with a probability Prob0。
In this embodiment, in the process of training an image forgery detection model, the image data set is generalized, the image and the image grayscale image are randomly cropped into pictures of different sizes, and horizontal flipping, vertical flipping, random cropping, random angle rotation, contrast adjustment, or brightness adjustment is performed to obtain the image data set. The method comprises the steps of collecting original images under various environments, generalizing the images to expand data, and performing data expansion on detected target images on a collection search website, wherein an image counterfeiting detection model obtained by performing iterative training on a data set constructed by frame images after data expansion has strong generalization capability, can perform counterfeiting detection on images at different angles and different types under various scenes, and can effectively avoid the problems of blurring, overturning, inclining and the like of the original images caused by various insurability in the actual application process, so that the YOLOv3 improves the target detection model, and even cannot detect the target for the target detection error.
The invention also provides an image forgery detection device, which comprises a memory and a processor, wherein the memory stores at least one program, and the at least one program is executed by the processor to realize the image forgery detection method. As an executable solution, the image falsification detection apparatus may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The system/electronic device may include, but is not limited to, a processor, a memory. It will be understood by those skilled in the art that the above-described constituent structures of the system/electronic device are only examples of the system/electronic device, and do not constitute a limitation on the system/electronic device, and may include more or less components than those described above, or some components in combination, or different components. For example, the system/electronic device may further include an input/output device, a network access device, a bus, and the like, which is not limited in this embodiment of the present invention.
Further, as an executable solution, the Processor may be a Central Processing Unit (CPU), other general purpose Processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), an off-the-shelf Programmable Gate Array (FPGA) or other Programmable logic device, a discrete Gate or transistor logic device, a discrete hardware component, and the like. The general purpose processor may be a microprocessor or the processor may be any conventional processor or the like, the processor being the control center for the system/electronic device and various interfaces and lines connecting the various parts of the overall system/electronic device.
The memory may be used to store computer programs and/or modules that the processor implements by running or executing the computer programs and/or modules stored in the memory and invoking data stored in the memory, various functions of the system/electronic device. The memory can mainly comprise a program storage area and a data storage area, wherein the program storage area can store an operating system and an application program required by at least one function; the storage data area may store data created according to the use of the mobile phone, and the like. In addition, the memory may include high speed random access memory, and may also include non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) Card, a Flash memory Card (Flash Card), at least one magnetic disk storage device, a Flash memory device, or other volatile solid state storage device.
The present invention also provides a computer-readable storage medium, in which a computer program is stored, and the computer program, when executed by a processor, implements the steps of the above-mentioned method according to the embodiment of the present invention. The system/electronic device integrated modules/units, if implemented in the form of software functional units and sold or used as separate products, may be stored in a computer readable storage medium. Based on such understanding, all or part of the flow in the method according to the embodiments of the present invention may also be implemented by a computer program to instruct related hardware, where the computer program may be stored in a computer readable storage medium, and when the computer program is executed by a processor, the computer program may implement the steps of the embodiments of the method. Wherein the computer program comprises computer program code, which may be in the form of source code, object code, an executable file or some intermediate form, etc. The computer readable medium may include: any entity or device capable of carrying computer program code, recording medium, U-disk, removable hard disk, magnetic disk, optical disk, computer Memory, Read-only Memory (ROM), Random Access Memory (RAM), software distribution medium, and the like. It should be noted that the computer readable medium may contain content that is appropriately increased or decreased as required by legislation and patent practice in the jurisdiction.
The reader should understand that in the description of this specification, reference to the description of the terms "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the invention. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described method embodiments are merely illustrative, and for example, the division of steps into only one logical functional division may be implemented in practice in another way, for example, multiple steps may be combined or integrated into another step, or some features may be omitted, or not implemented.
While the invention has been described with reference to specific embodiments, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (8)
1. An image forgery detection method characterized by: comprises the following steps of (a) carrying out,
s1, acquiring an original image, and carrying out graying processing on the original image to obtain a grayscale image;
s2, inputting the original image and the gray image into a pre-trained image forgery detection model to obtain an image feature classification result;
and S3, performing average voting according to the image feature classification result, and acquiring an image forgery detection probability value and an image authenticity result.
2. An image forgery detection method according to claim 1, characterized by: a step of training the image forgery detection model is further included before the step of S2,
acquiring an original image, carrying out graying processing to obtain an image data set comprising the original image and a grayscale image, and dividing the image data set into a training set and a verification set according to a preset proportion;
constructing an image forgery detection model based on a deep convolutional neural network;
and performing iterative training on the image counterfeiting detection model by using the training set, verifying the image counterfeiting detection model after each training through a cross entropy loss function and the verification set, and selecting the image counterfeiting detection model with the highest prediction precision as a pre-trained image counterfeiting detection model.
3. An image forgery detection method according to claim 2, characterized by: in the process of training the image forgery detection model, the image data set can be generalized, specifically including randomly cutting the original and the gray level images into pictures with different sizes, and performing horizontal turning or/and vertical turning or/and random cutting or/and random angle rotation or/and contrast adjustment or/and brightness adjustment to obtain the image data set.
4. An image forgery detection method according to claim 2, wherein: the method for constructing the image forgery detection model based on the deep convolutional neural network specifically comprises the following steps: a plurality of depth convolution neural networks ResNet50 are constructed based on a bottleneck layer, at least two depth convolution neural networks ResNet50 in the plurality of depth convolution neural networks ResNet50 are respectively used for extracting feature information of RGB images and gray level images in image data sets, image feature classification results are obtained according to the feature information, and weight is not shared between the at least two depth convolution neural networks ResNet 50.
5. An image forgery detection method according to claim 1, characterized by: in S1, the original image is grayed to obtain a grayscale image, and the graying expression is as follows: gray 0.299R + 0.587G + 0.114B;
wherein R, G, B represent the three channels in the RGB color system, respectively.
6. An image forgery detection method according to claim 1, characterized by: the S2 specifically includes the following steps,
s201, inputting the original image and the gray image into a feature extraction layer in the image forgery detection model respectively for feature extraction, and obtaining RGB features F respectively1∈RH×W×CAnd a gray scale feature F2∈RH×W×CWherein H, W, C denotes the height, width and channel of the image feature, respectively;
s202, performing feature fusion on the RGB features and the gray features, and performing batch normalization operation on the fused features through a batch normalization layer in the image forgery detection model and performing nonlinear mapping operation on an activation function, wherein an expression for fusing the RGB features and the gray features is as follows:
s203, according to the characteristic size pair F12Splitting, and taking the first half of the characteristic image as F3∈R(H/2)×(W)×(2C)The second half of the feature image is taken as F4∈R(H/2)×(W)×2C；
S204, adding F1、F2、F3And F4By global averagingAnd performing characteristic flattening on the uniform pooling layer, and obtaining an image characteristic classification result through a classification layer in the image counterfeiting detection model.
7. An image forgery detection apparatus comprising a memory and a processor, the memory storing at least one program, the at least one program being executed by the processor to implement the image forgery detection method according to any one of claims 1 to 6.
8. A computer-readable storage medium characterized by: comprising a memory in which a computer program is stored which, when being executed by a processor, carries out the image forgery detection method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111542176.6A CN114444566B (en) | 2021-12-16 | 2021-12-16 | Image forgery detection method and device and computer storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111542176.6A CN114444566B (en) | 2021-12-16 | 2021-12-16 | Image forgery detection method and device and computer storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114444566A true CN114444566A (en) | 2022-05-06 |
| CN114444566B CN114444566B (en) | 2024-07-23 |
Family
ID=81363909
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111542176.6A Active CN114444566B (en) | 2021-12-16 | 2021-12-16 | Image forgery detection method and device and computer storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114444566B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115456917A (en) * | 2022-11-11 | 2022-12-09 | 中国石油大学(华东) | Image enhancement method, device, equipment and medium beneficial to accurate target detection |
| CN117238015A (en) * | 2023-08-28 | 2023-12-15 | 浙江大学 | General depth forging detection method based on generation model |
| CN117611923A (en) * | 2024-01-08 | 2024-02-27 | 北京锐融天下科技股份有限公司 | Identification method and system for identity document authenticity |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112990031A (en) * | 2021-03-23 | 2021-06-18 | 大连理工大学 | Method for detecting tampered face video and image based on improved twin network |
| WO2021137946A1 (en) * | 2019-12-30 | 2021-07-08 | Microsoft Technology Licensing, Llc | Forgery detection of face image |
| CN113361474A (en) * | 2021-07-01 | 2021-09-07 | 中国科学技术大学 | Double-current network image counterfeiting detection method and system based on image block feature extraction |
| CN113468954A (en) * | 2021-05-20 | 2021-10-01 | 西安电子科技大学 | Face counterfeiting detection method based on local area features under multiple channels |
| CN113537173A (en) * | 2021-09-16 | 2021-10-22 | 中国人民解放军国防科技大学 | Face image authenticity identification method based on face patch mapping |
| CN113723295A (en) * | 2021-08-31 | 2021-11-30 | 浙江大学 | Face counterfeiting detection method based on image domain frequency domain double-flow network |
-
2021
- 2021-12-16 CN CN202111542176.6A patent/CN114444566B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2021137946A1 (en) * | 2019-12-30 | 2021-07-08 | Microsoft Technology Licensing, Llc | Forgery detection of face image |
| CN112990031A (en) * | 2021-03-23 | 2021-06-18 | 大连理工大学 | Method for detecting tampered face video and image based on improved twin network |
| CN113468954A (en) * | 2021-05-20 | 2021-10-01 | 西安电子科技大学 | Face counterfeiting detection method based on local area features under multiple channels |
| CN113361474A (en) * | 2021-07-01 | 2021-09-07 | 中国科学技术大学 | Double-current network image counterfeiting detection method and system based on image block feature extraction |
| CN113723295A (en) * | 2021-08-31 | 2021-11-30 | 浙江大学 | Face counterfeiting detection method based on image domain frequency domain double-flow network |
| CN113537173A (en) * | 2021-09-16 | 2021-10-22 | 中国人民解放军国防科技大学 | Face image authenticity identification method based on face patch mapping |
Non-Patent Citations (2)
| Title |
|---|
| 量子孤岛: "RGB转换成灰度图像的一个常用公式Gray = R*0.299 + G*0.587 + B*0.114", pages 1, Retrieved from the Internet <URL:https://blog.scdn.net/fly_wt/article/details/86432886> * |
| 马颖哲;孙劲光;: "一种多视角人脸检测方法", 计算机工程与科学, no. 01, 15 January 2011 (2011-01-15), pages 1 - 4 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115456917A (en) * | 2022-11-11 | 2022-12-09 | 中国石油大学(华东) | Image enhancement method, device, equipment and medium beneficial to accurate target detection |
| CN117238015A (en) * | 2023-08-28 | 2023-12-15 | 浙江大学 | General depth forging detection method based on generation model |
| CN117238015B (en) * | 2023-08-28 | 2024-08-02 | 浙江大学 | General depth forging detection method based on generation model |
| CN117611923A (en) * | 2024-01-08 | 2024-02-27 | 北京锐融天下科技股份有限公司 | Identification method and system for identity document authenticity |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114444566B (en) | 2024-07-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| De Rezende et al. | Exposing computer generated images by using deep convolutional neural networks | |
| CN112686812B (en) | Bank card inclination correction detection method and device, readable storage medium and terminal | |
| CN114444566A (en) | Image counterfeiting detection method and device and computer storage medium | |
| CN111444881A (en) | Fake face video detection method and device | |
| CN108197644A (en) | A kind of image-recognizing method and device | |
| CN111783085B (en) | Defense method and device for resisting sample attack and electronic equipment | |
| CN114444565B (en) | Image tampering detection method, terminal equipment and storage medium | |
| CN113792659B (en) | Document identification method and device and electronic equipment | |
| CN113344110B (en) | Fuzzy image classification method based on super-resolution reconstruction | |
| CN112926595B (en) | Training device of deep learning neural network model, target detection system and method | |
| CN114724218A (en) | Video detection method, device, equipment and medium | |
| CN112215076B (en) | Deep handwriting identification method and device based on double-tower network | |
| El-Gayar et al. | A novel approach for detecting deep fake videos using graph neural network | |
| CN113111880A (en) | Certificate image correction method and device, electronic equipment and storage medium | |
| CN112232336A (en) | Certificate identification method, device, equipment and storage medium | |
| CN114842478A (en) | Text area identification method, device, equipment and storage medium | |
| Singh et al. | SDCN2: A shallow densely connected CNN for multi-purpose image manipulation detection | |
| CN112613341A (en) | Training method and device, fingerprint identification method and device, and electronic device | |
| Xu et al. | Application of Neural Network in Handwriting Recognition | |
| Richards et al. | Deep Fake Face Detection using Convolutional Neural Networks | |
| CN117197543A (en) | Network anomaly detection method and device based on GMD imaging and improved ResNeXt | |
| CN114445877A (en) | Intelligent face changing method and device and computer storage medium | |
| CN116391189A (en) | Object sequence identification method, network training method, device, equipment and medium | |
| Alzamil | Image Forgery Detection with Machine Learning | |
| Mahale et al. | Copy-Move Image Forgery Detection Using Discrete Wavelet Transform |
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 |