CN103903271B - Image forensics method for natural image and compressed and tampered image based on DWT - Google Patents
Image forensics method for natural image and compressed and tampered image based on DWT Download PDFInfo
- Publication number
- CN103903271B CN103903271B CN201410144039.0A CN201410144039A CN103903271B CN 103903271 B CN103903271 B CN 103903271B CN 201410144039 A CN201410144039 A CN 201410144039A CN 103903271 B CN103903271 B CN 103903271B
- Authority
- CN
- China
- Prior art keywords
- image
- coefficient
- value
- compression
- tampered
- 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.)
- Expired - Fee Related
Links
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention provides an image forensics method for a natural image and a compressed and tampered image based on DWT. According to the method, the natural image and the compressed image based on the DWT can be effectively distinguished, meanwhile, good distinguishability is achieved on certain specific image tampering carrying out compression trace elimination on the compressed image, the joint probability histogram of a wavelet transform coefficient of the natural image and the tampered image is calculated through the method, the histogram is normalized, then Hough transform is carried out, the mean value, variance value, skewness value and kurtosis value of a Hough transform coefficient matrix are extracted as characteristic values of a support vector machine, and a training set is formed by the characteristic values. A classification model is generated by the support vector machine through the training set in a training mode, unknown characteristic value samples are classified through the model, and whether compression or anti-compression forensics processing is carried out on an image or not is judged. The method is stable in performance, easy and convenient to implement, efficient, high in accuracy and suitable for forensics detection of the natural image and the tampered image in other aspects.
Description
Technical field
The present invention relates to the technical field of image forensics is and in particular to a kind of usurped for natural image with based on dwt compression
Change plan picture image evidence collecting method.
Background technology
Fast development with current multimedia technology and network technology is so that transmission and shared multimedia become more
Convenient, people can be carried out for corresponding multimedia (such as picture, video, audio frequency etc.) by network and multimedia technology soon
The acquisition of speed, does not house and doubts, we live in the world of a vision purpose, and seeing is believing is that our traditional ideas are recognized
Know.But digital technology at this stage bring we huge convenient while, also distort along with for multimedia, wherein
Most importantly image is distorted.The false of picture material is likely to bring certain harm to the life of people,
Such as South China Tiger event, causes tremendous influence to entire society.
Traditional picture material protection is all to be designed according to external mode, for example add in the picture watermark or
Fingerprint.But under many circumstances, external protection scheme can not be implemented it is therefore desirable to for image itself well
Intrinsic fingerprint characteristic is studied, thus promoting effectively carrying out of evidence obtaining work.But, science is constantly present two-sidedness,
After some researchers discuss out that the evidence obtaining for a kind of distorted image detects, have researcher and carry out accordingly for this detection
Cover attack process so that existing evidence obtaining detection method lost efficacy.For example, in the compression evidence obtaining for image, there is research
Person is compressed and uncompressed classification according to the characteristics of image after compression, so that it is guaranteed that the authenticity of image.Meanwhile,
There is researcher and be directed to the anti-evidence obtaining of compression of these methods so that existing compression evidence collecting method lost efficacy.Therefore, seek nature
Intrinsic fingerprint characteristic in image just becomes particularly important.
In digital evidence obtaining, the compression evidence obtaining for image and the anti-evidence obtaining of compression focus primarily upon jpeg image.Because
In today's society, people are jpeg forms using the coded system of most cameras.For jpeg image compression algorithm
Detection in, such as farid etc. is (referring to h.farid.digital ballistics from jpeg quantization:a
Follow up study.dept.comp.sci..dartmouth college, tech.rep.tr2008-638,2008.) root
According to different camera guns using different jpeg quantization tables so that in image quantization to be detected and the database that existed
The quantization table of different camera lenses contrasted, thus detecting that this image is the information being shot by which kind of camera.Fan etc.
(referring to z.fan and r.dequeiroz.identification of bitmap compression history:jpeg
detection and quantizer estimation.ieee trans.image process.vol.12,no.2,
Pp.230 235, feb.2003.) using original image dct the continuity of coefficient histogram and compression after image dct
There is the property being spaced to be compressed collecting evidence in coefficient histogram.For compression anti-forensics technology, stamm etc. (referring to
matthew c.stamm,k.j.ray liu.anti-forensics of digital image compression.ieee
Transactions on information forensics andsecurity.vol.6, no.3,2011.) for jpeg pressure
The interval problem between dct coefficient histogram after contracting, carries out modeling using laplace model for histogram, so
Afterwards shake is carried out for compartment to add, thus fill up interval between coefficient histogram so that lin etc. (referring to
w.s.lin,s.k.tjoa,h.v.zhao,and k.j.r.liu.digital image source coder forensics
via intrinsic fingerprints.ieee trans.inf.forensics security.vol.4,no.3,
Pp.460 475, sep.2009.) etc. proposition evidence obtaining algorithm lost efficacy.
However, with the fast development of digital technology, jpeg2000 and spiht(set partitioning in
Hierarchicaltrees, SPIHT algorithm) these image compression algorithms based on dwt have also obtained widely
Application.Compress the forensic technologies of image for dwt, lin etc. is (referring to w.s.lin, s.k.tjoa, h.v.zhao, and
k.j.r.liu.digital image source coder forensics via intrinsic
fingerprints.ieee trans.inf.forensics security.vol.4,no.3,pp.460–475,
Sep.2009.) evidence obtaining detection is carried out using the interval between the coefficient histogram after conversion.Meanwhile, farid etc. (referring to
hany farid and siwei lyu.higher-order wavelet statistics and their
application to digital forensics.ieee workshop on statistical in computer
Vision (in conjunction with cvpr), 2003.) propose using the father after image wavelet transform, between coefficient
Linear relationship between node and child nodes and neighbor node and uncle's node carries out natural image and compression image, or
The thought of the evidence obtaining detection between natural image and computer composograph.For anti-forensics technology, stamm etc. (referring to
m.c.stamm and k.j.r.liu.wavelet-based image compression anti-forensics.in
Proc.ieee int.conf.image process.pp.1737 1740,2010.) it is proposed for the anti-of spiht compression algorithm
Forensic technologies, Main Basiss laplace model carries out the filling of pectination coefficient histogram.
Also obtained extensively using the method that the characteristic of the feature invariance based on natural image carries out image forensics simultaneously
Research.AsDeng (referring to janjessica fridrich.calibration
revisited.mm&sec'09proceedings of the11th acm workshop on multimedia and
Security.pp.63-74,2009.) propose based on natural image after carrying out fraction pixel cutting, overall count special
Property change the technology collected evidence of less characteristic, and valenzise etc. is (referring to giuseppe valenzise, marco
tagliasacchi,stefano tubaro.revealing the traces of jpeg compression anti-
forensics.ieee transactions on information forensics and security.vol.8,no.2,
2013.) anti-forensics technology changing realization based on total variance between different compression ratios (total variation) proposing
Evidence obtaining algorithm.
When collecting evidence to compression of images, the mode collected evidence for distinctive characteristic in specific compression algorithm is held very much
Easily made to cover this characteristic by some way by other researchers, so that evidence obtaining algorithm lost efficacy.And it is directed to the anti-of researcher
Evidence obtaining algorithm, researcher can find out evidence obtaining algorithm further again so that whole evidence obtaining and anti-evidence obtaining enter a kind of vicious circle
Competitive stage.Therefore, image forensics are carried out using characteristic specific to natural image, be a kind of trend of future development and grind
Study carefully focus.For the characteristic of wavelet transformation, seek natural image fingerprint characteristic after the wavelet transform, for different researchers
Compression for image or the anti-evidence obtaining of compression make it all have well can to distinguish row, can effectively ensure the true of image
Property, the game simultaneously as far as possible forgone between evidence obtaining and anti-evidence obtaining, is that more unified side is found in the authenticity detection of general image
Method.
Content of the invention
The technical problem to be solved in the present invention is: overcomes the shortcomings of existing forensic technologies, provides one kind to be directed to natural image
With based on dwt(discrete wavelet transform, wavelet transform) compression tampered image image evidence obtaining side
Method, the method uses the incidence relation between dwt different decomposition classification coefficient, obtains this between natural image and tampered image
The difference of incidence relation, is further processed using Hough transformation, obtains difference characteristic value, realize for natural image with
And the high-quality evidence obtaining of tampered image.
The technical solution adopted for the present invention to solve the technical problems: one kind is usurped for natural image with based on dwt compression
Change plan picture image evidence collecting method, the method can be collected evidence for based on the image of dwt compression, or can be directed to warp
The carrying out crossing the image that anti-evidence obtaining is processed collect evidence it is characterised in that: calculate the wavelet transformation system of natural image and tampered image
The joint probability histogram of number, is normalized for this histogram, then carries out Hough transformation, extracts Hough transformation coefficient square
The average of battle array, variance yields, skewness value and kurtosis value are as the characteristic value of SVMs, the composition training set of this feature value；?
Holding vector machine uses this training set training to generate disaggregated model, using this model, unknown characteristic value sample is classified,
Judge whether image is through distorting, described is process through overcompression or through back-pressure contracting evidence obtaining through distorting；Extract described
The characteristic value of SVMs specifically include following steps:
The first step, carries out wavelet transform to image, obtains corresponding wavelet coefficient, to the discrete wavelet coefficient obtaining
Carry out rounding operation, according to the characteristic of wavelet transformation, picture breakdown is n rank, wherein in n rank, wherein n=1,
2, n 1, different according to wave filter service condition, it can be divided into lnLow frequency sub-band, hnHorizontal subband, dnVertical subband
And vnDiagonal subband, then lnLow frequency sub-band may proceed to be filtered using wave filter, obtains the l of n+1 leveln+1Low frequency sub-band,
hn+1Low frequency sub-band, dn+1Low frequency sub-band and vn+1Low frequency sub-band；
Second step, for the h after wavelet transformation, the subband in tri- directions of v and d carries out joint probability distribution Nogata respectively
Figure calculates；It is assumed that in h on h directionnOne of coefficient hn(r, s), wherein (r, s) represent positional information, define its " father
Coefficient is parent "So coefficient hn(r, s) is just corresponding " child " coefficient；According to this kind of corresponded manner, one
Individual " child " coefficient corresponds to " father " coefficient, so that the relation between this " child " and " father " node is brighter
Aobvious, adopt and with 10 the bottom of for, all of coefficient is taken the logarithm；Then joint probability histogram is asked for " child " and " father " coefficient；
The histogrammic solution of joint probability is carried out using formula (1),
In ηhIn being horizontally oriented of representing of subscript h, what wherein δ represented is target function, defines δ (x, y)=1, when
And if only if x=0 and y=0；Wherein [] represents and rounds, mhRepresent that all " father " and " child " corresponds in h direction
Mapping relations number,WithRepresent the value of " father " and " child " in i level dwt is decomposed；Simultaneously for ηvWith
ηdIt is also adopted by same calculation；ForWithValue, the range of definition be -100, -
99,…,100}；Wherein k, l are in integer set ζ；
3rd step, for the joint probability histogram obtaining, is normalized operation, is carried out using formula (2),
Using this kind of mode, joint probability histogram is normalized to { 0,1 ..., 255 } so as to an ash can be become
Then normalized joint probability histogram is processed by the span of degree image as gray level image；
4th step, the probability histogram for three directions obtaining after normalized is carried out suddenly using formula (3)
Husband's map function, is transformed in the coefficient domain of Hough transformation,
ρ=xcos θ+ysin θ (3)
Wherein ρ represents initial point to the distance of image cathetus, and θ represents the folder of the vertical line perpendicular to image cathetus and x-axis
Angle size；
5th step, the coefficient domain for three different directions obtaining is averaged, variance, skewness and kurtosis, and will
Four characteristic value combinations in three directions, form the characteristic vector of one 12 dimension.
Further, using svm(support vector machine, SVMs) carry out data model instruction
Practice, the characteristic vector of natural image and tampered image is trained, obtains training pattern.
Further, the feature value vector of the image of unknown classification is carried out data training using training pattern, instructed
Practice type.
The principle of the present invention is:
A kind of evidence collecting method of the image for natural image with based on dwt compression tampered image of the present invention, including image
Feature extraction, model training and model use three processes, and image characteristics extraction needs natural image and tampered image are entered
The corresponding feature extraction of row；Extraction process includes the wavelet transformation to image, then obtains coefficient according to corresponding computing formula
Between joint probability histogram, for joint probability histogram carry out part block, obtain the data of most expression power, with
When histogram is normalized, finally carry out Hough transformation, the coefficient after being converted, average, side carried out for coefficient
The calculating of difference, skewness and kurtosis value, the characteristic value in three different directions of joint, obtain the characteristic vector of one 12 dimension；Right
In the characteristic vector training set of the natural image comprising same number bar number and tampered image, using the side of cross validation
Formula, obtains the most optimized parameter of svm grader, thus obtaining training pattern；For the characteristic vector of unknown classification, it is trained
The classification of model, then obtains corresponding type output, obtains the differentiation classification of image；
The process of image characteristics extraction is:
(1) image is carried out dwt wavelet transformation；
(2) for the coefficient obtaining after dwt conversion, take with 10 for low logarithm, and be multiplied by 10, by the coefficient for 0
Logarithm be set as 0；
(3) to the logarithmic coefficient matrix processing obtaining with step (2), using " father " coefficient shown in Fig. 1 and " child
Relation between son " coefficient, respectively obtains dwt conversion father's node in rear three directions and reflecting correspondingly of child nodes
Penetrate relation pair, and limit the value of father's node and child nodes in { -100, -99 ..., 100 }；
(4) the histogrammic calculating of joint probability is carried out using formula (1)；
(5) the joint probability histogram obtaining is normalized using formula (2), is normalized to integer
0,1 ..., 255 } in the range of；
(6) for the joint probability histogram obtaining, carry out rim detection；
(7) data that step (6) obtains is carried out Hough transformation using formula (2), obtain corresponding coefficient domain；
(8) coefficient domain obtaining is carried out the calculating of average, variance, skewness and kurtosis value；
(9) characteristic value in three directions of joint, generates the characteristic vector of one 12 dimension；
Model training process is as follows:
(1) classification design is carried out using svm；
(2) select c-svm as the model of classification；
(3) data in training set is carried out cross validation according to natural image and tampered image data volume identical mode
Model training, then obtain optimum c-svm parameter；
Unknown category classification process is as follows:
(1) image feature data of unknown classification is carried out normalized；
(2) classified using the model of training, obtained classification results；
The present invention is compared with prior art advantageously:
(1) the method applied in the present invention is to be made a distinction based on characteristic specific to natural image, can be good at getting rid of
Except the game between evidence obtaining and anti-evidence obtaining, provide a kind of algorithm with more popularity for later image forensics.
(2) present invention is simple for the feature extraction of image, and the dimension of the characteristic vector ultimately producing is low, for big
During the data of scale, can have more preferable time high efficiency.
(3) feature of present invention point is directed to natural image stability height, has very strong robustness.
Brief description
Fig. 1 is the relation between involved " father " and " child " node in the present invention；
Fig. 2 is the general frame flow chart in the present invention；
Fig. 3 is the compression image forensics flow chart in the present invention；
Fig. 4 is the anti-evidence obtaining compression image forensics flow chart in the present invention；
Fig. 5 is involved anti-evidence obtaining compressed image processing flow chart in the present invention.
Specific embodiment
Below in conjunction with the accompanying drawings and specific embodiment further illustrates the present invention.
Before carrying out the experimental implementation of the present invention, need to be generated and process for experiment image used.This reality
Testing adopted natural image is ucid-v2(g.schaefer and m.stich, " ucid-an uncompressed
colour image database,”in proc.spie:storage and retrieval methods and
Applications for multimedia, 2004, vol.5307, pp.472 480) in 1338 images, for this 1338
Image, needs to obtain compression ratio from 0.5 to 8, and is spaced apart 0.5 16 kinds of compression image and the counter of ad hoc approach takes
Card image.It is the acquisition process of the compression image based on spiht and jpeg2000 first, its acquisition process is as follows:
Step 1: spiht image compression algorithm code is arranged, the compression algorithm of jpeg2000 adopts simultaneously
The algorithm having existed in matlab carries out image procossing；
Step 2: for different images, the rate that is compressed is from 0.5 to 8, and is spaced apart compression image in the 16 of 0.5
Generate, thus obtaining the image of the spiht compression algorithm of the different compression ratios of 1338 × 16, and 1338 × 16 different pressures
The compression image of the jpeg2000 of shrinkage.
Step 3: for different compression ratios and different compression algorithms, different names are carried out for image and deposits
Put operation.
Again, need to obtain the image of the evidence obtaining algorithm process that contracts through back-pressure, acquisition process such as accompanying drawing 5 institute of this image
Show, its acquisition process () as follows taking the image of spiht compression algorithm compression as a example:
Step 1: carry out 9/7dwt conversion for the image between different compression ratios through spiht compression algorithm, obtain
The frequency domain information of respective image；
Step 2: in frequency domain, the wavelet coefficient histogram of different brackets and different sub-band carries out laplace model
Parameter Estimation；
Step 3: according to the model parameter obtaining, carry out the Jitter Calculation at the corresponding coefficient of different brackets different sub-band；
Step 4: add jitter at corresponding coefficient, thus obtaining the wavelet coefficient after processing；
Step 5: anti-9/7dwt process is carried out for the wavelet coefficient processing；
Step 6: carry out image forensics operation, obtain the image of back-pressure compression algorithm process, generate 1338 × 16 warps herein
Cross the image of back-pressure compression algorithm process.
Then, carry out the experimental implementation of the present invention.The detail flowchart of the present invention is as shown in Fig. 2 whole evidence obtaining process
Use three processes including image characteristics extraction, model training and model, wherein of paramount importance part is exactly feature extraction
Process, as the characteristic extraction procedure being marked in Fig. 2；Extraction process carries out wavelet transformation firstly the need of to image, due to
In the standard of jpeg2000, lossy compression method is using 9/7 wavelet transformation, is therefore also adopted by this small echo herein and carries out discrete wavelet transformer
Change, then the calculating of probability histogram and the operation processing further are carried out according to different directions；Characteristic vector has been asked for
After one-tenth, need to carry out the model training of svm, realize parameter Estimation by the way of cross validation, finally for unknown images
Carry out discriminant classification.
In the present invention, for evidence obtaining be compression image and the compression image processing through anti-evidence obtaining, as Fig. 3 and
Shown in Fig. 4, its flow process is the same, and the image simply processing is different, is collectively referred to as tampered image by two herein and carries out whole implementation behaviour
Make the explanation of flow process.Whole implementation step is as follows:
The process of image characteristics extraction is:
Step 1: the image by natural image and after distorting carries out dwt wavelet transformation, is carried out little using 9/7 small echo herein
Wave conversion；
Step 2: the coefficient obtaining is taken with 10 for low logarithm, and is multiplied by 10 and then expands natural image and distort
The otherness of image, the logarithm of the coefficient for 0 is set as 0；
Step 3: to the logarithmic coefficient matrix obtaining with step 2, for three different directions in dwt conversion, according to
Relation between father shown in Fig. 1 and child nodes, finds one one-to-one correspondence pair, and limits father's node and child
The value of child node is in { -100, -99 ..., 100 }；
Step 4: the joint probability histogram of the form calculus different directions according to formula (1)；
Step 5: the joint probability histogram obtaining is normalized using formula (2), is normalized to
Integer 0,1 ..., 255 } in the range of；
Step 6: rim detection is carried out for the joint probability histogram obtaining, is used herein as sobel operator and carries out edge
The calculating of detection；
Step 7: the data that step 5 is obtained carries out Hough transformation using formula (2), obtains corresponding coefficient domain；
Step 8: the coefficient domain obtaining is carried out the calculating of average, variance, skewness and kurtosis value；
Step 9: the characteristic value in three directions of joint, generate the characteristic vector of one 12 dimension；
Model training process is as follows:
Step 1: classification design is carried out using svm, the libsvm being used herein as having designed is trained；
Step 2: select c-svm as the model of classification；
Step 3: the data in training set is intersected according to natural image and tampered image data volume identical mode
The model training of checking, then obtains the parameter of the c-svm of optimum, herein can be using the calculation of the training optimal solution having existed
Method；
Unknown category classification process is as follows:
Step 1: the image feature data of unknown classification is carried out the normalization of libsvm data form；
Step 2: the model using training is classified, and obtains classification results；
The experiment effect of the present invention, this place adopts auc(area under curve, TG-AUC) for the effect tested
Fruit illustrates, and auc, closer to 1, illustrates that experiment effect is better.The experiment effect of the present invention is attained by more than 0.97.
Table 1 shows that the present invention is directed to spiht, and jpeg2000 compresses the auc value that image is collected evidence, and is directed to
The auc value of the evidence obtaining of spiht back-pressure contracting evidence obtaining algorithm process image.
Table 1 image forensics auc value table
Claims (3)
1. a kind of evidence collecting method of the image for natural image with based on dwt compression tampered image, the method can be directed to base
Image in dwt compression is collected evidence, or can collect evidence for the carrying out of the image processing through anti-evidence obtaining, and its feature exists
In: calculate the joint probability histogram of the wavelet conversion coefficient of natural image and tampered image, returned for this histogram
One change, then carries out Hough transformation, extracts the average of Hough transformation coefficient matrix, variance yields, skewness value and kurtosis value as
Hold the characteristic value of vector machine, the composition training set of this feature value；SVMs uses this training set training to generate classification mould
Type, is classified to unknown characteristic value sample using this model, judges whether image is through distorting, described through distorting for
Process through overcompression or through back-pressure contracting evidence obtaining；The characteristic value extracting described SVMs specifically includes following steps:
The first step, carries out wavelet transform to image, obtains corresponding wavelet coefficient, and the discrete wavelet coefficient obtaining is carried out
Rounding operates, and according to the characteristic of wavelet transformation, picture breakdown is n rank, wherein in n rank, wherein n=1,
2, n 1, different according to wave filter service condition, it can be divided into lnLow frequency sub-band, hnHorizontal subband, dnVertical subband
And vnDiagonal subband, then lnLow frequency sub-band may proceed to be filtered using wave filter, obtains the l of n+1 leveln+1Low frequency sub-band,
hn+1Horizontal subband, dn+1Vertical subband and vn+1Diagonal subband；
Second step, for the h after wavelet transformation, the subband in tri- directions of v and d carries out joint probability distribution histogram meter respectively
Calculate；It is assumed that in h on h directionnOne of coefficient hn(r, s), wherein (r, s) represent positional information, define its " father " system
Number isSo coefficient hn(r, s) is just corresponding " child " coefficient；According to this kind of corresponded manner, " a child
Son " coefficient corresponds to " father " coefficient, so that the relation between this " child " and " father " node becomes apparent from, adopts
All of coefficient is taken the logarithm the bottom of in order to 10；Then joint probability histogram is asked for " child " and " father " coefficient；Joint
The solution of probability histogram is carried out using formula (1),
In hhIn being horizontally oriented of representing of subscript h, what wherein δ represented is target function, defines δ (x, y)=1, and if only if
X=0 and y=0；Wherein [] represents and rounds, mhRepresent all " father " and " child " one-to-one mapping in h direction
The number of relation,WithRepresent the value of " father " and " child " in i level dwt is decomposed；Simultaneously for hvAnd hdIt is also adopted by
Same calculation；ForWithValue, the range of definition be { -100, -99 ..., 100 }；Its
Middle k, l are in integer set z；
3rd step, for the joint probability histogram obtaining, is normalized operation, is carried out using formula (2),
Using this kind of mode, joint probability histogram is normalized to { 0,1 ..., 255 } so as to a gray-scale map can be become
Then normalized joint probability histogram is processed by the span of picture as gray level image；
4th step, the probability histogram for three directions obtaining after normalized carries out Hough change using formula (3)
Change operation, be transformed in the coefficient domain of Hough transformation,
ρ=xcos θ+ysin θ (3)
Wherein ρ represents initial point to the distance of image cathetus, and θ represents big perpendicular to the vertical line of image cathetus and the angle of x-axis
Little；
5th step, the coefficient domain for three different directions obtaining is averaged, variance, skewness and kurtosis, and by three
Four characteristic value combinations in direction, form the characteristic vector of one 12 dimension.
2. the evidence obtaining side of a kind of image for natural image with based on dwt compression tampered image according to claim 1
Method, will it is characterised in that carry out the model training of data using svm (support vector machine, SVMs)
The characteristic vector of natural image and tampered image is trained, and obtains training pattern.
3. the evidence obtaining side of a kind of image for natural image with based on dwt compression tampered image according to claim 1
Method, it is characterised in that the feature value vector of the image of unknown classification is carried out data training using training pattern, obtains training class
Type.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410144039.0A CN103903271B (en) | 2014-04-11 | 2014-04-11 | Image forensics method for natural image and compressed and tampered image based on DWT |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410144039.0A CN103903271B (en) | 2014-04-11 | 2014-04-11 | Image forensics method for natural image and compressed and tampered image based on DWT |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103903271A CN103903271A (en) | 2014-07-02 |
| CN103903271B true CN103903271B (en) | 2017-01-18 |
Family
ID=50994578
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410144039.0A Expired - Fee Related CN103903271B (en) | 2014-04-11 | 2014-04-11 | Image forensics method for natural image and compressed and tampered image based on DWT |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103903271B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105513059A (en) * | 2015-12-01 | 2016-04-20 | 北京航空航天大学 | Urban multi-mode media data consistency detection method |
| CN106056639B (en) * | 2016-05-24 | 2019-05-14 | 大连理工大学 | It is a kind of based on the anti-hidden decryption method collected evidence of camera source |
| CN107025647B (en) * | 2017-03-09 | 2020-02-28 | 中国科学院自动化研究所 | Image tampering evidence obtaining method and device |
| CN110555792B (en) * | 2019-08-16 | 2022-05-17 | 广东外语外贸大学南国商学院 | Image tampering blind detection method based on normalized histogram comprehensive feature vector |
| CN113259665B (en) * | 2020-02-07 | 2022-08-09 | 华为技术有限公司 | Image processing method and related equipment |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102184537A (en) * | 2011-04-22 | 2011-09-14 | 西安理工大学 | Image region tamper detection method based on wavelet transform and principal component analysis |
| CN102609947A (en) * | 2012-02-10 | 2012-07-25 | 浙江理工大学 | Forgery detection method for spliced and distorted digital photos |
| CN102968793A (en) * | 2012-11-20 | 2013-03-13 | 百年金海安防科技有限公司 | Method for identifying natural image and computer generated image based on DCT (Discrete Cosine Transformation)-domain statistic characteristics |
| CN103034995A (en) * | 2012-11-18 | 2013-04-10 | 百年金海安防科技有限公司 | Splicing image detecting algorithm based on statistical properties of Benford rule |
-
2014
- 2014-04-11 CN CN201410144039.0A patent/CN103903271B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102184537A (en) * | 2011-04-22 | 2011-09-14 | 西安理工大学 | Image region tamper detection method based on wavelet transform and principal component analysis |
| CN102609947A (en) * | 2012-02-10 | 2012-07-25 | 浙江理工大学 | Forgery detection method for spliced and distorted digital photos |
| CN103034995A (en) * | 2012-11-18 | 2013-04-10 | 百年金海安防科技有限公司 | Splicing image detecting algorithm based on statistical properties of Benford rule |
| CN102968793A (en) * | 2012-11-20 | 2013-03-13 | 百年金海安防科技有限公司 | Method for identifying natural image and computer generated image based on DCT (Discrete Cosine Transformation)-domain statistic characteristics |
Non-Patent Citations (1)
| Title |
|---|
| Image Tamper Detection and Recovery System Based on Discrete Wavelet Transformation;Kwang-Fu Li et al;《Communications, Computers and signal Processing, 2001》;20010828;164 - 167 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103903271A (en) | 2014-07-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103903271B (en) | Image forensics method for natural image and compressed and tampered image based on DWT | |
| Wang et al. | Optimized feature extraction for learning-based image steganalysis | |
| Ye et al. | No-reference image quality assessment using visual codebooks | |
| CN103116763B (en) | A kind of living body faces detection method based on hsv color Spatial Statistical Character | |
| Zhang et al. | A dense u-net with cross-layer intersection for detection and localization of image forgery | |
| CN102917227B (en) | Compressive sensing-based adaptive video information hiding method | |
| WO2017080196A1 (en) | Video classification method and device based on human face image | |
| CN104778457A (en) | Video face identification algorithm on basis of multi-instance learning | |
| CN103037212B (en) | The adaptive block compressed sensing method for encoding images of view-based access control model perception | |
| CN103200421A (en) | No-reference image quality evaluation method based on Curvelet transformation and phase coincidence | |
| CN106056523B (en) | Blind checking method is distorted in digital picture splicing | |
| CN110121109A (en) | Towards the real-time source tracing method of monitoring system digital video, city video monitoring system | |
| CN103561274B (en) | Video time domain tamper detection method for removing moving object shot by static camera lens | |
| Si et al. | A full‐reference stereoscopic image quality assessment index based on stable aggregation of monocular and binocular visual features | |
| CN104715440A (en) | Remote-sensing image perceptual hash authentication method based on Gabor filter bank and DWT converting | |
| CN105227964A (en) | Video-frequency identifying method and system | |
| Luo et al. | Image universal steganalysis based on best wavelet packet decomposition | |
| CN105868778A (en) | Method for detecting median filtering based on frequency domain fluctuation characteristics | |
| CN102968793B (en) | Based on the natural image of DCT domain statistical property and the discrimination method of computer generated image | |
| WO2021088176A1 (en) | Binary multi-band power distribution-based low signal-to-noise ratio sound event detection method | |
| Alaql et al. | Classification of image distortions for image quality assessment | |
| Zhang et al. | Depth combined saliency detection based on region contrast model | |
| Su et al. | Multimedia source identification using an improved weight photo response non-uniformity noise extraction model in short compressed videos | |
| Amin et al. | Exposing deepfake frames through spectral analysis of color channels in frequency domain | |
| CN106650678A (en) | Face identification method through utilization of Gabor wavelet sub-band related structure |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170118 Termination date: 20170411 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |