CN102096895A - Video digital fingerprint method based on run-length coding and one-dimensional discrete forurier transform - Google Patents
Video digital fingerprint method based on run-length coding and one-dimensional discrete forurier transform Download PDFInfo
- Publication number
- CN102096895A CN102096895A CN 201110029329 CN201110029329A CN102096895A CN 102096895 A CN102096895 A CN 102096895A CN 201110029329 CN201110029329 CN 201110029329 CN 201110029329 A CN201110029329 A CN 201110029329A CN 102096895 A CN102096895 A CN 102096895A
- Authority
- CN
- China
- Prior art keywords
- sequence
- fingerprint
- group
- video
- length
- 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
Abstract
The invention relates to a video digital fingerprint method based on run-length coding and one-dimensional discrete forurier transform, belonging to the field of information processing technology. The video digital fingerprint method comprises the steps of distributing a group number for each user and using the only information sequence in the group as an identification code to serve as the fingerprint of a buyer, embedding the user fingerprint into an original video by utilizing a fingerprint embedding method to generate a copy containing the user fingerprint before the video is published and finally after the copy video to be detected is received, accurately judging one of the conspirators by utilizing a conspirator detection method. The video copy generated by the invention has favorable visual effect, very strong robustness to attacks of conspiracy attack, video compression, frame dropping, frame changing and the like and not high complexity and good real-time in the extraction and the detection of the fingerprint.
Description
Technical field
What the present invention relates to is a kind of method of technical field of information processing, specifically is a kind of digital video fingerprint method based on Run-Length Coding and one-dimensional discrete Fourier transform.
Background technology
The fast development of infotech makes all kinds of digital medium informations propagate fast by the diverse network approach, only needs very low cost just can obtain high quality multimedia product copy.How digital product is carried out copyright protection and become one of key problem of information age copyright protection.Digital finger-print is a kind of new Digital copyright protection technology.Before digital media products issue, in primary products, embed the information relevant earlier with the buyer, i.e. user's digital finger-print information, each user's fingerprint has nothing in common with each other; After the publisher finds copy right piracy, just can pass through to extract the fingerprint in the pirate product, determine the source of bootlegging, the bootlegger is prosecuted, thereby play the effect of copyright protection.This patent is an object with the video, has proposed the real-time video fingerprint technique of a kind of anti-conspiracy attack and video compress.
Through existing literature search is found that correlation technique is as follows:
People such as Wang are at paper " Group-oriented fingerprinting for multimedia forensics (the packet-based fingerprint that is used for the multimedia messages evidence obtaining) " (EURASIP journal on applied signal processing (EURASIP uses the signal Processing periodical), vol.2004, no.14, pp.2153-2173 has proposed packet-based fingerprint schemes in Oct.2004).This scheme uses the pseudo-random sequence of obeying standardized normal distribution as finger print information.At first, distribute mutually orthogonal pseudo-random sequence as user profile to different user; Then, the user is divided into groups, the user who most possibly initiates conspiracy attack is each other divided in same group according to prior imformation; User in giving same group distributes the group information of same pseudo-random sequence as this group user, and group information on the same group is not mutually orthogonal; Group information is connected the fingerprint that just obtains the user with user profile; At last, fingerprint is embedded in discrete cosine transform (the DCT:Discrete Cosine Transform) coefficient of carrier.When detecting the collaborator, judge the group at collaborator place earlier by related operation, in group, judge the collaborator by related operation then.Compare the quadrature fingerprint, this algorithm can improve correct detection collaborator's probability, and the related operation number of times reduces; But shortcoming is employed pseudo-random sequence quantity still to be directly proportional with total number of users.
People such as Naoki are at paper " Collusion-resistant fingerprinting scheme based on the CDMAtechnique (based on the anti-collusion digital fingerprinting scheme of CDMA (Code Division Multiple Access)) " (International Workshop onSecurity, Nara, Japan (information security international symposium in 2007), Oct.2007, LNCS, vol.4752, a kind of fingerprint schemes based on CDMA (CDMA:Code Division Multiple Access) technology has been proposed pp.28-43), its method is: the user is divided into groups, and each user distributes a group # and Customs Assigned Number as its identification code; Picture is carried out overall dct transform, choose the DCT coefficient sequence that a part of medium and low frequency coefficient is formed two equal in length; Adopt CDMA technology that Customs Assigned Number and group # are embedded into respectively in two DCT coefficient sequence; All DCT coefficients are carried out the picture that the DCT inverse transformation obtains containing fingerprint.After finding suspicious copy, at first picture is carried out dct transform, select the DCT coefficient sequence of having carried user identification code; Adopt the CDMA technology information that takes the fingerprint in conjunction with original image, detect a plurality of collaborators.This scheme does not need related operation when detecting the collaborator, and detection complexity is low, but when the conspiracy number of users is big, a lot of innocent persons can be judged to be the collaborator.
On the other hand, because the video data volume after the digitizing is very huge, be not easy to transmission and storage; And exist extremely strong correlativity in the video data, and reduce the video data volume by the data compression means, store and transmission of video in the mode of compressed encoding, can effectively save storage space and improve video transmission efficient.Therefore, a good video finger print method at first must be highly resistant to various video-frequency compression methods, as MPEG4, H.264 etc.Find that by literature search according to the difference of embedded location, present video finger print method can be divided three classes, promptly in original video stream, embed fingerprint, in compression domain, embed and the embedding of the video flowing after compression.Z.Zhao etc. have proposed a kind of fast video watermarking algorithm in article " A novel video watermarking scheme in compressed domain based onfast motion estimation " (a kind of new type of compression domain video watermark scheme based on fast motion estimation), promptly the motion vector of compressed video bitstream is done minimum change, simultaneously thereby their method for quick estimating that also proposes a kind of motion vector makes this watermarking algorithm highly effective, but this method is confined to a certain concrete video compression standard, if with another compress mode video is compressed, this method promptly lost efficacy.Watermark is directly embedded in the bit stream after overcompression, biggest advantage is that the fingerprint that embeds does not need through complete coding and decoding process, influence to vision signal is less, this is very important for real-time watermark embedded technology, because the motion estimation pass of many compression schemes all requires high-intensity calculating; But video system will limit the embedding quantity of information of watermark to the constraint of rate of video compression code, may exert an influence to the motion compensation loop simultaneously, obviously increase the complexity of this method for compensating measure that this influence takes; Also having some video watermark technology is to embed in the spatial domain of original video or other transform domains, Deguillaume etc. have proposed the video finger print method of a kind of 3 d-dem Fourier transform (3DDFT:3-Dimensional Discrete Fourier Transform) in article " Robust 3D DFT videowatermarking " (the 3 d-dem Fourier transform video watermark of robust), this method can not be confined to a certain video compression standard, but, often not too be applicable to real-time video watermarking because the complexity that they calculate is very high.
In sum, present digital video fingerprint method still can not and extract real-time in resistance to compression, embedding, conspire to detect and reach gratifying effect simultaneously aspect four of accuracy and the finger print information amounts.
Summary of the invention
The present invention is directed to the prior art above shortcomings, a kind of digital video fingerprint method based on Run-Length Coding and one-dimensional discrete Fourier transform is provided, the finger print information amount that this method needs seldom, the video copy visual effect that generates is good, to conspiracy attack, video compress, frame losing, change attacks such as frame and have very strong robustness, and the extraction of fingerprint and the complexity of detection are not high, and real-time is good.
The present invention is achieved by the following technical solutions, the present invention includes following steps:
The first step, the user is divided into groups, distributes information sequence unique in a group # and the group as its identification code for each user,, specifically may further comprise the steps as buyer's fingerprint:
1.1) according to features such as the region of living in of buyer in the actual environment, social bonds, all users are divided into the N group, the user of most possible generation contact is divided in same group;
1.2) group information sequence g that to give length of each set of dispense be lg
i(1≤i≤N), mutually orthogonal, i.e. g between the group information sequence
i* g
i=0,1≤i≤N, 1≤j≤N, i ≠ j;
1.3) suppose that every group can be held 2 at most
M-1-1 user, give each user's allocated length in the group for the m first place be 1 positive and negative 1 sequence as user profile w and non-complete 1 sequence of w, the information sequence that i organizes j user is designated as w
Ij, wherein: 1≤i≤N, 1≤j≤2
M-1-1;
1.4) information sequence w is carried out Run-Length Coding: even 1 section is 1 distance of swimming, even-1 section is-1 distance of swimming, 1 distance of swimming and-1 distance of swimming distribute alternately, the length of this distance of swimming is that even 1 or even-1 number is formed a sequence with the length value of each distance of swimming by former order, generates a polynary distance of swimming sequence w ';
1.5) in w, except that the 1st, 1 or-1 probability that occurs was 1/2 during other were every.According to the character of Run-Length Coding, length is that the probability that 30 the distance of swimming occurs is (1/2)
30, can be approximately 0.Generate the mapping table of a S (x) to E (x), wherein, S (i)=i, E (i) is that a length is positive and negative 1 sequence of n, 1≤i≤30, and E (i) is mutually orthogonal in twos, with reference to above-mentioned S (x)～E (x) table of comparisons, each i=S (i) of sequence w ' is replaced with corresponding E (i) sequence, thereby multi-element code sequence w ' has been become binary code sequence w ";
1.6) with the group number information sequence g of each group
i, the user profile sequence w in 1≤i≤N and this group
Ij", 1≤j≤num
i, num
iThe number of users of representing the i group, combining is exactly the original fingerprint sequence orf of j user in the i group
Ij, to orf
IjSpread spectrum obtains final fingerprint preface rice and draws up f
Ijf
IjBy gpn
iAnd wpn
Ij" two parts are formed, wherein, and gpn
iAnd wpn
Ij" be respectively by to g
iAnd w
Ij" spread spectrum obtains.
Described spread spectrum comprises two steps:
I) producing a length, to equal L and element be positive and negative 1 binary pseudo-random sequence PN={a
1, a
2..., a
L, a wherein
i=1 or-1,1＜=i＜=L.This binary pseudo-random sequence is by being that 0 in 0 and 1 the pseudo-random sequence is mapped as-1 and obtains with element;
Ii) with sequence PN and orf
IjThe element of each multiplies each other, and with the sequence that obtains that multiplies each other this position is replaced then, so just obtains embedding the fingerprint sequence f of video
Ijf
IjLength can not surpass [N+m*n] * L.That is,
orf
ij＝{g
i?w
j″}＝{g
i1，g
i2，...，g
ilg，w
ij1″，w
ij2″，...，w
ijlw″}，
f
ij＝{g
i1＊PN，g
i2＊PN，...，g
ilg＊PN，w
ij1″＊PN，w
ij2″＊PN，...，w
ijlw″＊PN}。
Described element is that 0 and 1 pseudo-random sequence is m sequence, M sequence or Gold sequence.
Second step, before the issue video, adopt fingerprint embedding method that user fingerprints is embedded in the original video, generate the copy that contains user fingerprints, specifically may further comprise the steps:
2.1) utilize the camera lens split plot design that video is divided into several GOP, contain the D two field picture among each GOP, wherein: D is certain values to all GOP that obtained by same Video Segmentation, the size of each frame is the M*N pixel;
2.2) each GOP is done 1D DFT conversion along time shaft, obtain a D dimension group F (u, v, x), 1≤u≤N, 1≤v≤M, 1≤x≤D is with F (u, v, x) form with matrix is saved in the database, and a pair of intermediate frequency frame of choosing about DC frame symmetry embeds, and the telescopiny between them is independent of fully identical;
2.3) each pixel of each frame all is a plural number after the DFT conversion, each of user fingerprints is embedded in the amplitude information of choosing pixel corresponding in the frame successively, two-dimensional array M (u after obtaining embedding, v, τ)=a*f (i)+amplitude (F (u, v, i)), wherein: (u, v are to embed the two-dimensional array that obtains behind the fingerprint τ) to M; F (i) is the i position of the fingerprint sequence that will embed; A is the intensity that digital finger-print embeds, and is a regulatable parameter; Amplitude information is got in amplitude () expression.
The position of described embedding is: embed fingerprint f
IjLength be designated as lenf
Ij, in the τ frame of GOP, select lenf
IjIndividual pixel composition sequence v
Ij={ v
Ij(1), v
Ij(1) ...., v
Ij(lenf
Ij) be used to embed fingerprint.
2.4) transform frame that embeds fingerprint is carried out 1D DFT inverse transformation on the time shaft, obtain embedding the video behind the fingerprint.
The 3rd step, after receiving copy video to be measured, adopt collaborator's detection method, accurately judge one of them collaborator, specifically may further comprise the steps:
3.1) the suspicious video that will embed fingerprint utilizes the camera lens split plot design that video is divided into the GOP of several length for D, the size of D value is consistent with built-in end;
3.2) each GOP is done ID DFT conversion along time shaft, obtain a D dimension group F (u, v, x), 1≤u≤N, 1≤v≤M, 1≤x≤D;
3.3) select the target frame F ' that embeds fingerprint (u, v, τ), by ((τ)-sequence that F ' (u, v, τ))/a obtains is exactly the fingerprint sequence f ' that extracts, τ is the embedding frame of fingerprint to F for u, v;
3.4) fingerprint sequence is separated spread spectrum: is that unit carries out segmentation with f ' with the L bit, and wherein L is the length of built-in end frequency expansion sequence PN, that is: f '={ f ' (1), f ' (2), f ' (3) ... }={ seg (1), seg (2), seg (3), ..., wherein: seg (i)={ f ' ((i-1) * L+1), f ' ((i-1) * L+2), ..., f ' is (i*L) }; Successively each section and the PN sequence that built-in end produces are carried out computing cross-correlation then, when the result who obtains greater than zero, judge that then this section separates that the output bit is 1 behind the spread spectrum, when the result less than zero, then be judged to be-1, so obtain separating the positive and negative 1 information sequence f behind the spread spectrum ";
3.5) judge the group at assailant place by computing cross-correlation: with f " in the sequence FI that forms of the top n bit of set of landmarks information organize information sequence g with built-in end N
i, 1≤i≤N makes similarity relatively successively, if FI and g among the gained result
k(similarity of 1≤k≤N) is the highest, thinks that then the k group is the group at assailant place.
3.6) with the information sequence w before the spread spectrum of all users in the k group
Kj" (1≤j≤num
k, num
kThe number of users of representing k group) and f " in remove group information part be that unit carries out segmentation with the n bit, n is the length of E (x) sequence in built-in end S (x)～E (x) table of comparisons, generation length is the sequence w of n
Kj1", w
Kj2" ..., w
Kjlw", f "
1, f "
2..., f "
Len, len=[length (f ")-length (FI)]/n, with w
Kjm" and f
m" carry out similarity successively relatively, comparative result is placed on vectorial SIM
Kj(m) in, 1≤m≤len
KjIf, lw
Kj＜len
Kj, SIM then
Kj(m)=1, lw
Kj＜m≤len
KjIf lw
Kj＞len
Kj, SIM then
KjBe a null vector, final by statistics SIM
KjIn zero number, by relatively, the user of the SIM sequence correspondence that the zero number is minimum is defined as the assailant.
Compare with existing video finger print method, the invention has the advantages that: first, in traditional fingerprint spectrum spreading method, the quantity of orthogonal spreading sequence is directly proportional with total number of users, the present invention is by introducing Run-Length Coding in cataloged procedure, make fingerprint sequence become multi-element code by dual code, the quantity of orthogonal spreading sequence equals symbol value maximum in all multi-element code, has greatly reduced the quantity of orthogonal spreading sequence.The second, the present invention has reduced the embedding data volume effectively by original fingerprint being carried out Run-Length Coding and twice spread spectrum obtains embedding fingerprint, has avoided the huge problem of fingerprint data volume in traditional spectrum spreading method.The 3rd, the present invention does 1D DFT conversion along time shaft to each GOP, thereby obtain the embedding territory of fingerprint, this embeds the territory in retaining space information, also preserved the time-domain information of video well, therefore in the amplitude information in this territory, embed fingerprint and be highly resistant to video compress, frame losing, change attack such as frame, and have good vision disguise.
Embodiment
Below embodiments of the invention are elaborated, present embodiment is being to implement under the prerequisite with the technical solution of the present invention, provided detailed embodiment and concrete operating process, but protection scope of the present invention is not limited to following embodiment.
It is initial carrier for the bus.yuv video of 352*288 that present embodiment adopts size, and this video is exactly a GOP in fact, comprises 50 two field pictures, specifically may further comprise the steps:
The first step generates buyer's fingerprint: distribute a group number and unique user identification code for the buyer of each digital product.
1.1) in the present embodiment, all users are divided into 15 groups, every group of 1000 users.
1.2) 15 groups group information represents with the 2nd row to the 16th row column vectors of 16 dimension Hadamard matrixes respectively, i.e. g
i=Hadamard (:, i+1), 1＜=i＜=15.
1.3) suppose that every group can be held 2 at most
20-1 user, give every group in different users distribute unique length be 21 first places be 1 positive and negative 1 sequence as user profile w, the information sequence that i organizes j user is designated as w
Ij(1＜=i＜=15,1＜=j＜=1000).Then this system of fingerprints maximum number of users that can hold is Nu=15* (2
20-1).
1.4) with user profile sequence w
IjCarry out Run-Length Coding, the length value that is about to positive and negative 1 distance of swimming is formed a sequence by former order, generates a polynary distance of swimming sequence w
Ij'.
1.5) generate the table of comparisons of a S (x)～E (x).Wherein, S (i)=i, owing to need to generate 30 orthogonal vector, adopt dimension more than or equal to 30 Hadamard matrix for making things convenient in the present embodiment, consider the character of Hadamard matrix, generate the Hadamard matrix H dam of one 32 dimension in the present embodiment, E (i) is the i+1 row column vector of Hdam matrix, 1≤i≤30.With reference to above-mentioned S (x)～E (x) table of comparisons, with each sequence displacement of sequence w ' with correspondence in the table, thereby with multi-element code sequence w
Ij' become binary code sequence w
Ij".For example, suppose w
Ij={ 1-111-1-1-1-1111-111-1-1-1-1-1-11}, the w that generates behind the Run-Length Coding
Ij'={ 11243126 1}, this w
Ij' corresponding w
Ij"={ Hdam (:, 2) ', Hdam (:, 2) '; Hdam (:, 3) ', Hdam (:, 5) '; Hdam (:, 4) ', Hdam (:, 2) '; Hdam (:, 3) ', Hdam (:; 7) ', Hdam (:, 2) ' }; wherein Hdam (:, j) ' expression Hadamard matrix in the commentaries on classics rank vector of column vector of j row, 2≤j≤31.
1.6) user's group information sequence adds the user fingerprints orf before information sequence has just formed spread spectrum
IjUse primitive polynomial x
4+ x generate length be 31 and element be 0 and 1 m sequence, be mapped to-1 with 0, obtain an element and be positive and negative 1 binary pseudo-random sequence PN={a
1, a
2..., a
31.The user fingerprints orf that will not have spread spectrum
IjEach multiply each other with the PN sequence successively, be that 31 pseudo-random sequence replaces the bit value on this with the length that obtains then.The sequence of gained is exactly the fingerprint sequence that will embed video.
Second step, embed buyer's fingerprint: suppose and to give 100 users with the bus.yuv video distribution, then before this digital product of issue, adopt fingerprint embedding method that these 100 parts of different user fingerprints are embedded into respectively in the product original copy earlier, generate 100 parts of video copies that contain user fingerprints.
2.1) in the present embodiment, the video of choosing itself is exactly a complete camera lens, includes abundant movable information, includes 50 two field pictures in this camera lens, the size of every width of cloth image is 352*288;
2.2) camera lens is done 1D DFT conversion along time domain, obtain one 50 dimension group F (u, v, x), 1≤x≤50, wherein F (u, v, 1) is direct current (DC) frame, chooses intermediate frequency frame the 10th frame and embeds.
2.3) select suitable position that fingerprint is embedded in the amplitude of the 10th frame respective pixel, embedding formula is as follows:
M(u，v，10)＝a＊w(i)+amplitude(F(u，v，10)
Wherein, M (u, v, 10) embeds the two-dimensional array that obtains behind the fingerprint; W (i) is the i position of the fingerprint sequence that will embed; A is the intensity that digital finger-print embeds, and is a regulatable parameter.
The embedding point is chosen rule and is: will embed fingerprint f
IjLength be designated as lenf
Ij, select lenf at the 10th frame of GOP
IjIndividual pixel composition sequence v
Ij={ v
Ij(1), v
Ij(2) ..., v
Ij(lenf
Ij) be used to embed fingerprint, v
Ij(k) (1≤k≤lenf
Ij) take from frame be positioned at 10+ (kmod 31)+[k/ (31*20)] OK, the pixel of 300-(k mod 31)-[k/ (31*20)] row, wherein the maximum integer that is not more than x is got in [x] expression.
2.4) frame of video that embeds fingerprint is done 1D DFT inverse transformation on the time shaft, obtain embedding the video behind the fingerprint.
After finishing said process, just can give corresponding user with the video distribution that has embedded fingerprint.
In the present embodiment,, calculate these the 100 parts Y-PSNR PSNR that embed the legal video of fingerprints respectively by following formula for testing the fidelity of the legal video that embeds fingerprint:
Wherein g (m, n) and h (m n) represents original embedding frame and contain the gray-scale value of fingerprint frame respectively.Calculating these the 100 parts mean P SNR that embed the fingerprint video is 39.2571dB, and the fidelity of visible video is fine.
In addition, in the present embodiment, for test described fingerprint embedded mode to video compress, frame losing, change the robustness that frame etc. is attacked, these the 100 parts video copies that embedded fingerprint are carried out H.264 compression and decompression, the average probability that can correctly detect fingerprint is 0.98, is 0.01 with innocent person's erroneous judgement for assailant's average probability; Then these 100 parts of video copies are carried out frame losing and attack, lose last frame, the average probability that can correctly detect fingerprint is 0.88, the probability of erroneous judgement is 0.01, lose last two frames, the average probability that can correctly detect fingerprint is 0.71, and the probability of erroneous judgement is 0.03.As seen, the present invention has well anti-aggressive.
After obtaining having embedded the legal video of fingerprint, some dishonest users can join together their copy to carry out conspiracy attack, generate the copy that digital finger-print is weakened, distort or delete.Five users of picked at random conspire in the present embodiment, and they average the respective pixel value of the color video that obtains, obtain a new yuv video, H.264 video are compressed to obtain conspiring copy then, carry out illegal distribution at last.In the present embodiment, with this Cheng Chongfu of this average attack 20 times.
In the 3rd step, the collaborator detects: the copyright owner adopts collaborator's detection method after having obtained the suspicious copy of illegal distribution, accurately obtain one of them collaborator.This method adopts non-blind Detecting to survey, and detects step and is:
3.1) GOP that obtains is done ID DFT conversion along time shaft, (u, v x), select the target frame F ' (u, v, 10) that embeds fingerprint to obtain one 50 dimension group F ';
3.2) be exactly the fingerprint sequence f ' that extracts with the formula sequence that (F ' (u, v, 10)-F (u, v, 10))/a obtains.
3.3) be that unit carries out segmentation (L is the length of built-in end frequency expansion sequence PN) with f ' with the L bit, be f '={ f ' (1), f ' (2), f ' (3), ...={ seg (1), seg (2), seg (3) ..., seg (i)={ f ' ((i-1) * L+1), f ' ((i-1) * L+2) ..., f ' is (i*L) }.Successively each section and the PN sequence that built-in end produces are carried out computing cross-correlation then, if the result who obtains, then judges this section greater than zero and separates that the output bit is 1 behind the spread spectrum, if less than zero, then be judged to be-1, so obtain separating the positive and negative 1 information sequence f behind the spread spectrum ".
3.4) will separate the sequence FI that the top n bit of the information sequence that spread spectrum obtains forms and do the similarity comparison with the 2nd to N column vector of N dimension Hadamard matrix, choose the group k (k is not unique) of similarity maximum, 1＜=k＜=N-1, for example: the similarity maximum of the 5th row column vector of FI and Hadamard matrix, then one of seat offence person is the user in the 4th group.
3.5) with the spread spectrum information sequence w of all users in the 4th group
4j" (1＜=j＜=num
4, num
4Represent the 4th group number of users) and f " in remove group information part be that unit carries out segmentation (n is the length of E (x) sequence in built-in end S (x)～E (x) table of comparisons) with the n bit respectively, generate the sequence w that length is n
4j"={ w
4j1", w
4j2" ..., w
4jlw" }, f
i"={ f
1", f
2" ..., f
Len" }, len=[length (f ")-length (FI)]/n.With w
4j" and f
i" carry out similarity successively relatively, comparative result is placed on vectorial SIM
4j(k) in, 1＜=k＜=len; If lw
4j＜len, then SIM
4j(k)=1, lw
4j＜k≤len; If lw then
4j＞len, then SIM
4jIt is a null vector.Statistics SIM
4jIn zero number, by relatively, the SIM that the zero number is minimum
4j(1＜=j＜=num
4, num
4Representing the 4th group number of users) user of sequence correspondence is defined as the assailant.
In the present embodiment, the copy that 20 average attack are obtained detects, when signal to noise ratio (S/N ratio) be-during 5db, the average probability that can correctly detect one of them assailant is 0.98.As seen, the present invention has very high conspiracy detection performance.
Claims (7)
1. the digital video fingerprint method based on Run-Length Coding and one-dimensional discrete Fourier transform is characterized in that, may further comprise the steps:
The first step, the user is divided into groups, distribute information sequence unique in a group # and the group as its identification code, as buyer's fingerprint for each user;
Second step, before the issue video, adopt fingerprint embedding method that user fingerprints is embedded in the original video, generate the copy that contains user fingerprints;
The 3rd step, after receiving copy video to be measured, adopt collaborator's detection method, accurately judge one of them collaborator.
2. the digital video fingerprint method based on Run-Length Coding and one-dimensional discrete Fourier transform according to claim 1 is characterized in that the described first step specifically may further comprise the steps:
1.1) according to features such as the region of living in of buyer in the actual environment, social bonds, all users are divided into the N group, the user of most possible generation contact is divided in same group;
1.2) group information sequence g that to give length of each set of dispense be lg
i(1≤i≤N), mutually orthogonal, i.e. g between the group information sequence
i* g
j=0,1≤i≤N, 1≤j≤N, i ≠ j;
1.3) suppose that every group can be held 2 at most
M-1-1 user, give each user's allocated length in the group for the m first place be 1 positive and negative 1 sequence as user profile w and non-complete 1 sequence of w, the information sequence that i organizes j user is designated as w
Ij, wherein: 1≤i≤N, 1≤j≤2
M-1-1;
1.4) information sequence w is carried out Run-Length Coding: even 1 section is 1 distance of swimming, even-1 section is-1 distance of swimming, 1 distance of swimming and-1 distance of swimming distribute alternately, the length of this distance of swimming is that even 1 or even-1 number is formed a sequence with the length value of each distance of swimming by former order, generates a polynary distance of swimming sequence w ';
1.5) in w, except that the 1st, 1 or-1 probability that occurs was 1/2 during other were every.According to the character of Run-Length Coding, length is that the probability that 30 the distance of swimming occurs is (1/2)
30, can be approximately 0.Generate the mapping table of a S (x) to E (x), wherein, S (i)=i, E (i) is that a length is positive and negative 1 sequence of n, 1≤i≤30, and E (i) is mutually orthogonal in twos, with reference to above-mentioned S (x)～E (x) table of comparisons, each i=S (i) of sequence w ' is replaced with corresponding E (i) sequence, thereby multi-element code sequence w ' has been become binary code sequence w ";
1.6) with the group number information sequence g of each group
i, the user profile sequence w in 1≤i≤N and this group
Ij", 1≤j≤num
i, num
iThe number of users of representing the i group, combining is exactly the original fingerprint sequence orf of j user in the i group
Ij, to orf
IjSpread spectrum obtains final fingerprint preface rice and draws up f
Ijf
IjBy gpn
iAnd wpn
Ij" two parts are formed, wherein, and gpn
iAnd wpn
Ij" be respectively by to g
iAnd w
Ij" spread spectrum obtains.
3. the digital video fingerprint method based on Run-Length Coding and one-dimensional discrete Fourier transform according to claim 1 is characterized in that, described spread spectrum comprises two steps:
I) producing a length, to equal L and element be positive and negative 1 binary pseudo-random sequence PN={a
1, a
2..., a
L, a wherein
i=1 or-1,1＜=i＜=L.This binary pseudo-random sequence is by being that 0 in 0 and 1 the pseudo-random sequence is mapped as-1 and obtains with element;
Ii) with sequence PN and orf
IjThe element of each multiplies each other, and with the sequence that obtains that multiplies each other this position is replaced then, so just obtains embedding the fingerprint sequence f of video
Ijf
IjLength can not surpass [N+m*n] * L.That is,
orf
ij＝{g
i?w
j″}＝{g
i1，g
i2，...，g
ilg，w
ij1″，w
ij2″，...，w
ijlw″}，
f
ij＝{g
i1＊PN，g
i2＊PN，...，g
ilg＊PN，w
ij1″＊PN，w
ij2″＊PN，...，w
ijlw″＊PN}。
4. the digital video fingerprint method based on Run-Length Coding and one-dimensional discrete Fourier transform according to claim 1 is characterized in that, described element is that 0 and 1 pseudo-random sequence is m sequence, M sequence or Gold sequence.
5. the digital video fingerprint method based on Run-Length Coding and one-dimensional discrete Fourier transform according to claim 1 is characterized in that, described second step specifically may further comprise the steps:
2.1) utilize the camera lens split plot design that video is divided into several GOP, contain the D two field picture among each GOP, wherein: D is certain values to all GOP that obtained by same Video Segmentation, the size of each frame is the M*N pixel;
2.2) each GOP is done 1D DFT conversion along time shaft, obtain a D dimension group F (u, v, x), 1≤u≤N, 1≤v≤M, 1≤x≤D is with F (u, v, x) form with matrix is saved in the database, and a pair of intermediate frequency frame of choosing about DC frame symmetry embeds, and the telescopiny between them is independent of fully identical;
2.3) each pixel of each frame all is a plural number after the DFT conversion, each of user fingerprints is embedded in the amplitude information of choosing pixel corresponding in the frame successively, two-dimensional array M (u after obtaining embedding, v ,=a*f (i)+amplitude (F (u, v, i)), wherein: (u, v are to embed the two-dimensional array that obtains behind the fingerprint τ) to M; F (i) is the i position of the fingerprint sequence that will embed; A is the intensity that digital finger-print embeds, and is a regulatable parameter; Amplitude information is got in amplitude () expression;
2.4) transform frame that embeds fingerprint is carried out 1D DFT inverse transformation on the time shaft, obtain embedding the video behind the fingerprint.
6. the digital video fingerprint method based on Run-Length Coding and one-dimensional discrete Fourier transform according to claim 5 is characterized in that the position of described embedding is: embed fingerprint f
IjLength be designated as lenf
Ij, in the τ frame of GOP, select lenf
IjIndividual pixel composition sequence v
Ij={ v
Ij(1), v
Ij(1) ...., v
Ij(lenf
Ij) be used to embed fingerprint.
7. the digital video fingerprint method based on Run-Length Coding and one-dimensional discrete Fourier transform according to claim 5 is characterized in that, described the 3rd step specifically may further comprise the steps:
3.1) the suspicious video that will embed fingerprint utilizes the camera lens split plot design that video is divided into the GOP of several length for D, the size of D value is consistent with built-in end;
3.2) each GOP is done ID DFT conversion along time shaft, obtain a D dimension group F (u, v, x), 1≤u≤N, 1≤v≤M, 1≤x≤D;
3.3) select the target frame F ' that embeds fingerprint (u, v, τ), by ((τ)-sequence that F ' (u, v, τ))/a obtains is exactly the fingerprint sequence f ' that extracts, τ is the embedding frame of fingerprint to F for u, v;
3.4) fingerprint sequence is separated spread spectrum: is that unit carries out segmentation with f ' with the L bit, and wherein L is the length of built-in end frequency expansion sequence PN, that is: f '={ f ' (1), f ' (2), f ' (3) ... }={ seg (1), seg (2), seg (3), ..., wherein: seg (i)={ f ' ((i-1) * L+1), f ' ((i-1) * L+2), ..., f ' is (i*L) }; Successively each section and the PN sequence that built-in end produces are carried out computing cross-correlation then, when the result who obtains greater than zero, judge that then this section separates that the output bit is 1 behind the spread spectrum, when the result less than zero, then be judged to be-1, so obtain separating the positive and negative 1 information sequence f behind the spread spectrum ";
3.5) judge the group at assailant place by computing cross-correlation: with f " in the sequence FI that forms of the top n bit of set of landmarks information organize information sequence g with built-in end N
i, 1≤i≤N makes similarity relatively successively, if FI and g among the gained result
k(similarity of 1≤k≤N) is the highest, thinks that then the k group is the group at assailant place;
3.6) with the information sequence w before the spread spectrum of all users in the k group
Kj" (1≤j≤num
k, num
kThe number of users of representing k group) and f " in remove group information part be that unit carries out segmentation with the n bit, n is the length of E (x) sequence in built-in end S (x)～E (x) table of comparisons, generation length is the sequence w of n
Kj1", w
Kj2" ..., w
Kjlw", f "
1, f "
2..., f "
Len, len=[length (f ")-length (FI)]/n, with w
Kjm" and f
m" carry out similarity successively relatively, comparative result is placed on vectorial SIM
Kj(m) in, 1≤m≤len
KjIf, lw
Kj＜len
Kj, SIM then
Kj(m)=1, lw
Kj＜m≤len
KjIf lw
Kj＞len
Kj, SIM then
KjBe a null vector, final by statistics SIM
KjIn zero number, by relatively, the user of the SIM sequence correspondence that the zero number is minimum is defined as the assailant.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110029329 CN102096895A (en) | 2011-01-21 | 2011-01-21 | Video digital fingerprint method based on run-length coding and one-dimensional discrete forurier transform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN 201110029329 CN102096895A (en) | 2011-01-21 | 2011-01-21 | Video digital fingerprint method based on run-length coding and one-dimensional discrete forurier transform |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN102096895A true CN102096895A (en) | 2011-06-15 |
Family
ID=44129978
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN 201110029329 Pending CN102096895A (en) | 2011-01-21 | 2011-01-21 | Video digital fingerprint method based on run-length coding and one-dimensional discrete forurier transform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN102096895A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110717151A (en) * | 2019-09-04 | 2020-01-21 | 湖南遥昇通信技术有限公司 | Digital fingerprint processing and signature processing method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1266587A (en) * | 1998-03-04 | 2000-09-13 | 皇家菲利浦电子有限公司 | Method and device for watermark |
| WO2004044820A1 (en) * | 2002-11-12 | 2004-05-27 | Koninklijke Philips Electronics N.V. | Fingerprinting multimedia contents |
| CN101079147A (en) * | 2007-06-25 | 2007-11-28 | 中山大学 | Multiple bit digital watermark method capable of resisting printing, scanning and geometric transformation |
| CN101330610A (en) * | 2008-07-22 | 2008-12-24 | 华为技术有限公司 | Method and apparatus for embedding and extracting watermark as well as processing system |
| CN101872398A (en) * | 2010-06-13 | 2010-10-27 | 上海交通大学 | Anti-collusion digital fingerprinting method based on code division multiple access and diversity technology |
-
2011
- 2011-01-21 CN CN 201110029329 patent/CN102096895A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1266587A (en) * | 1998-03-04 | 2000-09-13 | 皇家菲利浦电子有限公司 | Method and device for watermark |
| WO2004044820A1 (en) * | 2002-11-12 | 2004-05-27 | Koninklijke Philips Electronics N.V. | Fingerprinting multimedia contents |
| CN101079147A (en) * | 2007-06-25 | 2007-11-28 | 中山大学 | Multiple bit digital watermark method capable of resisting printing, scanning and geometric transformation |
| CN101330610A (en) * | 2008-07-22 | 2008-12-24 | 华为技术有限公司 | Method and apparatus for embedding and extracting watermark as well as processing system |
| CN101872398A (en) * | 2010-06-13 | 2010-10-27 | 上海交通大学 | Anti-collusion digital fingerprinting method based on code division multiple access and diversity technology |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110717151A (en) * | 2019-09-04 | 2020-01-21 | 湖南遥昇通信技术有限公司 | Digital fingerprint processing and signature processing method |
| CN110717151B (en) * | 2019-09-04 | 2021-05-04 | 湖南遥昇通信技术有限公司 | Digital fingerprint processing and signature processing method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | A new video watermarking algorithm based on 1D DFT and Radon transform | |
| Bhatnagar et al. | A new robust adjustable logo watermarking scheme | |
| CN104053074B (en) | Video watermarking method based on depth image and Otsu segmentation | |
| Chen et al. | Robust contourlet-based blind watermarking for depth-image-based rendering 3D images | |
| CN106454367B (en) | Multimedia file identification method and system | |
| Singh | A robust rotation resilient video watermarking scheme based on the SIFT | |
| Cui et al. | A robust DIBR 3D image watermarking algorithm based on histogram shape | |
| Masoumi et al. | A blind spatio-temporal data hiding for video ownership verification in frequency domain | |
| CN102117474B (en) | Digital picture watermark embedding and detecting method and device | |
| Kumar et al. | A multi-image hiding technique in dilated video regions based on cooperative game-theoretic approach | |
| Luo et al. | Stereo image watermarking scheme for authentication with self-recovery capability using inter-view reference sharing | |
| Masoumi et al. | Content protection in video data based on robust digital watermarking resistant to intentional and unintentional attacks. | |
| Chen et al. | Robust contourlet-based watermarking for depth-image-based rendering 3D images | |
| Guan et al. | A digital blind watermarking scheme based on quantization index modulation in depth map for 3D video | |
| CN102096895A (en) | Video digital fingerprint method based on run-length coding and one-dimensional discrete forurier transform | |
| Ahuja et al. | Robust Video Watermarking Scheme Based on Intra-Coding Process in MPEG-2 Style. | |
| Chongtham et al. | A copyright protection scheme for videos based on the SIFT | |
| Jamali et al. | Robust image watermarking by multi resolution embedding in wavelet transform coefficients | |
| KR101549983B1 (en) | The self-adaptive method for inserting/detecting a forensic mark using particle swarm optimization in the multiple transform domain and the apparatus thereof | |
| He et al. | A bilateral attention based generative adversarial network for DIBR 3D image watermarking | |
| CN109447888A (en) | A kind of watermark handling method of food packaging false-proof digital image | |
| CN106507216B (en) | The method that three-dimensional video-frequency is embedded in the method, apparatus of watermark and extracts watermark | |
| Liang et al. | Ridgelet-based robust and perceptual watermarking for images | |
| Vo et al. | A high-capacity invertible steganography method for stereo image | |
| Kerbiche et al. | A robust video watermarking based on image mosacing and multi-frequential embedding |
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 |
Application publication date: 20110615 |