CN101706944A - Quantization table evaluation based method for detecting JPEG image tampering - Google Patents
Quantization table evaluation based method for detecting JPEG image tampering Download PDFInfo
- Publication number
- CN101706944A CN101706944A CN200910198229A CN200910198229A CN101706944A CN 101706944 A CN101706944 A CN 101706944A CN 200910198229 A CN200910198229 A CN 200910198229A CN 200910198229 A CN200910198229 A CN 200910198229A CN 101706944 A CN101706944 A CN 101706944A
- Authority
- CN
- China
- Prior art keywords
- image
- category
- quantization table
- quantization
- distorted
- 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
Images
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a quantization table evaluation based method for detecting JPEG image tampering, which is characterized by first selecting a plurality of sub-blocks of the images to be detected, respectively evaluating the quantization table of each image sub-block and comparing every two quantization tables, detecting whether the images to be detected are subjected to tampering operation of clipping, then dividing the images to be detected into blocks again and extracting the eigenvector of the quantization table of each image sub-block or the quantization table of each image sub-block according to whether the images to be detected are subjected to clipping, finally classifying the image blocks according to the difference between the eigenvectors of the quantization tables of adjacent image blocks or the quantization tables of adjacent image blocks, and judging whether the images are subjected to synthetic clipping operation and detecting the tampering areas. The method is established based on the quantization table evaluation algorithm with low complexity and high accuracy, can improve the efficiency of detecting tampering of images and can simultaneously detect tampering of the images subjected to clipping and synthesis to further improve the accurate rate of image detection.
Description
Technical field
The invention belongs to the multi-media information security technical field, be specifically related to the method that a kind of detection jpeg image of estimating based on quantization table is distorted, this method is applied to the passive authentication of jpeg image.
Background technology
In current digital Age, along with popularizing and the widespread use of simple image processing software of the image digitazation pickup apparatus of various low price, make people can make the digital picture that a width of cloth is forged at an easy rate, and naked eyes often are difficult to distinguish the true and false of these forgery images.What be worth people's worry is, utilize the convenience that digital picture is propagated on the internet, the image of these forgeries is as being used to some responsive field, and for example politics, military affairs and the administration of justice etc. will produce immeasurable influence to our social stability and even the safety of whole country.Therefore, the authenticity verification of picture material becomes the Information Security problem that modern society needs to be resolved hurrily.Digital watermarking and digital signature are a kind of comparatively effectively authentication means, but owing to need embed authentication information in the source images of not distorting, make it can not effectively be used in some cases.Therefore need prior imformation few, do not need the artificial passive authentication method that in image source, embeds authentication information to be fit to the application of more occasions.
Jpeg image is one of current digital image format the most commonly used, thereby has more practical significance at the detection of distorting of jpeg image.Need train the proper vector of extracting by support vector machine at the passive authentication method overwhelming majority of jpeg image at present, and then testing image classified, though such authentication method can detect distorting of jpeg image, but, these class methods often need a large amount of training samples to guarantee to detect the accuracy that jpeg image is distorted, computation complexity height; In addition, destroyed the piece grid through the jpeg image after cutting out, what cause that most method all can not detected image distorts.
Summary of the invention
The problem and shortage that exists of prior art in view of the above, the method that the object of the present invention is to provide a kind of detection jpeg image of estimating based on quantization table to distort, this method is based upon on the quantization table algorithm for estimating basis that a kind of complexity is little and precision is high, can improve the efficient of distorting of detected image, and can detect simultaneously through cutting out and the distorting of the jpeg image that synthesize, the raising detected image accuracy.
In order to achieve the above object, the present invention adopts following technical proposals:
The method that a kind of detection jpeg image of estimating based on quantization table is distorted.The concrete steps that it is characterized in that this method are: at first choose the some image subblocks of testing image, estimate the quantization table of every image subblock respectively and make comparisons the operation whether process is cut out of detection testing image in twos.Then testing image is drawn piece again, whether cut out quantization table proper vector or the quantization table that extracts every image subblock according to the testing image process, at last, difference degree by adjacent image piece quantization table proper vector or quantization table is classified to image block, judge that whether image is through the synthetic operation of distorting, and detect the tampered region, its step is as follows:
(1), choosing the n block size is the not overlapped image block I of L * L
1, I
2... I
n
(2), estimate the quantization table of every image block to obtain Qt
1, Qt
2..., Qt
n
(3), with Qt
1, Qt
2..., Qt
nQt is calculated in computing in twos respectively
iAnd Qt
jBetween difference measurement Diff
I, j
(4), add up among the vectorial Diff greater than threshold value T
1The number Num of element, judgment threshold T
1The number Num of element, if Num>3n (n-1)/8 judges that then testing image is through cutting out execution in step (5); Otherwise judge that testing image not through cutting out, jumps to step (10);
(5), choose the image subblock I of jpeg image I to be measured
0, and it is divided into not overlapped image block B
1,1, B
2,1..., B
M, 1
(6), transitional slide B
I, 1(i=1 2...m), gets B
I, j(j=1,2 ..., 64);
(7), estimate B
I, 1(i=1, quantization table proper vector Q 2...m)
i(i=1,2 ..., m);
(8), with image block B
I, 1(i=1,2 ..., m) be divided into: category-A and category-B.Initialization: with B
1,1Demarcate and be category-A.Calculate adjacent image piece B
I-1,1And B
I, 1Diversity factor vector D between the quantization table proper vector
i=[d
I, 1, d
I, 2..., d
I, j..., d
I, 64], wherein
Get element minimum in the diversity factor vector then as image block B
I-1,1And B
I, 1The difference measurement Δ of quantization table
i
(9), judge distorted image: if D
i<T
2, then with B
I, 1Demarcate and be and B
I-1,1Identical class; Otherwise with B
I, 1Demarcate and be and B
I-1,1Different classes, after the classification, if all images piece is category-A, then testing image is not distorted through synthetic; If existing category-A also has category-B, think that then category-A and category-B belong to different image sources, promptly to distort, and judge that category-A or category-B are the tampered region through synthetic, whole testing process finishes;
(10), choose the image subblock I of jpeg image I to be measured
0, and it is divided into not overlapped image block B
1, B
2..., B
m
(11), estimated image piece B
i(i=1,2 ..., quantization table Q m)
i(i=1,2 ..., m);
(12), with image block B
i(i=1,2 ..., m) be divided into: category-A and category-B.Initialization: with B
1,1Demarcate and be category-A.Calculate adjacent image piece B
I-1,1And B
I, 1Difference measurement Δ between the quantization table
i
(13), judge distorted image: if Δ
i<T
2, then with B
I, 1Demarcate and be and B
I-1,1Identical class; Otherwise with B
I, 1Demarcate and be and B
I-1,1Different classes, after the classification, if all images piece is category-A, then testing image is not distorted through synthetic; If existing category-A also has category-B, then category-A belongs to different image sources with category-B, promptly distorts through synthetic, and judges that category-A or category-B are the tampered region, and whole testing process finishes.
Estimation quantization table described in above-mentioned steps (2), (7) and (11), its concrete steps are as follows:
(2-1), calculate discrete cosine transform coefficient;
(2-2), the statistical straggling cosine transform coefficient, histogram draws;
(2-3), find out the maximum point of i place, position discrete cosine transform coefficient statistic histogram;
(2-4), estimate quantization step q
2And q
3
(2-5), estimate quantization step q
1: use quantization step q
2And q
3Provide q
1Estimated value,
(2-6), estimate quantization step q
i(i=4,5..., 64) have obtained quantization step q successively
2, q
3, q
1, q
4..., q
63, q
64Value, promptly finished the estimation of whole quantization table.
Qmask weighting matrix described in the above-mentioned steps (3), diversity factor between two quantization tables of this weighting matrix is by they quantization step decisions in Qmask non-zero position, can effectively avoid, the influence that the evaluated error that quantization step is caused when less or big is brought, the detection accuracy of raising entire method.
Transitional slide described in the above-mentioned steps (6), it is with image block from right to left, be that step-length is carried out transitional slide with a pixel from bottom to up, after image block whenever is moved to the left 7 pixels, the pixel that moves up, and then move to left, 7 pixels have altogether moved up, accumulative total transitional slide 63 times, thus estimate to obtain the quantization table proper vector, solve by image and cause the institute that do not match of JPEG piece grid to cause the quantization table misjudgment through cutting out.
The advantage that the method that the detection jpeg image of estimating based on quantization table of the present invention is distorted compared with prior art specifically has is: this method directly from the statistic histogram of jpeg image discrete cosine transform coefficient, utilizes the maximum point of the positive and negative semiaxis of histogram to estimate quantization table on the link of estimating quantization table.And existing JPEG quantization table method of estimation has maximum likelihood estimate, the histogrammic power Spectral Estimation method of discrete cosine transform conversion coefficient etc., and algorithm complex is higher.Show estimated time of the present invention obviously than lacking the estimated time of other prior art by the contrast that provides among the embodiment, and then explanation the present invention detects jpeg image to distort the time few, the raising detection efficiency.Take the image subblock slide displacement in addition and estimate the method for quantization table proper vector, solved, cause the quantization table misjudgment and problem that can not tamper detection owing to image causes JPEG piece grid not match through cutting out.
Description of drawings
Fig. 1 is 64 position zigzag ordering synoptic diagram of 8 * 8 image blocks;
The process flow diagram of the method that Fig. 2 distorts for the detection jpeg image of estimating based on quantization table of the present invention;
Fig. 3 is step among Fig. 2 (2), (7) and (11)) process flow diagram of described estimation quantization table.
Embodiment
Below in conjunction with accompanying drawing embodiments of the invention are described in further detail.
The method that the above-mentioned detection jpeg image of estimating based on quantization table is distorted, as shown in Figure 2, its concrete steps are as follows:
(1), choosing the n block size is the not overlapped image block I of L * L
1, I
2... I
n
Choosing the n block size from jpeg image I to be measured is the not overlapped image block I of L * L
1, I
2... I
nI
Ix, I
IySatisfy respectively with L: I
IxMod8=1, I
IyMod8=1, Lmod8=0, wherein I
IxAnd I
IyDifference representative image piece I
iHorizontal ordinate and ordinate, mod represents complementation;
(2), estimate the quantization table of every image block to obtain Qt
1, Qt
2..., Qt
n, Qt wherein
i(i=1,2 ..., n) representative image piece I
iThe estimation quantization table, be one 8 * 8 matrix;
(3), with Qt
1, Qt
2..., Qt
nQt is calculated in computing in twos respectively
iAnd Qt
jBetween difference measurement Diff
I, j, do computing in twos:
Obtain an one-dimensional vector Diff=[Diff who contains n (n-1)/2 element
1,2, Diff
1,3... Diff
N-1, n].Wherein, Qt
i(k) represent quantization table Qt
i64 quantization steps in k quantization step;
It is a weighting matrix; Diff
I, jRepresent Qt
iAnd Qt
jBetween difference measurement;
(4), whether add up among the vectorial Diff greater than threshold value T
1The number Num of element, judge that whether original image is through cutting out judgment threshold T
1The number Num of element, if Num>3n (n-1)/8 judges that then testing image is through cutting out execution in step (5); Otherwise judge that testing image not through cutting out, jumps to step (10).Wherein, be meant the part that sanction is removed original image through the image of cutting out, remaining original image is called the image through cutting out;
(5), choose the image subblock I of jpeg image I to be measured
0, wherein, I
0The eighth row that is I is capable to M, and the 8th is listed as N row, i.e. I
0=I (8:N, 8:M).With I
0From left to right, be divided into the not overlapped image block B that the m block size is R * R from top to bottom
1,1, B
2,1..., B
M, 1Wherein,
Rmod8=0, M represents I
0Height, N represents I
0Wide,
Representative is smaller or equal to the maximum integer of x;
(6), transitional slide B
I, 1(i=1 2...m), gets B
I, j(j=1,2 ..., 64).
With B
I, 1From right to left, be that step-length is carried out transitional slide, B with a pixel from bottom to up
I, 1Whenever after being moved to the left 7 pixels, the pixel that moves up, and then move to left, 7 pixels altogether move up.So B
I, 1Be shifted altogether 63 times, obtained image block B
I, 2, B
I, 3..., B
I, 64
(7), estimate B
I, j(j=1, quantization table 2...64) constitute B
I, 1Quantization table proper vector Q
i=[Q
I, 1..., Q
I, i..., Q
I, 64], Q wherein
I, jRepresent B
I, jThe estimation quantization table;
(8), with image block B
I, 1(i=1,2 ..., m) be divided into: category-A and category-B.Initialization: with B
1,1Demarcate and be category-A.Calculate adjacent image piece B
I-1,1And B
I, 1Diversity factor vector D between the quantization table proper vector
i, D
i=[d
I, 1, d
I, 2..., d
I, j..., d
I, 64], wherein
Get element minimum in the diversity factor vector then as image block B
I-1,1And B
I, 1The difference measurement Δ of quantization table
i, Δ
i=min (d
I, 1, d
I, 2..., d
I, j..., d
I, 64), wherein min (x) is the element of amount of orientation x intermediate value minimum;
(9), judge distorted image: if D
i<T
2, then with B
I, 1Demarcate and be and B
I-1,1Identical class; Otherwise with B
I, 1Demarcate and be and B
I-1,1Different classes.After the classification,, think that then testing image does not have not distort through synthetic if all images piece is category-A; If existing category-A also has category-B, think that then category-A and category-B belong to different image sources, think that promptly this image is to distort through synthetic, and judge that category-A or category-B are the tampered region, and whole testing process finishes.Wherein, syntheticly distort a part separately that is meant two width of cloth different images and be spliced into another width of cloth image;
(10), choose the image subblock I of jpeg image I to be measured
0, with jpeg image I to be measured from left to right, being divided into the m block size from top to bottom is the not overlapped image block B of R * R
1, B
2..., B
mWherein,
M represents I
0Height, N represents I
0Wide;
(11), estimated image piece B
i(i=1,2 ..., quantization table Q m)
i(i=1,2 ..., m);
(12), with image block B
i(i=1,2 ..., m) be divided into two classes: category-A and category-B.Initialization: with B
1,1Demarcate and be category-A.Calculate adjacent image piece B
I-1,1And B
I, 1Difference measurement Δ between the quantization table
i,
(13), judge distorted image: if Δ
i<T
2, then with B
I, 1Demarcate and be and B
I-1,1Identical class; Otherwise with B
I, 1Demarcate and be and B
I-1,1Different classes, if classification back all images piece is category-A, then testing image is not distorted through synthetic; If existing category-A also has category-B, then category-A belongs to different image sources with category-B, promptly distorts operation through synthetic, and judges that category-A or category-B are the tampered region, and whole testing process finishes.
Estimated image piece I described in above-mentioned steps (2), (7) and (11)
1Quantization table, obtain Qt
1, Qt
2..., Qt
nAs shown in Figure 3, at first be nonoverlapping 8 * 8 image blocks and do discrete cosine transform, then 64 positions are pressed the zigzag ordering, according to draw corresponding discrete cosine transform coefficients statistics histogram and obtain maximum points all in the histogram of different positions image division.Maximum point from histogrammic positive and negative semiaxis estimates 64 locational quantization parameters at last, thereby estimates quantization table, and its concrete steps are as follows:
(2-1), calculate discrete cosine transform coefficient:
Get image block I
1Green channel I
1g, with image I
1gBe divided into nonoverlapping 8 * 8 image blocks, every image block is carried out discrete cosine transform, discrete cosine transform is defined as:
Wherein, C (u), C are (v)=(2)
-1/2, work as u, v=0; C (u), (v)=1, u, v get other values to C.(i j) is pixel value to f, and (u v) is a discrete cosine transform coefficient to F;
(2-2), the statistical straggling cosine transform coefficient, histogram draws:
(2-2-1), as shown in Figure 1, for each 8 * 8 image block, with 64 positions by the zigzag numbering that sorts.Add up all 8 * 8 image block position i (i=2 ..., 64) discrete cosine transform coefficient located.
(2-2-2), the discrete cosine coefficient value with position i place uniformly-spaced is divided into H the H interval, that different position i is corresponding different:
During i ∈ [2,3], H=600; During i ∈ [4,6], H=350; During i ∈ [7,10], H=150; During i ∈ [11,15], H=130;
During i ∈ [16,21], H=125; During i ∈ [22,28], H=120; During i ∈ [29,36], H=115; During i ∈ [37,43], H=110;
During i ∈ [44,64], H=100.
(2-2-3), construct vectorial dctx
i[dx
1, dx
2..., dx
H], dcty
i[dy
1, dy
2..., dy
H].With vectorial dctx
iBe horizontal ordinate, vectorial dcty
iBe ordinate, the statistic histogram of i place, the position discrete cosine transform coefficient that draws.Wherein, dctx
iIn element value represent the discrete cosine transform coefficient value at i place, position, dcty
iIn element represent the number of corresponding discrete cosine transform coefficient value.
(2-3), find out the maximum point of i place, position discrete cosine transform coefficient statistic histogram:
(2-3-1), to vectorial dcty
iAsk first order derivative, get vectorial dcty '
i, dcty '
i=[ddy
1, ddy
2..., ddy
H-1], wherein, ddy
i=dy
I+1-dy
i(i=1,2 ... H-1);
(2-3-2), with vectorial dcty '
iAdjacent element multiply each other, obtain vectorial hy
i, hy
i=[h
1, h
2..., h
H-2], wherein, h
i=ddy
I+1Ddy
i(i=1,2 ..., H-2);
(2-3-3), search and satisfy h
i≤ 0 element h
i, and the value of record subscript i is stored in vectorial tp
iIn, establish and write down t subscript value, tp altogether
i=[tp
1, tp
2..., tp
t];
(2-3-4), investigate vectorial dcty '
iElement
(i=1,2 ..., t), if satisfy
Record subscript tp
iBe stored in vectorial th
iIn, establish and write down r subscript value, th altogether
i=[th
1, th
2..., th
r];
(2-3-5), the discrete cosine transform coefficient statistic histogram at position i place has r maximum point, for
(j=1,2 ..., r);
(2-4), estimate quantization step q
2And q
3
(2-4-1), establish image block I
1The estimation quantization table be Q, 64 quantization parameters of Q are pressed zigzag ordering, the quantization parameter at position i place is q
i(i=1,2 ..., 64).By obtaining position i (i=2 in the step (2-3), 3) discrete cosine transform coefficient histogram of locating and histogrammic maximum point, investigate histogrammic positive axis, take out all maximum points, wherein the most close null value place, except that maximum value of middle, and by the big or small descending sort of maximum value, the pairing horizontal ordinate xa of each maximum value behind the record ordering
j, composition of vector Xa
i, Xa
i=[xa
1, xa
2... xa
j..., xa
M], wherein the M representative is from the maximum value number of positive axis taking-up.
(2-4-2), negative semiaxis is done and the positive axis identical operations, obtain vectorial Xb
i, Xb
i=[xb
1, xb
2... xb
j..., xb
N], wherein the N representative is from the maximum value number of negative semiaxis taking-up;
(2-4-3), make quantization step
(2-5), estimate quantization step q
1: use quantization step q
2And q
3Provide q
1Estimated value,
(2-6), estimate quantization step q
i(i=4,5..., 64):
(2-6-1), by obtaining position i (i=4 in the step (2-3), 64) discrete cosine transform coefficient histogram of locating and histogrammic maximum point, investigate histogrammic positive axis, take out all maximum points (the most close null value place, be except that maximum value of middle), and by the big or small descending sort of maximum value, the pairing horizontal ordinate xa of each maximum value behind the record ordering
j, composition of vector Xa
i, i.e. Xa
i=[xa
1, xa
2... xa
j..., xa
M], wherein the M representative is from the maximum value number of positive axis taking-up.
If (2-6-2), judge satisfied (q
I-1-xa
j)/q
I-1>0.4 (j gets since 1), wealth makes constant a=xa
jOtherwise to xa
J+1Do identical judgement, up to finding vectorial Xa
iIn qualified element, if Xa
iMiddle M element do not satisfy inequality, then makes a=q
I-1
(2-6-3), negative semiaxis is done and the positive axis identical operations, according to step 1., the method that obtains constant a in is 2. obtained constant b.
(2-6-4), order
Obtained quantization step q successively
2, q
3, q
1, q
4..., q
63, q
64Value, promptly obtained image block I
1Estimation quantization table Qt
1
The method of estimating quantization table among the present invention is compared and is compared with maximal possibility estimation of the prior art [7], histogram power spectrum [5] method, and table 1 has provided the comparing data of estimated time, thereby makes that the whole detection algorithm efficient of distorting is higher.
Table 1 is estimated the time of quantization table
By the estimated time explanation that provides in the table: quantization table of the present invention obviously lacks than other method estimated time estimated time, and then illustrates and the present invention detect jpeg image to distort the time few, the raising detection efficiency.
Claims (4)
1. method that the detection jpeg image of estimating based on quantization table is distorted, it is characterized in that this method at first chooses the some image subblocks of testing image, the quantization table that estimates every image subblock is respectively also made comparisons in twos, detect the testing image operation whether process is cut out, then testing image is drawn piece again, whether cut out quantization table proper vector or the quantization table that extracts every image subblock according to the testing image process, at last, difference degree by adjacent image piece quantization table proper vector or quantization table is classified to image block, judge that whether image is through the synthetic operation of distorting, and detect the tampered region, its concrete steps are as follows:
(1), choosing the n block size is the not overlapped image block I of L * L
1, I
2... I
n
(2), estimate the quantization table of every image block to obtain Qt
1, Qt
2..., Qt
n
(3), with Qt
1, Qt
2..., Qt
nQt is calculated in computing in twos respectively
iAnd Qt
jBetween difference measurement Diff
I, j, do computing in twos:
(4), add up among the vectorial Diff greater than threshold value T
1The number Num of element, judgment threshold T
1The number Num of element, if Num>3n (n-1)/8 judges that then testing image is through cutting out execution in step (5); Otherwise judge not through cutting out, jump to step (10);
(5), choose the image subblock I of jpeg image I to be measured
0, and it is divided into not overlapped image block B
1,1, B
2,1..., B
M, 1
(6), transitional slide B
I, 1(i=1 2...m), gets B
I, j(j=1,2 ..., 64);
(7), estimate B
I, 1(i=1, quantization table proper vector Q 2...m)
i(i=1,2 ..., m);
(8), with image block B
I, 1(i=1,2 ..., m) be divided into: category-A and category-B, initialization: B
1,1Demarcate and be category-A, calculating adjacent image piece B
I-1,1And B
I, 1Diversity factor vector D between the quantization table proper vector
i
(9), judge distorted image: if D
i<T
2, then with B
I, 1Demarcate and be and B
I-1,1Identical class; Otherwise with B
I, 1Demarcate and be and B
I-1,1Different classes, after the classification, if all images piece is category-A, then testing image is not distorted through synthetic; If existing category-A also has category-B, think that then category-A and category-B belong to different image sources, promptly to distort, and judge that category-A or category-B are the tampered region through synthetic, whole testing process finishes;
(10), choose the image subblock I of jpeg image I to be measured
0, and it is divided into not overlapped image block B
1, B
2..., B
m
(11), estimated image piece B
i(i=1,2 ..., quantization table Q m)
i(i=1,2 ..., m);
(12), with image block B
i(i=1,2 ..., m) be divided into: category-A and category-B, initialization: B
1,1Demarcate and be category-A, calculating adjacent image piece B
I-1,1And B
I, 1Difference measurement Δ between the quantization table
i
(13), judge distorted image: if Δ
i<T
2, then with B
I, 1Demarcate and be and B
I-1,1Identical class; Otherwise with B
I, 1Demarcate and be and B
I-1,1Different classes, after the classification, if all images piece is category-A, then testing image is not distorted through synthetic; If existing category-A also has category-B, then category-A belongs to different image sources with category-B, promptly distorts through synthetic, and judges that category-A or category-B are the tampered region, and whole testing process finishes.
2. the method that the detection jpeg image of estimating based on quantization table according to claim 1 is distorted is characterized in that, the estimation quantization table described in above-mentioned steps (2), (7) and (11), and its concrete steps are as follows:
(2-1), calculate discrete cosine transform coefficient;
(2-2), the statistical straggling cosine transform coefficient, histogram draws;
(2-3), find out the maximum point of i place, position discrete cosine transform coefficient statistic histogram;
(2-4), estimate quantization step q
2And q
3
(2-5), estimate quantization step q
1: use quantization step q
2And q
3Provide q
1Estimated value,
(2-6), estimate quantization step q
i(i=4,5..., 64) have obtained quantization step q successively
2, q
3, q
1, q
4..., q
63, q
64Value, promptly finished the estimation of whole quantization table.
3. the method that the detection jpeg image of estimating based on quantization table according to claim 2 is distorted, it is characterized in that, Qmask weighting matrix described in the above-mentioned steps (3), the diversity factor between two quantization tables of this weighting matrix is by they quantization step decisions in Qmask non-zero position.
4. the method that the detection jpeg image of estimating based on quantization table according to claim 3 is distorted, it is characterized in that, transitional slide described in the above-mentioned steps (6), it is with image block from right to left, is step-length transitional slide 63 times with a pixel from bottom to up.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN200910198229A CN101706944A (en) | 2009-11-03 | 2009-11-03 | Quantization table evaluation based method for detecting JPEG image tampering |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN200910198229A CN101706944A (en) | 2009-11-03 | 2009-11-03 | Quantization table evaluation based method for detecting JPEG image tampering |
Publications (1)
Publication Number | Publication Date |
---|---|
CN101706944A true CN101706944A (en) | 2010-05-12 |
Family
ID=42377167
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN200910198229A Pending CN101706944A (en) | 2009-11-03 | 2009-11-03 | Quantization table evaluation based method for detecting JPEG image tampering |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101706944A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102801975A (en) * | 2012-07-31 | 2012-11-28 | 李斌 | Method and device for estimating processing of quantifying step size of pictures |
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 |
CN103501428A (en) * | 2013-10-22 | 2014-01-08 | 北京博威康技术有限公司 | Multimedia monitoring method and multimedia monitoring device |
CN103544692A (en) * | 2012-07-13 | 2014-01-29 | 深圳市智信达软件有限公司 | Blind detection method for tamper with double-compressed JPEG (joint photographic experts group) images on basis of statistical judgment |
CN106960435A (en) * | 2017-03-15 | 2017-07-18 | 华中师范大学 | A kind of double compression automatic testing methods of jpeg image |
CN110147800A (en) * | 2019-05-20 | 2019-08-20 | 哈尔滨工业大学 | Image duplication based on SIFT, which is pasted, distorts blind detection method |
CN112950556A (en) * | 2021-02-07 | 2021-06-11 | 深圳力维智联技术有限公司 | Image truth evaluation method, device and system and computer readable storage medium |
-
2009
- 2009-11-03 CN CN200910198229A patent/CN101706944A/en active Pending
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103544692A (en) * | 2012-07-13 | 2014-01-29 | 深圳市智信达软件有限公司 | Blind detection method for tamper with double-compressed JPEG (joint photographic experts group) images on basis of statistical judgment |
CN102801975A (en) * | 2012-07-31 | 2012-11-28 | 李斌 | Method and device for estimating processing of quantifying step size of pictures |
CN102801975B (en) * | 2012-07-31 | 2014-10-29 | 李斌 | Method and device for estimating processing of quantifying step size of pictures |
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 |
CN102957915B (en) * | 2012-11-15 | 2015-03-25 | 西安理工大学 | Double JPEG (Joint Photographic Experts Group) compressed image-targeted tamper detection and tamper locating method |
CN103501428A (en) * | 2013-10-22 | 2014-01-08 | 北京博威康技术有限公司 | Multimedia monitoring method and multimedia monitoring device |
CN106960435A (en) * | 2017-03-15 | 2017-07-18 | 华中师范大学 | A kind of double compression automatic testing methods of jpeg image |
CN110147800A (en) * | 2019-05-20 | 2019-08-20 | 哈尔滨工业大学 | Image duplication based on SIFT, which is pasted, distorts blind detection method |
CN112950556A (en) * | 2021-02-07 | 2021-06-11 | 深圳力维智联技术有限公司 | Image truth evaluation method, device and system and computer readable storage medium |
CN112950556B (en) * | 2021-02-07 | 2024-05-10 | 深圳力维智联技术有限公司 | Image reality evaluation method, device, system and computer readable storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Wang et al. | Detection of image region duplication forgery using model with circle block | |
CN101706944A (en) | Quantization table evaluation based method for detecting JPEG image tampering | |
Dirik et al. | Image tamper detection based on demosaicing artifacts | |
CN104303192B (en) | Unique identifying information from the feature of labelling | |
CN104933721B (en) | Stitching image altering detecting method based on color filter array characteristic | |
Hussain et al. | Image forgery detection using multi-resolution Weber local descriptors | |
CN102226920B (en) | Cutting-resistant JPEG image compression history and synthetic tamper detection method | |
CN102176208B (en) | Robust video fingerprint method based on three-dimensional space-time characteristics | |
CN103561271B (en) | The moving target of static camera shooting is removed video spatial domain altering detecting method | |
CN106530200A (en) | Deep-learning-model-based steganography image detection method and system | |
CN104598933A (en) | Multi-feature fusion based image copying detection method | |
CN101493938B (en) | Method for detecting cooked image based on noise distribution discipline | |
CN102892048B (en) | Video watermark anti-counterfeiting method capable of resisting geometric attacks | |
Wang et al. | LBP-SVD based copy move forgery detection algorithm | |
CN104408728A (en) | Method for detecting forged images based on noise estimation | |
Hussain et al. | Copy-move image forgery detection using multi-resolution weber descriptors | |
CN102521606A (en) | Method for classifying pixel blocks of JPEG images and image falsification detecting and falsified area positioning methods based on same | |
CN102957915A (en) | Double JPEG (Joint Photographic Experts Group) compressed image-targeted tempertamper detection and tempertamper locating method | |
Hilal et al. | Copy-move forgery detection using principal component analysis and discrete cosine transform | |
CN103310235A (en) | Steganalysis method based on parameter identification and estimation | |
Dixit et al. | Copy-move forgery detection exploiting statistical image features | |
CN106709915B (en) | Image resampling operation detection method | |
Li et al. | Robust median filtering detection based on the difference of frequency residuals | |
CN100481872C (en) | Digital image evidence collecting method for detecting the multiple tampering based on the tone mode | |
Nguyen et al. | Detecting resized double jpeg compressed images–using support vector machine |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Open date: 20100512 |