CN100512440C - Water mark method and device for digital video signal and detecting method and device - Google Patents

Abstract

This invention discloses a method and a device for adding Rodust watermark to digital video signals and a method and a device for testing if the watermark is inserted into the digital video signals, in which, the method includes: dividing a digital video signal into lens sequence and selecting adjacent lens from the sequence to form a video segment, inserting modulated bipolar binary watermark vector Wr, and Wr=(w(0),..., w(n)..., w(N-1)), N is a positive integer greater than 1, which can resist against multiple kinds of time synchronous distortion, space synchronous distortion, luminosity distortion and their combined distortion.

Description

Water mark method, equipment and the detection method, the equipment that are used for digital video signal

Technical field

The invention belongs to the Digital Video Processing field, relate in particular to the method and apparatus that adds watermark in the digital video signal, and detect the method and apparatus that whether has embedded watermark in the digital video signal.

Background technology

Video watermark is the key technology of digital video copyright protection.The most challenging problem of digital video watermarking is exactly how to make watermark in standing the various video application after the conventional Video processing, can preserve, and can correctly be detected, the robustness problem of watermark that Here it is.For video, normal signal is handled and is comprised: time synchronized distortion (as frame insertion, frame deletion, frame-rate conversion etc.), spatial synchronization distortion (as various geometric transformations such as cutting, scaling, translation, rotation, perspective, nonlinear transformation, spherical conversion, local geometric conversion), luminosity distortion (as diminishing compression, noise, filtering, contrast adjustment etc.) etc.

In order to solve the robustness problem of watermark, the researcher has proposed many thoughts and method.The general watermark robustness method commonly used that improves has: spread-spectrum watermark embeds, embeds in the coefficient of known robustness in redundant embedding, the significant coefficient in perception etc.The general thoughts of antagonism time and spatial synchronization distortion then is exhaustive search, synchronization template, auto-correlation, constant watermark etc.These ways of thinking have promoted the research of watermark robustness effectively.But these methods thoroughly do not solve the robustness problem that a certain class or a few class distortions are resisted in watermark.Based on these methods, the researcher studies respectively at time synchronized distortion, spatial synchronization distortion, this three major types distortion of luminosity distortion, proposed some can resist wherein a certain class or the water mark method of a few class distortions.

Along the watermarked comparison effective video water mark method of the opposing spatial synchronization distortion of proposition recently that is considered to of time shaft.Haitsma etc. have proposed by the mean flow rate of each frame of modulation, the water mark method of embed watermark information frame by frame.This method can be resisted polytype spatial synchronization distortion preferably, but the resistance of the time synchronized distortion that causes for interleave, frame losing, frame frequency conversion etc. at random a little less than.Sang etc. have proposed to regulate the water mark method of the variance of each frame and have resisted geometric distortion.But owing to will calculate the discrete Fourier transform (DFT) (DFT) of each frame, so the computation complexity of this method is higher, the ability of anti-simultaneously time synchronized distortion still a little less than.

Niu etc. propose along the watermarked method of time shaft, can resist the time synchronized distortion that the frame losing of certain frequency causes, can resist the spatial synchronization distortion that multiple geometric transformation causes, the luminosity distortion for lossy compression method causes also has resistance preferably.But this method can not be resisted the time synchronized distortion that frame frequency conversion causes.The method of propositions such as Chen can be resisted spatial synchronization distortion and slight random frame is inserted and frame deletion causes time synchronized distortion preferably, still the time synchronized distortion resistance that causes for frame frequency conversion a little less than.

In addition, embedding for resistance time synchronous distortion along time shaft also is a good selection.Lin etc. as frame of video watermark keys maker, can resist the time synchronized distortion that at random frame deletion, frame insertion, frame exchange etc. cause with finite state machine, but this scheme for the resistance of spatial synchronization distortion a little less than.

More than these along methods that time shaft embeds, all be based on frame of video, all can not resist the time synchronized distortion that frame frequency conversion causes.And those methods based on scene or camera lens can be resisted more distortion.The video watermark process based on scene that Jung etc. propose can resist the spatial synchronization distortion that polytype time synchronized distortion and simple geometry conversion cause preferably, but amount of calculation is big, and video quality descends obviously.Propositions such as Sun cut apart the water mark method that combines based on independent component analysis (ICA) and camera lens, can resist polytype time synchronized distortion and common luminosity distortion, as compression etc., the image watermark method that then depends on method and adopted for the resistance of spatial synchronization distortion.

In addition, the part scholar has proposed the method based on 3 d-dem cosine transform (3D DCT), 3 d-dem wavelet transformation (3D DWT).The distortion that these methods cause in the processing that can resist the part type in varying degrees, but all have problems such as complexity height.

Recently the video watermark process that proposes such as Harmanci can be resisted the spatial synchronization distortion that geometric transformation to a certain degree causes, and the time synchronized distortion that causes such as the editor of frame interpolation at random, scene, shearing, exchange.But this method utilize each two field picture of video hash (HASH) information as detecting supplementary, be not the blind Detecting water mark method.

Although the researcher has proposed to comprise the multiple water mark method of above water mark method, can resist a class or a few class distortion to a certain extent, but all resistance time synchronous distortion, spatial synchronization distortion and luminosity distortion simultaneously and the combination distortions thereof effectively of these water mark methods.This makes the practical application of watermark be subjected to considerable restraint.

Summary of the invention

The invention provides a kind of new for digital video signal adds watermark and detects the robust watermarking method whether digital video signal has embedded watermark, in order to solve the problem of the polytype time synchronized distortion of effective opposing, spatial synchronization distortion, luminosity distortion and combination distortion thereof.

A kind of robust watermarking method that is used for digital video signal comprises:

Digital video signal is divided into shot sequence, and from shot sequence, selects adjacent camera lens to form video segment;

In this video segment, embed the bipolarity binary watermarking vector W of modulation _r, W _r=(w (0) ..., w (n) ..., w (N-1)) and, wherein, w (n) is W _rN component, n=0～N-1, N are the positive integer greater than 1;

Wherein, the described step of selecting adjacent camera lens to form video segment from shot sequence further comprises:

Step 1, from the camera lens cut-point selected camera lens cut-point as the video segment starting point, the next camera lens cut-point of video segment starting point as the video segment terminal point;

Step 2, calculating current video fragment length;

If step 3 current video fragment length is less than predetermined value L _Min, then the next camera lens cut-point of current video sheet segment endpoint as the new terminal point of video segment, return step 2, wherein predetermined value L _MinBe positive number; Otherwise, with the video segment of current video fragment as composition.

A kind of method of detecting watermarks that uses the digital video signal of said method comprises:

Step 1, digital video signal to be detected is divided into a shot sequence;

Step 2, from the camera lens cut-point selected camera lens cut-point as video segment starting point and terminal point;

If also have after the terminal point of step 3 current video fragment camera lens not obtain detecting, then the next camera lens cut-point of current video sheet segment endpoint as the new terminal point of video segment; Otherwise, detect and finish;

Step 4, calculating current video fragment length are if video segment length is less than predetermined value D _Min, return step 3, wherein predetermined value D _MinBe positive number, and smaller or equal to predetermined value L _MinOtherwise enter step 5;

Whether step 5, the described video segment of detection have watermark W _r

If step 6 has watermark, then get starting point and the terminal point of the terminal point of current video fragment as the new video fragment, return step 3; Otherwise enter step 7;

If step 7 current video fragment length is less than predetermined value D _Max, return step 3, wherein predetermined value D _MaxMore than or equal to predetermined value L _MinOtherwise enter step 8;

Step 8, the next camera lens cut-point of video segment starting point as new video segment starting point and terminal point, return step 3.

A kind of equipment that is used for watermark information is embedded into digital video signal comprises:

Being used for the watermark information coded modulation is the watermark modulating unit of bipolarity binary watermarking vector, described bipolarity binary watermarking vector W _r=(w (0) ..., w (n) ..., w (N-1)) and, wherein, w (n) is W _rN component, n=0～N-1, N are the positive integer greater than 1;

Be used for digital video signal is divided into the video lens cutting unit of shot sequence;

Be used for choosing cinestrip, form the video segment of suitable video segment and choose the unit from shot sequence;

Be used in video segment, embedding the watermark embedding unit of bipolarity binary watermarking vector;

Wherein, described watermark embeds the unit and further comprises:

Be used for video segment is divided into the video segment cutting unit of a plurality of video units;

Be used for 1 video unit is divided into the video subelement cutting unit of a plurality of video subelements;

Be used to regulate the intermediate frequency energy of video subelement with watermarked vectorial W _rThe subelement intermediate frequency energy adjustment unit of one-component.

A kind of being used to detected the equipment whether digital video signal contains watermark, comprising:

Being used for the watermark information coded modulation is the watermark modulating unit of bipolarity binary watermarking vector, described bipolar binary vector W _r=(w (0) ..., w (n) ..., w (N-1)) and, wherein, w (n) is W _rN component, n=0～N-1, N are the positive integer greater than 1;

Be used for digital video signal is divided into the video lens cutting unit of shot sequence;

Be used for choosing cinestrip, form the video segment of suitable video segment and choose the unit from shot sequence;

Be used to judge whether video segment contains the video segment identifying unit of described bipolarity binary watermarking vector;

Wherein, described video segment identifying unit further comprises:

Be used for video segment is divided into the video segment cutting unit of a plurality of video units;

Be used in 1 video unit, extracting the characteristic vector component extraction unit of the one-component of characteristic vector;

The correlation that is used for calculated characteristics vector sum watermark vector, and judge whether to contain the identifying unit of watermark according to the correlation size.

Video watermark process disclosed in this invention can effectively be resisted polytype time synchronized distortion, spatial synchronization distortion, luminosity distortion and combination distortion thereof, compare with present existent method, on the degree of the kind of the distortion that can resist and distortion, all had significantly to promote.

Description of drawings

Fig. 1 is embedded into watermark for the embodiment of the invention structural representation of the equipment in the vision signal;

Fig. 2 embeds the structural representation of unit for the watermark that is used for video segment among Fig. 1;

Fig. 3 is the structural representation of 1 watermark embedding unit among Fig. 2;

Fig. 4 is the embodiment of the invention detects the equipment of watermark in vision signal a structural representation;

Fig. 5 is used to judge among Fig. 4 whether video segment contains the structural representation of the video segment identifying unit of watermark;

Fig. 6 is the structural representation of characteristic vector component extraction unit among Fig. 5;

Fig. 7 is used for the schematic flow sheet of the watermark embedding method of digital video signal for the present invention;

Fig. 8 is used for the schematic flow sheet of the method for detecting watermarks of digital video signal for the present invention.

Embodiment

Below in conjunction with Figure of description the specific embodiment of the present invention is described.

By research that video lens is cut apart as can be known:

Because shot boundary is the key character of video, content is not being constituted under the situation of obviously influence, even frame losing, the synchronous distortion of frame frequency variation equal time occur, the border of camera lens still can detect.The present invention utilizes camera lens to cut apart according to these characteristics and obtains the watermark embedding and detect required synchronizing information.

As shown in Figure 1, watermark is embedded into equipment in the vision signal, has comprised:

Being used for the watermark information coded modulation is bipolarity binary watermarking vector W _r Watermark modulating unit 100, W wherein _r=(w (0) ..., w (n) ..., w (N-1)) and, w (n) is W _rN component, n=0～N-1, N are the positive integer greater than 1;

Be used for digital video signal is divided into the video lens cutting unit 110 of shot sequence;

Be used for choosing cinestrip, form the video segment of suitable video segment and choose unit 120 from shot sequence;

Be used in video segment, embedding the watermark embedding unit 130 of bipolarity binary watermarking vector;

The memory cell 140 that is used for the video data after the stores processor.

Wherein, be used in video segment, embedding the watermark embedding unit 130 of bipolarity binary watermarking vector, as shown in Figure 2, further comprised:

Be used for video segment is divided into the first video segment cutting unit 131 of a plurality of video units;

Be used for watermarked vectorial W in 1 video unit _r1 watermark of one-component embed unit 132.

Wherein 1 watermark embeds unit 132, as shown in Figure 3, having comprised:

Be used for 1 video unit is divided into the first video subelement cutting unit 1321 of a plurality of video subelements;

Be used to regulate video subelement intermediate frequency energy with watermarked vectorial W _rThe subelement intermediate frequency energy adjustment unit 1322 of one-component.

According to the said equipment, the digital video signal that from digital video signal source, produces, 110 processing obtain shot sequence through the video lens cutting unit.From this sequence, choose unit 120 by video segment and select the video segment that the lens group of partial continuous becomes to be of convenient length.Then this video segment is divided into N video unit through video segment cutting unit 131, and each video unit is divided into R video subelement through video subelement cutting unit 1321.By subelement intermediate frequency energy adjustment unit 1322 the intermediate frequency energy of each video subelement is adjusted, to embed the watermark vector W that obtains according to watermark modulating unit 100 _r

As shown in Figure 4, be watermark to be embedded into the equipment whether the corresponding detection digital video signal of equipment in the vision signal contains watermark with above-mentioned, comprising:

Being used for the watermark information coded modulation is bipolarity binary watermarking vector W _r Watermark modulating unit 100, W wherein _r=(w (0) ..., w (n) ..., w (N-1)) and, w (n) is W _rN component, n=0～N-1, N are the positive integer greater than 1;

Be used for digital video signal is divided into the video lens cutting unit 110 of shot sequence;

Be used for choosing cinestrip, form the video segment of suitable video segment and choose unit 120 from shot sequence;

Be used to judge whether video segment contains the video segment identifying unit 160 of described bipolarity binary watermarking vector.

Wherein the video segment identifying unit 160, as shown in Figure 5, further comprise:

Be used for video segment is divided into the second video segment cutting unit 161 of a plurality of video units;

Be used in 1 video unit, extracting the characteristic vector component extraction unit 162 of the one-component of characteristic vector;

The correlation that is used for calculated characteristics vector sum watermark vector, and judge whether to contain the identifying unit 163 of watermark according to the correlation size.

Characteristic vector component extraction unit 162 wherein as shown in Figure 6, further comprises:

Be used for 1 video unit is divided into the second video subelement cutting unit 1621 of a plurality of video subelements;

The subelement intermediate frequency energy calculation unit 1622 that is used for the intermediate frequency energy of computation subunit;

The characteristic component computing unit 1623 that is used for the one-component of calculated characteristics vector.

According to the said equipment, the digital video signal that from digital video signal source to be detected, produces, 110 processing obtain shot sequence through the camera lens cutting unit.From this sequence, choose unit 120 by video segment and select the video segment that the lens group of partial continuous becomes to be of convenient length, by the second video segment cutting unit 161 video segment is divided into N video unit then, to each video unit, be divided into R subelement by the second video subelement cutting unit 1621, calculate the intermediate frequency energy of each subelement again by subelement intermediate frequency energy calculation unit 1622, by each component that characteristic component computing unit 1623 is calculated characteristic vector according to the intermediate frequency energy meter of each subelement, judge by identifying unit 163 whether this video segment contains watermark then.

In the such scheme, the first video segment cutting unit 131 can be done identical setting with the second video segment cutting unit 161; The first video subelement cutting unit 1321 also can be done identical setting with the second video subelement cutting unit 1621.

Illustrate that below in conjunction with the said equipment the present invention is used for the water mark method of digital video signal.As shown in Figure 7, be the schematic flow sheet of this watermark embedding method, as seen from the figure, mainly may further comprise the steps:

Step 701, be bipolarity binary watermarking vector with the watermark information coded modulation;

This step is modulated to bipolarity binary watermarking vector W with watermark information to be embedded through watermark modulating unit 100 _r=(w (0) ..., w (n) ..., w (N-1)), wherein w (n) ∈ 1,1}, n=0～N-1; N is the positive integer greater than 1.

Step 702, digital video signal is divided into a shot sequence according to the camera lens dividing method;

This step is divided into a series of camera lenses with the digital video signal of watermark to be embedded through video lens cutting unit 110, obtains the camera lens cut-point of digital video signal.

In this step, during the camera lens cutting, can a detector lens shear.The pixel based on the DC image that can adopt Yee and Liu to propose carries out Shot Detection and cutting to comparison method.

Step 703, the video segment of from above-mentioned shot sequence, selecting several adjacent camera lenses compositions to have appropriate length;

This step will be chosen unit 120 through video segment by the shot sequence that step 702 obtains and select the video segment that the lens group of partial continuous becomes to be of convenient length, and can carry out as follows:

Step (11) is selected a cut-point as the video segment starting point from the camera lens cut-point;

Step (12) the next camera lens cut-point of video segment starting point as the video segment terminal point;

Step (13) is calculated the current video fragment length;

Step (14) if video segment length less than L _Min, then the next camera lens cut-point of current video sheet segment endpoint as the new terminal point of video segment, change step (13), otherwise, enter step (15);

Step (15) with the current video fragment as the video segment that is of convenient length.

Step 704, in the video segment watermarked W _r

It is watermarked in the video segment that is obtained by step 703 that this step will embed unit 130 by watermark.Concrete embedding grammar is as follows:

At first the video segment that is obtained by step 703 is divided into N video unit through the first video segment cutting unit 131, each video unit is divided into R video subelement through the first video subelement cutting unit 1321.By subelement intermediate frequency energy adjustment unit 1322 the intermediate frequency energy of each video subelement is adjusted, to embed the watermark vector W that obtains according to watermark modulating unit 100 _r

In the such scheme, the intermediate frequency energy of each video subelement is the average of the intermediate frequency energy of the luminance component of all frames in this video subelement, can adopt following method when adjusting:

If R=3; If M is 8 * 8 a number in video one frame, B (p m) is m piece of p frame, DCT (p, m, i) be the dct transform coefficient of individual 8 * 8 of m of p frame according to i coefficient of zigzag order, i=S wherein ..., T, 0＜S＜T＜32.If E _m(p) be the intermediate frequency energy of video p frame, computational methods are as follows:

$E_m\left(p\right)=\underset{m=0}{\overset{M-1}{\mathrm{\Σ}}}\underset{i=S}{\overset{T}{\mathrm{\Σ}}}\left|\mathrm{DCT}(p,m,i)\right|$

If (n k) is the intermediate frequency energy of k video subelement of n video unit, i.e. the E of each frame of k video subelement of n video unit to g _mAverage:

$g(n,k)=\frac{1}{Q_{n,k}}\underset{p}{\mathrm{\Σ}}{E}_m\left(p\right)$

Wherein, Q _{N, k}Be the number of k the contained frame of video subelement of n video unit, n=0～N-1, k=0～R-1.Order:

$\mathrm{\ϵ}\left(n\right)=g(3n,1)-\frac{g(3n,0)+g(3n,2)}{2}$

When watermarked, adjust the DCT intermediate frequency coefficient of each frame of each subelement of n video unit, make ε (n) away from zero point.And, when w (n)=1, make that the absolute value of the DCT intermediate frequency coefficient of each frame increases in the 1st the video subelement of n video unit, make that the absolute value of the DCT intermediate frequency coefficient of each frame reduces in the 0th, 2 the video subelement of n video unit, ε (n) is changed to the direction greater than 0; When w (n)=-1, way is opposite, makes ε (n) change to the direction less than 0.

The method of adjusting the DCT intermediate frequency coefficient of each frame in the 1st the video subelement of n video unit is as follows:

Set up departments several DCT (p, m, i) adjusted new value be DCT (p, m, i) ＇, its computational methods are as follows:

DCT(p，m，i)＇＝DCT(p，m，i)+Δ(p，m，i)

Wherein Δ (p, m, i) for a change the amount, computational methods are as follows:

Δ(p，m，i)＝λ(i)·w(n)·φ(p，m，i)

$\mathrm{\&phi;}(p,m,i)=\left\{\begin{array}{cc}0,& \left|\mathrm{DCT}(p,m,i)\right|<=T_A\\ 1,& T_A<\mathrm{DCT}(p,m,i)<T(p,m)\\ -1,& -T(p,m)<\mathrm{DCT}(p,m,i)<-T_A\\ \frac{\mathrm{DCT}(p,m,i)}{T(p,m)},& T(p,m)\&le;\left|\mathrm{DCT}(p,m,i)\right|\end{array}\right.$

λ(i)＝max(1，DCT_Freq_Sensitivity(i)×Scale _D)

Wherein DCT_Freq_Sensitivity (i) is i the sensitivity value that DCT frequency sensitive kilsyth basalt is pressed the ZigZag order, Scale _DBe an adjustment factor, max () is for getting peaked function.T _ABe a positive constant, (p m) is two constant T to T _mAnd T _MLinear combination, and T _A≤ T _m≤ T (p, m)≤T _MT (p, m) computational methods are as follows:

T(p，m)＝α(p，m)·T _m+(1-α(p，m))·T _M

Wherein (p m) is combination coefficient to α, and (p, m) (p-1, (p, intermediate frequency energy m) is determined for variable quantity m), piece B with respect to B with respect to variable quantity, the piece B of p-1 by frame p for it.

In order to keep the not sentience of watermark, keep watermark to embed the quality of rear video, we with interval [C (and p, i), C (p, i)] come that (p, m i) carry out restricted intercepting to Δ.Wherein C (p, i) can calculate as follows:

Wherein D (p) is the variable quantity of p frame with respect to the p-1 frame.

When adjusting the DCT intermediate frequency coefficient of each frame in the 0th, 2 the video subelement of n video unit, usefulness-w (n) replaces the w (n) in the above-mentioned steps, can realize by above-mentioned steps.

Corresponding with above-mentioned watermark embedding method, as shown in Figure 8, be the schematic flow sheet of this method of detecting watermarks, as seen from the figure, mainly may further comprise the steps:

Step 801, be bipolarity binary watermarking vector with the watermark information coded modulation;

This step is modulated to bipolarity binary watermarking vector W with watermark information to be embedded through watermark modulating unit 100 _r=(w (0) ..., w (n) ..., w (N-1)), wherein w (n) ∈ 1,1}, n=0～N-1; N is the positive integer greater than 1.

Step 802, digital video signal is divided into a shot sequence according to the camera lens dividing method;

This step is divided into a series of camera lenses with the digital video signal of watermark to be embedded through video lens cutting unit 110, obtains the camera lens cut-point of digital video signal.

During the camera lens cutting, can a detector lens shear.The pixel based on the DC image that can adopt Yee and Liu to propose carries out Shot Detection and cutting to comparison method.

Step 803, from shot sequence, select several adjacent camera lenses to form to have the video segment of appropriate length, and detect this video segment and whether embedded watermark W _r

This step will be chosen unit 120 through video segment by the shot sequence that step 802 obtains and select the video segment that the lens group of partial continuous becomes to be of convenient length, and whether embed watermark by video segment identifying unit 160 these fragments of detection.Can carry out as follows:

Step (21), from the camera lens cut-point selected cut-point as video segment starting point and terminal point;

Step (22) is if also have camera lens not obtain detecting after the terminal point of current video fragment, then the next camera lens cut-point of current video sheet segment endpoint as the video segment terminal point; Otherwise, detect and finish.

Step (23), calculating current video fragment length.If video segment length is less than predetermined value D _Min, change step 22, wherein predetermined value D _MinBe positive number, and smaller or equal to predetermined value L _MinOtherwise enter step (24);

Whether step (24), the described video segment of detection have watermark W _r

Step (25) is then got starting point and the terminal point of the terminal point of current video fragment as the new video fragment if watermark is arranged, and changes step (22); Otherwise enter step (26);

Step (26) if the current video fragment length less than predetermined value D _Max, change step (22), wherein predetermined value D _MaxMore than or equal to predetermined value L _MinOtherwise enter step (27);

Step (27), the next camera lens cut-point of video segment starting point as new video segment starting point and terminal point.Change step (22).

Wherein, step (24) is finished by video segment identifying unit 160, concrete grammar is as follows: at first the video segment that is obtained by step (23) is divided into N video unit by the second video segment cutting unit 161 with video segment, to each video unit, be divided into R subelement by the second video subelement cutting unit 1621, calculate the intermediate frequency energy of each subelement again by subelement intermediate frequency energy calculation unit 1622, by each component that characteristic component computing unit 1623 is calculated characteristic vector according to the intermediate frequency energy meter of each subelement, judge by identifying unit 163 whether this video segment contains watermark then.

The method of each component of characteristic component computing unit 1623 calculated characteristics vectors is as follows:

Characteristic vector V=(v (0), v (1) ... v (n) ..., v (N-1)) the computational methods of component v (n) as follows:

$v\left(n\right)=\left\{\begin{array}{cc}1,& \mathrm{\&epsiv;}\left(n\right)>{\mathrm{\&epsiv;}}_0\\ -1,& \mathrm{\&epsiv;}\left(n\right)<-{\mathrm{\&epsiv;}}_0\\ 0,& -{\mathrm{\&epsiv;}}_0\&le;\mathrm{\&epsiv;}\left(n\right)\&le;{\mathrm{\&epsiv;}}_0\end{array}\right.$

Wherein, ε ₀It is a positive constant.

The method whether identifying unit 163 judgement video segments contain watermark is as follows:

At first calculated characteristics vector V and watermark vector W _rNormalization correlation τ.

$\mathrm{\&tau;}=\frac{V\&CenterDot;W_r}{\left|V\right|\left|W_r\right|}$

If τ _NcBe detection threshold.If τ 〉=τ _Nc, show that then detected this section video has watermark; Otherwise detected this section video does not have watermark.

Test shows, the spatial synchronization distortion that video watermark process disclosed in this invention causes for multiple common time synchronized distortion, for multiple geometric transformation and combination, the luminosity distortion of 4 kinds of main types, it is not that the detection accuracy rate of very violent video has all reached 100% that picture is changed.Change violent video for picture,, detect accuracy rate and also nearly all reached 100% as war film etc.This illustrates that this method can effectively resist polytype time synchronized distortion, spatial synchronization distortion and luminosity distortion, and this is that additive method is difficult to realize.

When two classes or the combination of the distortion more than two classes generation, video watermark process disclosed in this invention still can reach 100% detection accuracy rate for general vision signal, for changing violent vision signal, detects accuracy rate and has also reached more than 60%.This is that additive method is difficult to accomplish.

The present invention compares with present existing method, on the degree of the kind of the distortion that can resist and distortion Had significantly and promoted.

Obviously, those skilled in the art can carry out various changes and modification to the present invention and not break away from this Bright spirit and scope. Like this, if belonging to claim of the present invention, these modifications of the present invention and modification reach The range of its equivalent technologies, then the present invention also be intended to comprise these change and modification interior.

Claims (12)

1, a kind of water mark method that is used for digital video signal comprises:

Digital video signal is divided into shot sequence, and from shot sequence, selects adjacent camera lens to form video segment;

In this video segment, embed the bipolarity binary watermarking vector W of modulation _r, W _r=(w (0) ..., w (n) ..., w (N-1)) and, wherein, w (n) is W _rN component, n=0～N-1, N are the positive integer greater than 1;

Wherein, the described step of selecting adjacent camera lens to form video segment from shot sequence further comprises:

Step 1, from the camera lens cut-point selected camera lens cut-point as the video segment starting point, the next camera lens cut-point of video segment starting point as the video segment terminal point;

Step 2, calculating current video fragment length;

If step 3 current video fragment length is less than predetermined value L _Min, then the next camera lens cut-point of current video sheet segment endpoint as the new terminal point of video segment, return step 2, wherein predetermined value L _MinBe positive number; Otherwise, with the video segment of current video fragment as composition.

2, the method for claim 1 is characterized in that, watermarked W in video segment _rStep further comprise:

This video segment is divided into N video unit;

In n video unit, embed W _rN component w (n), n=0～N-1.

3, method as claimed in claim 2 is characterized in that, when described video segment is divided into N video unit, along time shaft, cuts apart by the duration equipartition principle.

4, method as claimed in claim 3 is characterized in that, and is described along time shaft, when cutting apart by the duration equipartition principle, may further comprise the steps:

Calculate the duration of this video segment by absolute time;

According to this duration, this video segment is divided into N video unit along time shaft;

According to each video unit starting and ending moment on time shaft, and the starting and ending of each frame of video constantly, each frame of video all is grouped in certain video unit, make in each video unit that contained video frame number is an integer, the longest video unit and the contained video frame number of the shortest video unit differ and are no more than 1 frame.

5, method as claimed in claim 2 is characterized in that, embeds W in n video unit _rN component w (n) time, adopt following steps:

This video unit is divided into R video subelement, and R is the positive integer greater than 1;

Regulate the intermediate frequency energy of each video subelement, according to the intermediate frequency energy relationship embedding w (n) of each video subelement.

6, method as claimed in claim 5 is characterized in that, the intermediate frequency energy of the luminance component of described video subelement is the average of the intermediate frequency energy of the luminance component of all frames in this video subelement.

7, the method for claim 1 is characterized in that, when digital video signal is divided into shot sequence, according to the camera lens dividing method video frequency program is divided into a shot sequence, and the only shear of detector lens.

8, a kind of method of detecting watermarks that uses the digital video signal of the described method of claim 1 comprises:

Step 1, digital video signal to be detected is divided into a shot sequence;

Step 2, from the camera lens cut-point selected camera lens cut-point as video segment starting point and terminal point;

If also have after the terminal point of step 3 current video fragment camera lens not obtain detecting, then the next camera lens cut-point of current video sheet segment endpoint as the new terminal point of video segment; Otherwise, detect and finish;

Step 4, calculating current video fragment length are if video segment length is less than predetermined value D _Min, return step 3, wherein predetermined value D _MinBe positive number, and smaller or equal to predetermined value L _MinOtherwise enter step 5;

Whether step 5, the described video segment of detection have watermark W _r

If step 6 has watermark, then get starting point and the terminal point of the terminal point of current video fragment as the new video fragment, return step 3; Otherwise enter step 7;

If step 7 current video fragment length is less than predetermined value D _Max, return step 3, wherein predetermined value D _MaxMore than or equal to predetermined value L _MinOtherwise enter step 8;

Step 8, the next camera lens cut-point of video segment starting point as new video segment starting point and terminal point, return step 3.

9, method as claimed in claim 8 is characterized in that, detects in this video segment whether embedded watermark W in the step 5 _rComprise:

Described video segment is divided into N video unit;

From n video unit, extract characteristic vector V=(v (0), v (1) ... v (n) ..., v (N-1)) n component v (n), n=0～N-1;

The correlation of calculated characteristics vector sum watermark vector, and judge according to correlation size and detection threshold whether video content contains video watermark.

10, method as claimed in claim 9 is characterized in that, during described n the component v (n) that extracts characteristic vector V from n video unit, comprising:

Described video unit is divided into R video subelement;

Calculate the intermediate frequency energy of each video subelement, utilize the relation of the intermediate frequency energy of each video subelement to extract v (n).

11, a kind of equipment that is used for watermark information is embedded into digital video signal is characterized in that, comprising:

Being used for the watermark information coded modulation is the watermark modulating unit of bipolarity binary watermarking vector, described bipolarity binary watermarking vector W _r=(w (0) ..., w (n) ..., w (N-1)) and, wherein, w (n) is W _rN component, n=0～N-1, N are the positive integer greater than 1;

Be used for digital video signal is divided into the video lens cutting unit of shot sequence;

Be used for choosing cinestrip, form the video segment of suitable video segment and choose the unit from shot sequence;

Be used in video segment, embedding the watermark embedding unit of bipolarity binary watermarking vector;

Wherein, described watermark embeds the unit and further comprises:

Be used for video segment is divided into the video segment cutting unit of a plurality of video units;

Be used for 1 video unit is divided into the video subelement cutting unit of a plurality of video subelements;

Be used to regulate the intermediate frequency energy of video subelement with watermarked vectorial W _rThe subelement intermediate frequency energy adjustment unit of one-component.

12, a kind of being used to detected the equipment whether digital video signal contains watermark, it is characterized in that, comprising:

Being used for the watermark information coded modulation is the watermark modulating unit of bipolarity binary watermarking vector, described bipolarity binary watermarking vector W _r=(w (0) ..., w (n) ..., w (N-1)) and, wherein, w (n) is W _rN component, n=0～N-1, N are the positive integer greater than 1;

Be used for digital video signal is divided into the video lens cutting unit of shot sequence;

Be used for choosing cinestrip, form the video segment of suitable video segment and choose the unit from shot sequence;

Be used to judge whether video segment contains the video segment identifying unit of described bipolarity binary watermarking vector;

Wherein, described video segment identifying unit further comprises:

Be used for video segment is divided into the video segment cutting unit of a plurality of video units;

Be used in 1 video unit, extracting the characteristic vector component extraction unit of the one-component of characteristic vector;

The correlation that is used for calculated characteristics vector sum watermark vector, and judge whether to contain the identifying unit of watermark according to the correlation size.

Images