CN107204007A - It is a kind of that blind evidence collecting method is pasted based on the duplication for blocking DCT domain coefficient - Google Patents
It is a kind of that blind evidence collecting method is pasted based on the duplication for blocking DCT domain coefficient Download PDFInfo
- Publication number
- CN107204007A CN107204007A CN201710365547.5A CN201710365547A CN107204007A CN 107204007 A CN107204007 A CN 107204007A CN 201710365547 A CN201710365547 A CN 201710365547A CN 107204007 A CN107204007 A CN 107204007A
- Authority
- CN
- China
- Prior art keywords
- vector
- msub
- image
- transfer vector
- matrix
- 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.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/14—Fourier, Walsh or analogous domain transformations, e.g. Laplace, Hilbert, Karhunen-Loeve, transforms
- G06F17/147—Discrete orthonormal transforms, e.g. discrete cosine transform, discrete sine transform, and variations therefrom, e.g. modified discrete cosine transform, integer transforms approximating the discrete cosine transform
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Discrete Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a kind of based on the blind evidence collecting method of duplication stickup for blocking DCT domain coefficient, including:Dct transform operation is carried out respectively to each subimage block, handled using quantization parameter, then carries out zigzag scanning and carries out break-in operation, characteristic vector is obtained, characteristic vector is saved in matrix A according to slip order;Dictionary sequence is carried out to matrix A and obtains matrix B, and to each row vector of matrix B, all carries out matching detection with adjacent row vector, if approximately the same, transfer vector is calculated;Traversal is all to be determined as approximately uniform row vector pair, obtains all transfer vectors, therefrom finds out main transfer vector, removes the sub-blocks different from main transfer vector, remaining sub-block be replicate stickup distort part;Identify in image to be detected be replicated, sticking area, remove isolated block, output image by opening operation.Present invention reduces time complexity, with more strong robustness, applied to any blind evidence obtaining that mode of distorting is pasted based on duplication.
Description
Technical field
Blind evidence obtaining field is pasted the present invention relates to the duplication of digital picture, more particularly to one kind is based on blocking DCT
Blind evidence collecting method is pasted in the duplication of (Discrete Cosine Transform, discrete cosine transform) domain coefficient.
Background technology
With the continuing to develop of Internet technology and multimedia technology, a large amount of popularizations of digital camera and picture editting
Software such as Photoshop etc. extensive use, people, which modify to image, edit and stored, becomes more and more easier.If
The digital picture of forgery changes original information, and is widely used in the neck such as news media, scientific research, court's exhibit
Domain, will influence the stabilization of our economy, our society and our politics[1].The Image Blind forensic technologies developed based on this phenomenon have wide answer
With prospect, turn into one of hot issue of image processing field research in recent years.
Image distorts that mode is numerous, and what one of which was most often used distorts mode is to be pasted with the duplication in width image
Distort, i.e., certain part in duplicating image is pasted it into another region in the width image, and both are without any
Occur simultaneously.Generally, because color, brightness with piece image are all without varying widely, thus distort part will very
Difficulty causes discovering for people.
At present, the blind forensic technologies of stickup are replicated to be divided into three classes:Matching based on image slices vegetarian refreshments, Autocorrelation Detection method and
Matching based on image block[2]。
1st, the matching based on image slices vegetarian refreshments:Tampered region is detected by being traveled through relatively to image pixel point.This
Method And Principle is simple, accuracy is high, logic chain is simple, for being detected without the duplication reproducing image Jing Guo any post-processing operation
Effect is obvious, but amount of calculation is huge, and robustness is extremely bad.
2nd, Autocorrelation Detection method[3]:Its principle is that duplication has identical or similitude with the image block of stickup, thus
With very strong autocorrelation.The algorithm comparison is simple, and the complexity of calculating is nor very big, but when this algorithm is applied to be answered
When the area size that system stickup is distorted is less than 25% situation of entire image, often there is situations such as accuracy rate is not high.
3rd, the matching based on image block:By find in image block same or similar piece to algorithm tampered region is entered
Row detection positioning.This algorithm principle is simple, it is easy to accomplish and it is widely applicable.
Three class algorithms respectively have quality, but because the matching based on image block can reduce time complexity to a certain extent,
And be used widely with preferably Detection results and certain robustness[4]。
By being found to above-mentioned analysis, the significant challenge that the blind evidence obtaining field of duplication stickup faces at present is:Algorithm when
Between complexity it is too high and bad to the robustness of various conventional post-processing operations, limit practical ranges.
The content of the invention
The invention provides a kind of based on the blind evidence collecting method of duplication stickup for blocking DCT domain coefficient, the present invention effectively drops
Low time complexity and with stronger robustness, can apply to any distort the blind of mode and take based on replicating to paste
Card, it is described below:
It is a kind of that blind evidence collecting method is pasted based on the duplication for blocking DCT domain coefficient, it the described method comprises the following steps:
Dct transform operation is carried out respectively to each subimage block, each subimage block handled using quantization parameter, then enters
Row zigzag scans and carries out break-in operation, obtains characteristic vector, characteristic vector is saved in matrix A according to slip order;
Dictionary sequence is carried out to matrix A and obtains matrix B, and to each row vector of matrix B, all with adjacent row vector
Matching detection is carried out, if approximately the same, transfer vector is calculated;
Traversal is all to be determined as approximately uniform row vector pair, obtains all transfer vectors, therefrom find out it is main shift to
Amount, removes sub-blocks different from main transfer vector, remaining sub-block be replicate stickup distort part;
Identify in image to be detected be replicated, sticking area, remove isolated block, output image by opening operation.
Wherein, judge that approximately the same step is specially:
Condition 1:Characteristic component
JudgeWhether set up, if it is, continuing more next characteristic component;If not, judging two
Individual characteristic vectorDiffer;Abs is absolute value sign,It is characterized component;
Condition 2:Characteristic component
1) calculateChange maxr and minr value simultaneously:If maxr < rl, then maxr=rl;If minr
> rl, then minr=rl;
2) judge whether maxr-minr < t set up, if set up, continue more next characteristic component, it is on the contrary then straight
Connect and judge the two characteristic vectorsDiffer;
Obtain and meet above-mentioned condition 1, two characteristic vectors of condition 2 simultaneously, two characteristic vectors are approximately uniform.
Wherein, the step of calculating transfer vector is specially:
Wherein, d is transfer vector, (x1,y1), (x2,y2) be two image subblocks upper left position.
Wherein, the main transfer vector is specially:
Traversal is all to be determined as approximately uniform row vector pair, calculates each transfer vector, all transfer vectors are entered
Row probability statistics, choose the most transfer vectors of frequency appearance in each transfer vector and are used as main transfer vector.
Wherein, methods described also includes:
When in the absence of the sub-block different from main transfer vector, show that image to be detected does not replicate stickup and distorted.
The beneficial effect for the technical scheme that the present invention is provided is:
1st, the DCT coefficient for quantifying to block is chosen as the characteristic vector of each sub-block, significantly reduces the complexity of time
Degree;
2nd, the post-processing operation commonly used after being distorted for addition white Gaussian noise and Gaussian Blur etc., with certain Shandong
Rod;
3rd, by find all similar matching blocks maximum main transfer vector and it is morphologic open operation handle, carry
The high accuracy of detection.
Brief description of the drawings
Fig. 1 is a kind of flow chart that blind evidence obtaining algorithm is pasted based on the duplication for blocking DCT domain coefficient;
Fig. 2 is zigzag scanning sequency schematic diagram;
Fig. 3 is the transfer vector schematic diagram between the duplication sticking area correspondence similar block of image;
Fig. 4 is the various different schematic diagrames for detecting samples replicated in the case of stickup is distorted;
Wherein, (a) is to be replicated part and the part of stickup to be horizontally situated;(b) it is to be replicated part and stickup
Part is located at upright position;(c) it is located at positive diagonal position to be replicated part and the part of stickup;(d) for be replicated part with
The part of stickup is located at anti-diagonal bits;(e) it is irregular area to be replicated part and the part of stickup.In addition, image is divided into
Three columns, from left to right respectively original image, image after being tampered and the design sketch detected using this method.
Fig. 5 is the schematic diagram of the robust detection sample of anti-noise jamming;
Wherein, (a) is the additive white Gaussian noise processing image and test result that signal to noise ratio is 35dB;(b) it is signal to noise ratio
Image and test result are handled for 25dB additive white Gaussian noise;(c) it is additive Gaussian white noise that signal to noise ratio is 15.035dB
Sonication image and test result.In addition, image is divided into three columns, it is respectively from left to right only to paste the figure after distorting by replicating
Picture, the effect added the image after additive white Gaussian noise again after replicating stickup and distorting and detected using this method
Figure.
Fig. 6 is the schematic diagram for the robust detection sample that anti is handled;
Wherein, (a) is set to for parameter:n1=n2=5, σ2Image to be detected and its inspection after=1 Gaussian Blur processing
Survey result;(b) it is set to for parameter:n1=n2=5, σ2Image to be detected and its detection knot after=2 Gaussian Blur processing
Really.In addition, image is divided into three columns, it is respectively from left to right only to paste the image after distorting by replicating, is usurped by replicating stickup
The image after Gaussian Blur and the design sketch detected using this method are added after changing again.
Fig. 7 is the schematic diagram of the detection performance comparision of three kinds of algorithms;
The schematic diagram that Fig. 8 compares for the time complexity of three kinds of algorithms.
Embodiment
To make the object, technical solutions and advantages of the present invention clearer, further is made to embodiment of the present invention below
It is described in detail on ground.
In order to solve problem above, it is desirable to be able to comprehensively, automatically, accurately extract the feature of sub-block and positioned and detected
Algorithm.Research shows:The energy compaction property of dct transform can to retain big portion in the case where only retaining low frequency coefficient
Divide important information, this embodies in the coefficient matrix after quantization to become apparent, the DCT coefficient for blocking quantization can be selected to make
For the feature of each sub-block[5]。
Embodiment 1
The embodiment of the present invention is proposed pastes blind evidence collecting method based on the duplication for blocking DCT domain coefficient, referring to Fig. 1, the party
Method comprises the following steps:
101:Dct transform operation is carried out respectively to each subimage block, each subimage block handled using quantization parameter,
Zigzag scanning is carried out again and break-in operation is carried out, characteristic vector is obtained, and characteristic vector is saved in matrix A according to slip order
In;
102:Dictionary sequence is carried out to matrix A and obtains matrix B, and to each row vector of matrix B, all with adjacent row
Vector carries out matching detection, if approximately the same, calculates transfer vector;
103:Traversal is all to be determined as approximately uniform row vector pair, obtains all transfer vectors, therefrom finds out main turn
The amount of shifting to, removes sub-blocks different from main transfer vector, remaining sub-block be replicate stickup distort part;
104:Identify in image to be detected be replicated, sticking area, remove isolated block, output image by opening operation.
Wherein, the approximately the same step that judges in step 102 is specially:
Condition 1:Characteristic component
JudgeWhether set up, if it is, continuing more next characteristic component;If not, judging two
Individual characteristic vectorDiffer;Abs is absolute value sign,It is characterized component;
Condition 2:Characteristic component
1) calculateChange maxr and minr value simultaneously:If maxr < rl, then maxr=rl;If minr
> rl, then minr=rl;
2) judge whether maxr-minr < t set up, if set up, continue more next characteristic component, it is on the contrary then straight
Connect and judge the two characteristic vectorsDiffer;
Obtain and meet above-mentioned condition 1, two characteristic vectors of condition 2 simultaneously, two characteristic vectors are approximately uniform.
Wherein, the main transfer vector in step 103 is specially:
Traversal is all to be determined as approximately uniform row vector pair, calculates each transfer vector, all transfer vectors are entered
Row probability statistics, choose the most transfer vectors of frequency appearance in each transfer vector and are used as main transfer vector.
In summary, the embodiment of the present invention by above-mentioned steps 101- steps 104 significantly reduce time complexity with
And with stronger robustness, can apply to any blind evidence obtaining that mode of distorting is pasted based on duplication, meet practical application
In a variety of needs.
Embodiment 2
The scheme in embodiment 1 is further introduced with reference to specific example, mathematical formulae, it is as detailed below
Description:
201:Image to be detected I is converted into gray level image, and is divided into several subimage blocks;
For example:If image to be detected I size is M × N, the size for choosing sliding shoe is B × B, makes it from mapping to be checked
As the I upper left corner starts according to from left to right, order from top to bottom enters line slip, finally successively in units of a pixel
Image to be detected I lower right corner is reached, Sum=(M-B+1) (N-B+1) individual subimage block can be finally obtained.
Whether when implementing, it is gray level image to first determine whether image to be detected I, if not being then converted into gray scale
Image.All experimental images that the embodiment of the present invention is used are from the digital image evidence collecting of T.-T.Ng et al. issues
Image measurement collection[6], all images are all the gray level image that size is 128 × 128, and preservation form is BMP.Chosen in parameter
When, choose M=N=128, B=8.
When implementing, M, N and B value can also not limited from other test sets, the embodiment of the present invention
System, is set according to the need in practical application.
202:Dct transform operation is carried out respectively to each subimage block of acquisition, and using quantization parameter Q to each subimage block
Quantification treatment is carried out, zigzag scanning (as shown in Figure 2) is then carried out again and break-in operation is carried out, the feature of all sub-blocks is obtained
Vector [pB2], then the characteristic vector of all sub-blocks is saved in matrix A according to slip order;
For example:For the digital picture f (x, y) of D × D sizes, its dct transform and its IDCT (Inverse Discrete
Cosine Transform, inverse discrete cosine transform) conversion be respectively:
Wherein, u, v=0,1,2 ..., D-1, x, y=0,1,2 ..., D-1.A (u) and a (v) is intermediate variable.
In above-mentioned formula, as u=v=0, cosine value takes 1, hasThis component is
The maximum of DCT coefficient after change, also referred to as DC component, other coefficients are referred to as AC compounent.
DCT coefficient matrix after conversion, corresponding frequency is by direct current and low frequency component from the upper left corner to the lower right corner, to high frequency
Component variation, and in a practical situation, the higher value of DCT coefficient matrix is all concentrated in direct current and low frequency component.Particularly
It is that high frequency coefficient substantially becomes 0 entirely after quantization.
It therefore, it can carry out dimension-reduction treatment to characteristic vector according to the characteristic, i.e., the DCT coefficient of each sub-block only considers
Low frequency part.However, can not be preserved with a matrix type again in DCT characteristic vector, it is necessary to by the coefficient after conversion according to
Order from low to high is arranged, and so can just facilitate dimensionality reduction and matching treatment afterwards.Therefore, it is necessary to by quantization
DCT coefficient matrix is scanned into a size for 1 × B according to zigzag arrangement mode2Row vector, and by p (0 < p≤
1) (p=0.25 in this example) blocks reservation k (k=[p × B as guillotine factor2]) individual component (k=16 in this example).
Characteristic vector can also be quantized coefficient Q quantizations before blocking, i.e., each component divided by Q in characteristic vector are then
Round.Then the characteristic vector after quantization and dimension-reduction treatment is:
203:Dictionary sequence is carried out to eigenvectors matrix A and obtains matrix B, its size is (M-B+1) (N-B+1) × [p
B2], and to each row vector of matrix BN all adjacent theretofIndividual row vectorCarry out matching detection, wherein NfIt should meet:
J-i < Nf≤ (M-B+1) (N-B+1), (N in this examplef=3) if matching result is approximately the same, calculates turning between the two
The amount of shifting to;
Before similar sub-block is found, dictionary sequence is carried out to eigenvectors matrix A first and obtains matrix B, immediately two-by-two
Compare between them the similarity degree of (i.e. between matrix B after eigenvectors matrix A sequences), if it is determined that it is approximately the same,
Then corresponding two sub-block is determined as similar sub-block pair, then records the transfer vector d between similar sub-block pair:
Wherein (x1,y1), (x2,y2) be two image subblocks upper left position coordinate.Replicate sticking area correspondence
There is unified transfer vector, as shown in Figure 3 between block.
The detailed process that two row vectors carry out matching detection is described in detail below:
When image to be detected is through overcompression plus makes an uproar, obscures and wait after attack, then be replicated characteristic vector and the stickup of sub-block
The characteristic vector of sub-block may not be essentially equal, but much like.Therefore needs are incorporated herein one kind and judge sub-block pair
Answer characteristic vector whether identical method.If each corresponding component of the characteristic vector of two sub-blocks is almost equal,
Then it is believed that the two sub-blocks are approximate related.
Illustrate to judge the whether approximately uniform tool of the corresponding characteristic vector of two sub-blocks with a specific example below
Body process:
For example:WithThe characteristic vector of respectively two sub-blocks, judge this two
Whether approximately uniform the corresponding characteristic vector process of individual sub-block be as follows:
Wherein, the two parameters of s and t are predefined in programming, and maxr is initialised a sufficiently small number, minr quilts
A sufficiently large number is initialized to,Respectively characteristic component.So specifically judge that flow is:
Condition 1:For each 1≤l≤k,
JudgeWhether set up, if it is, continuing more next characteristic component;If not, judging two
Individual characteristic vectorDiffer;
Wherein, abs is absolute value sign;S=4 in the embodiment of the present invention, can also be according to practical application when implementing
In the need for s value is set, the embodiment of the present invention is not repeated this.
Condition 2:For each 1≤l≤k,
1) calculateValue, while correspondence change maxr and minr value:If maxr < rl,
Then maxr=rl;If minr > rl, then minr=rl;
2) judge whether maxr-minr < t (t=0.0625 in this example) set up, if set up, continue more next
Characteristic component, it is on the contrary then directly judge the two characteristic vectorsDiffer.
, can also be according to being set to t value the need in practical application when implementing, the embodiment of the present invention pair
This is not repeated.
Obtain and meet above-mentioned condition 1, two characteristic vectors of condition 2 simultaneously, two characteristic vectors are approximately uniform.
204:Traversal is all to be determined as approximately uniform row vector pair, and counts all transfer vectors, and scanning is all
Transfer vector, find out main transfer vector therein, then remove all corresponding sub-blocks different from main transfer vector, remaining son
Block can be determined as replicate paste distort part, if do not found, you can judge do not have to replicate in the image to be detected
Stickup is distorted, and flow terminates, otherwise is transferred to next step;
Sub-block pair for removing erroneous matching, can use and determine the method for main transfer vector to realize that precisely duplication is viscous
Paste regional determination.Its principle is replication region D1With sticking area D2All have unified transfer vector, as shown in figure 3, and with pole
Transfer vector produced by a small number of similar areas is compared, and replicates sticking area D1, D2Transfer vector account for the overwhelming majority, so
Those indivedual a small number of transfer vectors are removed, left corresponding sub-block extremely replicates adhesive portion to region.
The detailed process for finding out main transfer vector is described in detail below:
Traversal is all to be determined as approximately uniform row vector pair, each transfer vector d is calculated, to all transfer vector d
Probability statistics are carried out, the most transfer vectors of frequency appearance in each transfer vector d is chosen and is used as main transfer vector.
205:Identify respectively in image to be detected and be replicated region and sticking area, removed using operation is opened on map
Possible isolated block, then exports the image viewing handled, and flow terminates.
Block and quantify because characteristic vector have passed through, might have the isolated area that some flase drops come out, therefore, use
It is morphologic to open operation[7], to remove the block that some are isolated.
The step of block is isolated in specific removal is known to those skilled in the art, and the embodiment of the present invention is not repeated this.
In summary, the embodiment of the present invention by above-mentioned steps 201- steps 205 significantly reduce time complexity with
And with stronger robustness, can apply to any blind evidence obtaining that mode of distorting is pasted based on duplication, meet practical application
In a variety of needs.
Embodiment 3
Feasibility checking is carried out to the scheme in Examples 1 and 2 with reference to specific calculation formula, Fig. 4-Fig. 8, referred to
It is described below:
The natural image for all experiments that this experiment is used both be from T.-T.Ng et al. issue be used for digital image evidence collecting
Image measurement collection[6], wherein it is 128 × 128 pixels that all images, which are all size, it is gray level image, preserving form is
BMP。
The accuracy of tampered region position can be detected in order to test detection algorithm, the following parameter is calculated.Note is defeated
The replication region for entering tampered image is D1, sticking area is D2, the replication region that detection algorithm is detected is R1, sticking area
For R2, effective detection acc, invalid detection f is respectively defined as:
Wherein, | | for the area in region, ∩ is the common factor in two regions, ∪ be two regions and.Acc and f bodies
The degree of closeness of the tampered region of physical presence in the region and image to be detected that detection algorithm detects is showed.
This algorithm is contrasted with following two algorithms in experiment:
Fridrich algorithms[2], the processing mode of its characteristics of image block is DCT;
Popescu algorithms[8], the processing mode of its characteristics of image block is PCA (Principal Component
Analysis)。
In order to preferably verify that this algorithm pastes the accuracy of tampering detection for replicating, to 300 nature figures of selection
Distort as having carried out duplication stickup, and considered it is various it is possible distort situation, to obtained all effective detections
Acc, invalid detection f makees average treatment, obtains average effective detectionAverage invalid detection
In addition, the robustness in order to verify this algorithm, has carried out the operations, such as Fig. 5 such as noise addition and Fuzzy Processing respectively
With shown in Fig. 6, and the detection performance of each algorithm has been subjected to list compared, as shown in Figure 7.When to image to be detected addition Gauss
Signal to noise ratio snr=35 after noise or during 25dB, Detection results are consistent with Detection results detection before addition noise;But work as SNR
During less than 20.035, Detection results drastically decline.
By being compared discovery with Popescu algorithms, Popescu algorithms are when signal to noise ratio is less than 25dB, Detection results
Will drastically it be deteriorated, therefore this algorithm has more preferable robustness in anti-noise sonication;When to image to be detected progress Gauss
During Fuzzy Processing, wherein parameter is set to:n1=n2=5, σ2=1 or σ2=2, Detection results are also fine, can verify this calculation
Method also has certain robustness in anti processing.In addition, the time complexity of three kinds of algorithms is compared, such as scheme
8 understand that this algorithm reduces time complexity to a certain extent.To sum up, the experiment show feasibility of this method with
Superiority.
Bibliography:
[1] Zhou Linna digital image blinds forensic technologies research [D] Beijing University of Post & Telecommunications, 2007.
[2]J.Fridrich,D.Soukal,J.Lukas.Detection of copy-move forgery in
digital images[A].In:Proceedings of the Digital Forensic Research Workshop
[C],USA,2003:1-10.
[3] list common vetch is based on digital image tampering detection research [D] the University Of Suzhou for replicating stickup, 2014.
[4] research [D] the South China Science & Engineering University of the red passive forensic technologies of picture materials of Zhao Jun, 2011.
[5]S.A.Khayam.The Discrete Cosine Transform(DCT):Theory and
Application[M].Upper Saddle River,NJ,USA:Prentice Hall Press,Aug.2003,32-41.
[6]T.-T.Ng,J.Hsu,S.-F.Chang.Columbia Image Splicing Detection
Evaluation Dataset,http://www.ee.columbia.edu/ln/dvmm/downloads/
AuthSplicedDataSet/AuthSplicedDataSet.htm.
[7]Rafeal C.Gonzalez,Richard E.Woods P.,“Digital Image Processing”,
Edition 2nd,Addison-Wesley pp.528-532,1992.
[8]A.Popescu,H.Farid.Exposing Digital Forgeries by Detecting
Duplicated Image Regions[R],Technical Report TR 2004-515,Dartmouth College,
2004.
It will be appreciated by those skilled in the art that accompanying drawing is the schematic diagram of a preferred embodiment, the embodiments of the present invention
Sequence number is for illustration only, and the quality of embodiment is not represented.
The foregoing is only presently preferred embodiments of the present invention, be not intended to limit the invention, it is all the present invention spirit and
Within principle, any modification, equivalent substitution and improvements made etc. should be included in the scope of the protection.
Claims (5)
1. a kind of paste blind evidence collecting method based on the duplication for blocking DCT domain coefficient, it is characterised in that methods described includes following step
Suddenly:
Dct transform operation is carried out respectively to each subimage block, handled using quantization parameter, then carries out zigzag scanning going forward side by side
Row break-in operation, obtains characteristic vector, and characteristic vector is saved in matrix A according to slip order;
Dictionary sequence is carried out to matrix A and obtains matrix B, and to each row vector of matrix B, is all carried out with adjacent row vector
Matching detection, if approximately the same, calculates transfer vector;
Traversal is all to be determined as approximately uniform row vector pair, obtains all transfer vectors, therefrom finds out main transfer vector, go
Except the sub-block different from main transfer vector, remaining sub-block be replicate paste distort part;
Identify in image to be detected be replicated, sticking area, remove isolated block, output image by opening operation.
2. according to claim 1 a kind of based on the blind evidence collecting method of duplication stickup for blocking DCT domain coefficient, its feature exists
In judging that approximately the same step is specially:
Condition 1:Characteristic component
JudgeWhether set up, if it is, continuing more next characteristic component;If not, judging two spies
Levy vectorDiffer;Abs is absolute value sign,It is characterized component;
Condition 2:Characteristic component
1) calculateChange max r and min r value simultaneously:If max r < rl, then max r=rl;If min
R > rl, then min r=rl;
2) judge whether max r-min r < t set up, if set up, continue more next characteristic component, it is on the contrary then direct
Judge the two characteristic vectorsDiffer;
Obtain and meet above-mentioned condition 1, two characteristic vectors of condition 2 simultaneously, two characteristic vectors are approximately uniform.
3. according to claim 1 a kind of based on the blind evidence collecting method of duplication stickup for blocking DCT domain coefficient, its feature exists
In, it is described calculating transfer vector the step of be specially:
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>d</mi>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>d</mi>
<mi>x</mi>
</msub>
<mo>,</mo>
<msub>
<mi>d</mi>
<mi>y</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<mi>d</mi>
<mi>x</mi>
</msub>
<mo>=</mo>
<msub>
<mi>x</mi>
<mn>1</mn>
</msub>
<mo>-</mo>
<msub>
<mi>x</mi>
<mn>2</mn>
</msub>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<msub>
<mi>d</mi>
<mi>y</mi>
</msub>
<mo>=</mo>
<msub>
<mi>y</mi>
<mn>1</mn>
</msub>
<mo>-</mo>
<msub>
<mi>y</mi>
<mn>2</mn>
</msub>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
Wherein, d is transfer vector, (x1,y1), (x2,y2) be two image subblocks upper left position.
4. according to claim 1 a kind of based on the blind evidence collecting method of duplication stickup for blocking DCT domain coefficient, its feature exists
In the main transfer vector is specially:
Traversal is all to be determined as approximately uniform row vector pair, calculates each transfer vector, all transfer vectors is carried out general
Rate is counted, and is chosen the most transfer vectors of frequency appearance in each transfer vector and is used as main transfer vector.
5. according to claim 1 a kind of based on the blind evidence collecting method of duplication stickup for blocking DCT domain coefficient, its feature exists
In methods described also includes:
When in the absence of the sub-block different from main transfer vector, show that image to be detected does not replicate stickup and distorted.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710365547.5A CN107204007A (en) | 2017-05-22 | 2017-05-22 | It is a kind of that blind evidence collecting method is pasted based on the duplication for blocking DCT domain coefficient |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710365547.5A CN107204007A (en) | 2017-05-22 | 2017-05-22 | It is a kind of that blind evidence collecting method is pasted based on the duplication for blocking DCT domain coefficient |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107204007A true CN107204007A (en) | 2017-09-26 |
Family
ID=59906388
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710365547.5A Pending CN107204007A (en) | 2017-05-22 | 2017-05-22 | It is a kind of that blind evidence collecting method is pasted based on the duplication for blocking DCT domain coefficient |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107204007A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108230411A (en) * | 2017-12-29 | 2018-06-29 | 成都工业学院 | The detection method and device of a kind of tampered image |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102957915A (en) * | 2012-11-15 | 2013-03-06 | 西安理工大学 | Double JPEG (Joint Photographic Experts Group) compressed image-targeted tempertamper detection and tempertamper locating method |
CN103345758A (en) * | 2013-07-25 | 2013-10-09 | 南京邮电大学 | Joint photographic experts group (JPEG) image region copying and tampering blind detection method based on discrete cosine transformation (DCT) statistical features |
-
2017
- 2017-05-22 CN CN201710365547.5A patent/CN107204007A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102957915A (en) * | 2012-11-15 | 2013-03-06 | 西安理工大学 | Double JPEG (Joint Photographic Experts Group) compressed image-targeted tempertamper detection and tempertamper locating method |
CN103345758A (en) * | 2013-07-25 | 2013-10-09 | 南京邮电大学 | Joint photographic experts group (JPEG) image region copying and tampering blind detection method based on discrete cosine transformation (DCT) statistical features |
Non-Patent Citations (2)
Title |
---|
YANPING HUANG ET AL: "Improved DCT-based detection of copy-move forgery in images", 《FORENSIC SCIENCE INTERNATIONAL》 * |
王俊文: "数字图像内容篡改盲取证研究", 《万方学位论文库》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108230411A (en) * | 2017-12-29 | 2018-06-29 | 成都工业学院 | The detection method and device of a kind of tampered image |
CN108230411B (en) * | 2017-12-29 | 2021-03-09 | 成都工业学院 | Method and device for detecting tampered image |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Qin et al. | Perceptual image hashing via dual-cross pattern encoding and salient structure detection | |
Kang et al. | Robust median filtering forensics using an autoregressive model | |
Wang et al. | Detection of image region duplication forgery using model with circle block | |
Popescu et al. | Exposing digital forgeries by detecting duplicated image regions | |
Huang et al. | Improved DCT-based detection of copy-move forgery in images | |
Kumar et al. | A fast DCT based method for copy move forgery detection | |
Ouyang et al. | Robust hashing for image authentication using quaternion discrete Fourier transform and log-polar transform | |
Ustubioglu et al. | A new copy move forgery detection technique with automatic threshold determination | |
Singh et al. | Fast and efficient region duplication detection in digital images using sub-blocking method | |
Muhammad et al. | Copy-move forgery detection using dyadic wavelet transform | |
Ketenci et al. | Copy-move forgery detection in images via 2D-Fourier transform | |
Kang et al. | Robust median filtering forensics based on the autoregressive model of median filtered residual | |
AlSawadi et al. | Copy-move image forgery detection using local binary pattern and neighborhood clustering | |
Yu et al. | Perceptual hashing with complementary color wavelet transform and compressed sensing for reduced-reference image quality assessment | |
Maind et al. | Image copy move forgery detection using block representing method | |
Al-Hammadi et al. | Curvelet transform and local texture based image forgery detection | |
Muhammad et al. | A non-intrusive method for copy-move forgery detection | |
Gong et al. | Image copy-move forgery detection using SURF in opponent color space | |
Yang et al. | Digital image forgery forensics by using undecimated dyadic wavelet transform and Zernike moments | |
Hamadouche et al. | A comparative study of perceptual hashing algorithms: Application on fingerprint images | |
Hamdy et al. | Quantization table estimation in JPEG images | |
Guo et al. | Content based image hashing via wavelet and radon transform | |
Soni et al. | Image forensic using block-based copy-move forgery detection | |
CN107204007A (en) | It is a kind of that blind evidence collecting method is pasted based on the duplication for blocking DCT domain coefficient | |
Ustubioglu et al. | A novel keypoint based forgery detection method based on local phase quantization and SIFT |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170926 |