CN102892048B - Video watermark anti-counterfeiting method capable of resisting geometric attacks - Google Patents

Abstract

The invention belongs to the fields of computer vision and image understanding, and relates to a video watermark anti-counterfeiting method capable of resisting geometric attacks, which comprises the steps of: first, extracting a maximally stable extrema region (MSER) of a video frame; second, screening fitted elliptic regions, eliminating regional overlapping, and selecting an elliptic region for final watermark embedding; and third, by modifying a horizontal high frequency vertical low-frequency band (HL) and a horizontal low frequency vertical low-frequency band (LH) of a wavelet transform coefficient, implementing the watermark embedding on the selected characteristic region. The ability of resisting affine geometric attacks is enhanced. Furthermore, the invention discloses an inverse process of the watermark embedding, namely a method of watermark extracting. According to the method, the modification on the original video is effectively reduced; and an excellent vision effect is guaranteed even under malicious geometric attacks.

Description

A kind of video watermark method for anti-counterfeit of resist geometric attacks

Art

The invention belongs to computer vision and image understanding field, relate to a kind of video watermark method for anti-counterfeit of resist geometric attacks.

Background technology

In recent years, more senior and more easy-operating video editing/distort/attack tool is popularized just gradually, causes the problem such as news, the evidence video for judiciary distorted of malice of forging to emerge in an endless stream; On the other hand, become Information Communication main carriers with forum, blog, blog, microblogging, social networks for the Society information net that representative is formed, its distinguishing feature is exactly that Information Communication is rapider.Both combinations, challenge the safety of video content worse.Video watermark is one of effective technology means of protection video content safety and authenticity.Video watermark refer to by watermark embedment to video sequence to prevent video from suffering fraudulent copying and amendment.For copyright protection, video watermark system is intended to resist various malice and non-malicious attack.Although the research and development of video watermark technology for some time, resist geometric attacks ability remains the difficult point of video watermark technology.

Early stage video watermark process by the pixel of amendment video single frames by watermark embedment to spatial domain.These class methods to common geometric attack and image procossing very sensitive.Watermark subsequently starts to be embedded into frequency domain to strengthen the robustness of watermaking system to image procossing.These class methods are come watermarked by the conversion coefficient revising frame sequence usually, and common transform method has discrete cosine transform, discrete Fourier transform, wavelet transform etc.But it is still responsive to Geometrical attack based on the method for frequency domain.

The video watermark embedding grammar of existing resist geometric attacks roughly can be divided into three major types: 1. geometry not political reform; 2. geometry restoring method; 3. the watermark method of feature based.In first kind method, watermark carrier is geometry field of invariants.In order to reach the object of anti-translation, rotation, change of scale, general stochastic transformation territory is used to watermarked and can keeps under geometric transformation synchronous [1].Certainly, also have other geometry fields of invariants [2] many to be suggested, as: logarithm-polar domain, common square, Zernike square.Although these methods can resist rotation to a certain extent and yardstick is attacked, watermark is subject to local geometric conversion and damages.Equations of The Second Kind method is by estimating that the parameter of geometric transformation is come watermarked, and these parameters are used to carry out inverse transformation to picture before watermark extraction.Typical geometric transformation parameter can use template to estimate [3] [4].But the major defect based on template method is template to be easy under attack and remove, these fatal errors make the method recovered based on geometry and impracticable.The 3rd class digital watermark being main feature with " localized watermark of feature based " is first comparatively early proposed by [5] [6] [7] [8] such as Kutter.The main points of view of these class methods is impacts that picture feature is not subject to geometric attack, watermark information and picture feature is associated the integrality that can ensure watermark information.The watermark embedding technique of feature based utilize picture feature can duplicate detection to determine the position of watermark embedment and extraction.This kind of technology is extract minutiae from picture first, and then distinguished point based structure local characteristic region is watermarked.

The people such as Bas [9] propose content-based synchronization watermarking mechanism.They use Harris detector to extract picture feature point and build Delaunay segmentation, in each spatial domain trigonum, then use classical additivity mechanism to come watermarked.This mechanism detects characteristic point and reconstructs original Delaunay segmentation to detect watermark after being attacked.But Harris point is very sensitive to dimensional variation, be in this way subject to the impact that yardstick is attacked.

There is a kind of local invariant feature crying Scale invariant features transform (SIFT) [10], because it is widely used in content-based digital watermark [11] to the consistency of rotation, yardstick.In [12], SIFT feature is used to generate the circular piece as embedded unit.Watermark embedment is in the spatial domain of these unit, and this mechanism can resist more generally geometric attack.Harris-Laplace detector is also used to the border circular areas of structure based on Harris angle point with watermarked.But SIFT detector and Harris-Laplace detector [13] are subject to the impact of length-width ratio distortion.A weakness of above-mentioned content-based watermark embedment mechanism is that embedding region is very responsive to asymmetric change of scale.In order to address this problem, affine invariant point detector has been put forward further.The affine area detector of Harris is used for extracting affine covariant region by people such as Hefei Ling in [14].

Maximum stable extremal region (MSER) possesses rotation, yardstick, Aspect Ratio change, illumination variation consistency.In this section of patent, we have proposed the healthy and strong mechanism of the video watermark based on MSER local feature.We select MSER be because: the first, under absolutely mostly count situation, as visual angle change, dimensional variation, illumination variation etc., it is more stable that MSER compares further feature territory; The second, MSER ultrahigh in efficiency, is close to Pixel-level linear complexity in leaching process.In addition, MSER has two good characteristics: the first, is closed under the MSER of acquisition is integrated into Picture Coordinate continuous transformation; The second, the MSER of acquisition be integrated into the monotonic transformation of picture intensity under be closed.One efficient (near-linear complexity) and the fast algorithm of detecting of practicality are at Matas[15] be suggested in original text.

Multi-resolution characteristics in view of wavelet field can provide spatial domain and frequency domain feature to make it compatible with human visual system (HVS), and we concentrate on wavelet transform (DWT) and add watermark to wavelet field HL and LH band instead of spatial domain.

Leading reference

[1] Dimitrios Simitopoulos, Dimitrios E.Koutsonanos, Michael Gerassimos Strintzis " the picture watermark based on general stochastic transformation " image procossing 2003.

[2] Picard Justin, Zhao Jian video watermark: the U.S., US2009220070A1 [P] 2009-09-03.

[3] Shelby Pereira, Thierry Pun " the anti-affine transformation picture watermark based on template matches " image procossing 2000.

[4] Macy William W, Holliman Matthew J watermark decode synchronization template: the U.S., US6707926B1 [P] 2004-03-16.

[5] M.Kutter, S.K.Bhattacharjee, T.Ebrahimi " towards second generation watermark mechanism " image procossing international conference 1999.

[6] Chih-Wei Tang, Hsueh-Ming Hang " a kind of digital picture watermark mechanism of feature based " signal transacting 2003.

[7] Xiang-yang Wang, Li-min Hou, Jun Wu " a kind of digital picture watermark of resist geometric attacks of feature based " picture and vision calculate 2008.

[8] N.R.Nantha Priya, S.Lenty Stuwart " the picture watermarking process of feature based " computer application International Periodicals 2010.

[9] Patrick Bas, Jean-Marc Chassery, and macq " the constant watermark of geometry of use characteristic point " image procossing 2002.

[10] David G.Lowe " object identification based on local scale invariant features " computer vision 1999.

[11] Lee Hae Yeoun, Lee Heung Kyu uses the picture watermark of Scale invariant features transform: Korea S, KR20070073332A [P] 2007-07-10.

[12] Viet Quoc PHAM, Takashi MIYAKI, Toshihiko YAMASAKI, Kiyoharu AIZAWA the mechanism of the watermark based on the constant object of geometry by SIFT feature " image procossing international conference 2007.

[13] Krystian Mikolajczyk, Cordelia Schmid " yardstick and affine constant point of interest detector " computer vision International Periodicals 2004.

[14] Hefei Ling, Liyun Wang, Fuhao Zou, Zhengding Lu, Ping Li " the compressed domain video watermark based on affine field of invariants " signal transacting 2006.

[15] J Matas, O Chum, M Urban, T P ajdla " the wide Baseline Stereo based on maximum stable extremal region " image and vision calculate 2004.

[16] Chun-Shien Lu, Hong-Yuan Mark Liao " watermark based on VS: the mechanism of a kind of anti-rotation and upset " image procossing international conference 2001.

Summary of the invention

The object of the invention is the above-mentioned deficiency overcoming prior art, propose a kind ofly to show superior video watermark method for anti-counterfeit in anti-affine geometry attack.Technical scheme of the present invention is as follows:

A video watermark method for anti-counterfeit for resist geometric attacks, comprises the following steps:

The first step: the maximum stable extremal region MSER extracting frame of video:

(1) use BINSORT sort algorithm to be according to pixels worth size to pixel in frame of video to arrange, and set up location index to the pixel after sequence, each pixel after sequence is considered as being communicated with assembly;

(2) use associating-lookup algorithm to carry out merging growth to connection assembly, record the area of each connection assembly;

(3) MSER is selected.Under assign thresholds, from the connection assembly that (2) obtain, pick out assembly area change obtains local minimum connection assembly with changes of threshold, obtain maximum stable extremal region.

(4) for each maximum stable extremal region, computation of mean values u and covariance ∑, recycling elliptic region equation (x-u) ^t∑ ^-1(x-u)=1, becomes elliptic region according to actual area form fit;

Second step: the elliptic region simulated is screened, eliminate region overlapping, choose the elliptic region for final watermark embedment:

(1) the excessive too small elliptic region of filter area, the elliptic region area of reservation meets τ ₁< | R _i| < τ ₂, wherein, | R _i| represent the area of i-th elliptic region, τ ₁and τ ₂represent lower bound and the upper bound of size respectively.

(2) design of graphics model, in graph model, the corresponding summit V of each elliptic region i _i; Summit is to V _i, V _jbetween by limit E _ijconnect; Limit E _ijweight w _ijequal summit to V _i, V _jbetween distance, (V, E) forms Undirected graph G;

(3) minimum spanning tree is used at Undirected graph G) algorithm obtains minimum spanning tree T;

(4) judge that in T, whether often pair of characteristic area is overlapping according to elliptic equation, if overlapping, then these two elliptic regions put identical cluster under, otherwise, put different clusters under;

(5) pick out the elliptic region that in each cluster, area is maximum, these elliptic regions will be used for final watermark embedment;

3rd step: by revising the vertical low-frequency band of horizontal high-frequent (HL) and the vertical high frequency band of horizontal low frequencies (LH) of wavelet conversion coefficient, carry out watermark embedment to the characteristic area selected, method is as follows:

(1) elliptic region normalization: first, to the inverse ∑ of the covariance of the elliptic region picked out ^-1carry out singular value decomposition, ∑ ^-1=Q Λ Q ^-1, wherein, Q represents real Orthogonal Symmetric unitary matrice, Q ^-1represent that Q's is inverse, Λ represents diagonal matrix.Then, z=Q is made ^t(x-μ) obtains z ^tΛ z=1, wherein, ^tthe transposition of representing matrix, z represents the circle coordinates after elliptic coordinates x normalization, finds transformation equation z=Q ^t(x-μ), normalizes to circle coordinates z by elliptic coordinates x;

(2) zero padding outside border circular areas after normalization, is filled to external square area;

(3) utilize Algorithms of Discrete Wavelet Transform to carry out rank transformation again to the advanced every trade conversion of obtained external square area, obtain wavelet conversion coefficient matrix;

(4) adaptive wavelet coefficient modifying, method is as follows: the vertical low-frequency band LH (x of horizontal high-frequent of note wavelet conversion coefficient matrix, y) with the vertical high frequency band HL of horizontal low frequencies (x, y) difference is D (x, y), if robustness threshold value is α, when watermark bit w (x, y) is embedded into, coefficient differentials must meet lower relation: D (x, y) >=α, if w (x, y)=1, D (x, y) <-a, if w (x, y)=0;

(5) reverse wavelet transform is carried out to amended matrix of wavelet coefficients, obtain amended square features region;

(6) get inscribed circle and oppositely normalization at square features intra-zone, obtain watermarked after oval feature region;

(7) characteristic area replace: remember watermarked after characteristic area R ₂with primitive character region R ₁difference be R _d, by R _dbe added in original video frame.

Watermark extraction step can be:

(1) to the frame of video being embedded with watermark, extract according to the method for the first step maximum stable extremal region MSER being embedded with the frame of video of watermark;

(2) carry out elliptic region screening according to the method for second step, obtain the elliptic region that may comprise watermark;

(3) to the elliptic region that may comprise watermark, (1), (2) and (3) of the 3rd step is performed successively;

(4) watermark bit blind Detecting: note w ' (x, y) detects watermark bit at (x, y), then the mode of watermark bit blind Detecting is as follows:

$w^{\&prime;}(x,y)=\left\{\begin{array}{cc}1,\mathrm{if}& D^{\&prime;}(x,y)\&GreaterEqual;0\\ 0,\mathrm{if}& D^{\&prime;}(x,y)<0\end{array}\right.,$ Wherein, D ' (x, y)=HL ' (x, y)-LH ' (x, y), HL ' (x, y) and LH ' (x, y) represents the wavelet coefficient at position (x, y) place;

(5) recovery rate is calculated, if the rate of getting is less than certain threshold tau _nc, be considered as in this region not with watermarked information, in the present invention, get τ _nc=0.68;

(6) final watermark w ^*(x, y) ballot obtains according to following ballot formula:

$w^{*}(x,y)=\left\{\begin{array}{cc}1,\mathrm{if}& {\mathrm{Num}}_1(x,y)\&GreaterEqual;{\mathrm{Num}}_0(x,y)\\ 0,\mathrm{if}& {\mathrm{Num}}_0(x,y)<{\mathrm{Num}}_1(x,y)\end{array}\right.$

Wherein, w ^*(x, y) represents the watermark bit voting results at position (x, y) place, Num ₁(x, y) and Num ₀(x, y) is illustrated respectively in the number of " 1 " that position (x, y) detects and the number of " 0 ".

It can be 0.09 that above-mentioned robustness puts threshold alpha.

The present invention has following technique effect:

1. the present invention is based on MSER Affinely invariant region and carry out watermark embedment, synchronization watermarking embeds and leaching process, enhances anti-affine geometry attacking ability.

2. the present invention is by revising wavelet coefficient to carry out watermark embedment at the vertical lower frequency region of the horizontal high-frequent of DWT (HL) and the vertical high-frequency domain of horizontal low frequencies (LH), effectively can reduce the amendment to original video, even if suffering also can keep fabulous visual effect in malice geometric attack situation.

3. the algorithm near-linear complexity that realizes of the present invention, its high efficiency and Real-time ensuring technology applying in actual production life.

Accompanying drawing explanation

The watermark embedment flow chart that Fig. 1 the present invention proposes.

R*r square region after Fig. 2 normalization.

Square region after Fig. 3 fills.

The MSER region that Fig. 4 is original.

The ellipse that Fig. 5 simulates.

MSER region after Fig. 6 area filters.

The region minimum spanning tree that Fig. 7 constructs.

Fig. 8 MSER region clustering.

The final selection result of Fig. 9.

Video sequence signal to noise ratio curve chart after Figure 10 is watermarked.

Figure 11 dimensional variation attacks lower robust analysis curve chart.

Robust analysis curve chart under Figure 12 rotation attack.

Embodiment

Below in conjunction with drawings and Examples, the present invention is described in detail.

The present invention is described in detail below:

Step one: generating pictures local affine invariant covariant region

MSER is detected by the layering set of analysis connection, weighted graph, mainly carries out in three steps: arrangement image pixel intensities; Merge and be communicated with assembly; Detect extremal region and confirm maximum stable extremal region.Concrete steps are as follows:

1. detect maximum stable extremal region.First BINSORT algorithm is used according to pixels to be worth arrangement picture pixel.After sequence, in picture, all pixels mark, and are communicated with assembly and are undertaken growing and merging by associating-lookup algorithm, and save the area of each connection assembly.In extremal region, maximum stable region refers to that those are under assign thresholds, the change of device region relative area obtains the device region of local minimum as the function that relative threshold changes, also namely, MSER refers to and can keep stable picture region in larger threshold range in picture binarization.

2. fitted ellipse region.Original MSER has irregularly shaped and very difficult description, so must build region shape descriptor.Usual extremal region can become oval according to actual area form fit.First average u and the covariance ∑ of each maximum stable extremal region R is calculated:

$u=\frac{1}{\left|R\right|}\underset{x\&Element;R}{\mathrm{\&Sigma;}}x$

$\mathrm{\&Sigma;}=\frac{1}{\left|R\right|}\underset{x\&Element;R}{\mathrm{\&Sigma;}}(x-u){(x-u)}^T.$

Wherein x=(x, y) represents the position of pixel, and R represents area pixel point location sets, | R| represents the number of pixel in MSER.Oval local affine invariant covariant characteristic area is by formula (x-u) ^t∑ ^-1(x-u)=1 determines.

Step 2: maximum stable extremal region filters.

The more overlap of the usual quantity in the region that MSER detector extracts covers whole picture (see accompanying drawing 4).After fit procedure, many elliptic regions have overlap (see accompanying drawing 5).This makes our watermark embedding method repeatedly watermarked in overlapping region, causes algorithm accurately cannot extract watermark.In order to eliminate this ambiguity, overlapping region needs to filter.

Based on MSER provincial characteristics, first the excessive too small elliptic region of area is filtered.As pre-treatment step, the elliptic region area that we retain meets τ ₁< | R _i| < τ ₂, wherein | R _i| represent the area of i-th MSER elliptic region, τ ₁and τ ₂represent lower bound and the upper bound of area respectively.τ ₁and τ ₂usually obtained by experiment experience, we get τ in the present invention ₁=2000, τ ₂=5000.(see accompanying drawing 6)

In order to filter all overlapping regions, our design of graphics category of model MSER region.In graph model, each elliptic region is regarded as a summit V _i; Be connected to each other between every opposite vertexes, (V, E) forms Undirected graph G.Every bar limit E _jweight w _jequal the distance between corresponding vertex.Our target generates unconnected graph to Undirected graph cutting.Each connected subgraph represents a cluster, and each summit only belongs to a specific cluster.In order to obtain stable zero lap MSER region, we remove weight limit based on minimum spanning tree (MST) algorithm.Minimum spanning tree is the spanning tree (see accompanying drawing 7) with minimal weight summation.MST algorithmic tendency is in deletion weight larger skirt, and two MSER regions apart from each other like this can be divided in different clusters.By arranging threshold tau, we delete the larger limit of weight further, make every bar weight limit meet w _i< w _τ.Therefore after further filtering, we obtain a dark woods of spanning tree or without connected graph (see accompanying drawing 8).Like this, we obtain the cluster set of a quantification, and each set comprises multiple MSER region.Finally, (see accompanying drawing 9) is selected as final in the MSER region that Retention area is maximum in each cluster.

MSER filter algorithm divides four steps to carry out:

1. filter out the MSER region that area is excessive or too small;

2. on the complete graph built, apply MST algorithm generate minimum spanning tree;

3. the cluster obtaining quantification is classified in the MSER set of couple candidate;

4. retain the MSER region that in each cluster, area is maximum as final selection.

Step 3: watermark mechanism.

1. watermark carrier channel selecting.

YUV color space, can the Human Perception effect of efficient coding colour picture or video typically for coloured image pipeline.Y represents brightness component, U and V is color component.The oversampling ratio of U and V is lower than Y, and this is called usually " color down-sampling ", and object is to improve compression efficiency.In order to keep chromaticity, we only add watermark at Y passage.

2. rotation-Scale invariant normalization.

In order to obtain yardstick and affine-invariant features, the MSER elliptic region of each selection needs to be normalized into circular block.The oval Cheng Yuan of normalization is equivalent to searching transformation equation.First, we describe the ellipse of matching by average μ and covariance ∑, meet (x-μ) ^t∑ ^-1(x-μ)=, wherein, x=(x, y).In order to normalization is oval, ∑ is by singular value decomposition:

∑ ^-1＝QΛQ ^-1

Wherein, Q represents real Orthogonal Symmetric unitary matrice, and Λ represents diagonal matrix.

Then, z=Q is made ^t(x-μ) obtains z ^tΛ z=1, wherein z represents the coordinate obtained from x normalization.Find transformation equation z=Q ^tafter (x-μ), normalization space can calculate thus and derive.

In order to obtain rotation attack consistency, we need to compose a fixed-direction to each normalization region.To each normalization region compute gradient build direction histogram in circular window.Histogrammic peak point is exactly the principal direction of characteristic area.

3. frequency domain.

The present invention brings based on discrete wavelet transformer to add watermark and be with to HL and LH of wavelet field.The main advantage that wavelet transform compares Fourier transform is that it can capture frequency and spatial positional information simultaneously.In the present invention, DWT algorithm is applied on normalized circular block by the mode of first line translation rank transformation again.Most concentration of energy of wavelet transformation are at LL subband and only account for the overall small part of wavelet coefficient.Watermark embedment is had two to the reason of small echo HL and LH subband:

1) LL subband comprises the most energy in signal.Therefore, when there is strenuous exercise in frame of video, the watermark of embedding cannot correctly be extracted after under attack.

2) HH subband comprises the detailed information of video, in the watermarked easy destruction video visual effect of this subband.

4. watermark embedment.

Watermark embed process is from decoded video obtains original frame sequence.For each frame, we utilize MSER detector and filter algorithm extract and select MSER characteristic area.Watermark by repeated embed in these MSER regions.Detailed watermark embed process is as follows:

1) MSER detector is used to extract local affine invariant covariant region from every frame.An independent frame comprises many MSER regions and overlaps each other.Overlapping region uses and eliminates based on the characteristic filter algorithm of minimum spanning tree.

2) in order to obtain affine and scale invariability, we become unit border circular areas at each elliptic region to be embedded of normalization.Rotational invariance is obtained to the compute gradient direction, region after each normalization and principal direction one.In order to obtain the square region that can input as DWT, zero padding operation is carried out on the border of our team's border circular areas.

3) each square region obtains DWT coefficient after wavelet transform.Watermark embedment is in HL and LH subband.In the present invention, Wavelet Coefficient Blocks is of a size of 2r*2r, so watermark is scaled to r*r to keep onesize with LH or HL subband.In experiment, we get r and equal 32.After watermark adjusted size is complete, carry out watermark embedment by amendment HL and LH sub-band coefficients.

4) last, based on wavelet transform, we by watermark adaptive feed-forward network in wavelet coefficient.Here, we remember that the difference of LH (x, y) and HL (x, y) is:

D(x，y)＝HL(x，y)-LH(x，y)

Select α as robustness threshold value, when watermark bit w (x, y) is embedded into, coefficient differentials must meet following relation:

D(x，y)＞＝α，if w(x，y)＝1，D(x，y)＜-a，if w(x，y)＝0

Larger threshold alpha can improve the robustness of algorithm but also reduce the perceived effect of video simultaneously.Balance robustness and perceived effect, we arrange threshold alpha is 0.09.

5. watermark extracting.

Similar watermark embedment, the first step of watermark extracting analyzes every content frame to extract local affine invariant covariant region.Then, use characteristic filter algorithm selects Non-overlapping Domain.Last watermark obtains from these regions.In order to obtain affine and scale invariability, it is necessary for being normalized each elliptic region being embedded with watermark, and the region after each normalization is filled to the square (see accompanying drawing 2,3) of 2r*2r.

We are from LH and the HL subband extraction watermark information of wavelet coefficient.The mode of watermark bit blind Detecting is as follows:

$w^{\&prime;}(x,y)=\left\{\begin{array}{c}1,\mathrm{if}{D}^{\&prime;}(x,y)\&GreaterEqual;0\\ 0,\mathrm{if}{D}^{\&prime;}(x,y)<0\end{array}\right.$

Wherein, D ' (x, y)=HL ' (x, y)-LH ' (x, y), HL ' (x, y) and LH ' (x, y) represents the wavelet coefficient at position (x, y) place.

After detecting watermark w ', we calculate the normalization relating value z between w and w ' _nc:

$z_{\mathrm{nc}}=\frac{(w\&CenterDot;w^{\&prime;})}{\left|w\right|}$

Wherein, | w| represents the length of w.If z _ncbe less than threshold tau _nc, we think not with watermarked information in this region.

In fact, we embedded in watermark in multiple invariant region, even if make so also can watermark be detected when video deforms.Final watermark result is by choosing in a vote.Note Num ₁(x, y) and Num ₀(x, y) is illustrated respectively in the number of " 1 " that position (x, y) detects and the number of " 0 ".Then, watermark result is determined as follows:

$w^{*}(x,y)=\left\{\begin{array}{c}1,\mathrm{if}{\mathrm{Num}}_1(x,y)\&GreaterEqual;{\mathrm{Num}}_0(x,y)\\ 0,\mathrm{if}{\mathrm{Num}}_0(x,y)<{\mathrm{Num}}_1(x,y)\end{array}\right.$

The present invention tests the Foreman video the most frequently used in watermark field, the visual quality of its principal measure watermark video and the robustness of watermark extracting under various attack.Y-PSNR (PSNR) is used to the visual quality measuring frame of video, and recovery rate (Normalized correlationvalue) is used to the robustness of measuring watermark mechanism.

Figure 10 is the visual effect that Y-PSNR (PSNR) is used for weighing video.From the experimental results, there is not vision difference in watermark video and original video.Figure 10 illustrates the Y-PSNR value of watermark video frame, and the average peak signal to noise ratio of watermark video frame is 43.763dB.

Figure 11 attacks for yardstick, and watermark video carries out dimensional variation between 0.6 to 2.0, and Experimental results show in fig. 11, can see that the present invention has robustness to yardstick attack.

Figure 12 rotation attack rotates watermark video frame between 0 degree to 180 degree, and as seen from Figure 12 when the anglec of rotation is lower than 100 degree, watermark extracting rate is greater than 90%, and along with the increase of the anglec of rotation, recovery rate does not decline rapidly.

Table 1 is contrast situation with the recovery rate of list of references [16] under method for anti-counterfeit multiclass of the present invention is attacked, and can find out, under various attack type, the present invention has surmounted [16].

Table 1 multiclass contrasts with the recovery rate of list of references [16] under attacking.

Claims (3)

1. a video watermark method for anti-counterfeit for resist geometric attacks, comprises the following steps:

The first step: the maximum stable extremal region MSER extracting frame of video:

(1) use BINSORT sort algorithm to be according to pixels worth size to pixel in frame of video to arrange, and set up location index to the pixel after sequence, each pixel after sequence is considered as being communicated with assembly;

(2) use associating-lookup algorithm to carry out merging growth to connection assembly, record the area of each connection assembly;

(3) select MSER: under assign thresholds, from the connection assembly that (2) obtain, pick out assembly area change obtains local minimum connection assembly with changes of threshold, obtain maximum stable extremal region;

(4) for each maximum stable extremal region, computation of mean values u and covariance Σ, recycling elliptic region equation (x-u) ^t∑ ^-1(x-u)=1, becomes elliptic region according to actual area form fit;

Second step: screen the elliptic region simulated, eliminates region overlapping, chooses the elliptic region for final watermark embedment;

(1) the excessive too small elliptic region of filter area, the elliptic region area of reservation meets τ ₁< | R _i| < τ ₂, wherein, | R _i| represent the area of i-th elliptic region, τ ₁and τ ₂represent lower bound and the upper bound of size respectively;

(2) design of graphics model, in graph model, the corresponding summit V of each elliptic region i _i; Summit is to V _i, V _jbetween by limit E _ijconnect; Limit E _ijweight w _ijequal summit to V _i, V _jbetween distance, (V, E) forms Undirected graph G;

(3) minimal spanning tree algorithm is used to obtain minimum spanning tree T at Undirected graph G;

(4) judge that in T, whether often pair of characteristic area is overlapping according to elliptic equation, if overlapping, then these two elliptic regions put identical cluster under, otherwise, put different clusters under;

(5) pick out the elliptic region that in each cluster, area is maximum, these elliptic regions will be used for final watermark embedment;

3rd step: by revising the vertical low-frequency band of horizontal high-frequent (HL) and the vertical high frequency band of horizontal low frequencies (LH) of wavelet conversion coefficient, carry out watermark embedment to the characteristic area selected, method is as follows:

(1) elliptic region normalization: first, to the inverse ∑ of the covariance of the elliptic region picked out ^-1carry out singular value decomposition, ∑ ^-1=Q Λ Q ^-1, wherein, Q represents real Orthogonal Symmetric unitary matrice, Q ^-1represent that Q's is inverse, Λ represents diagonal matrix; Then, z=Q is made ^t(x-μ) obtains z ^tΛ z=1, wherein, the transposition of T representing matrix, z represents the circle coordinates after elliptic coordinates x normalization, finds transformation equation z=Q ^t(x-μ), normalizes to circle coordinates z by elliptic coordinates x;

(2) zero padding outside border circular areas after normalization, is filled to external square area;

(3) utilize Algorithms of Discrete Wavelet Transform to carry out rank transformation again to the advanced every trade conversion of obtained external square area, obtain wavelet conversion coefficient matrix;

(4) adaptive wavelet coefficient modifying, method is as follows: the vertical high frequency band LH of the horizontal low frequencies (x of note wavelet conversion coefficient matrix, y) with the vertical low-frequency band HL (x of horizontal high-frequent, y) difference is D (x, y), if robustness threshold value is α, when watermark bit w (x, y) is embedded into, coefficient differentials must meet lower relation: D (x, y) >=α, if w (x, y)=1, D (x, y) <-α, ifw (x, y)=0;

(5) reverse wavelet transform is carried out to amended matrix of wavelet coefficients, obtain amended square features region;

(6) get inscribed circle and oppositely normalization at square features intra-zone, obtain watermarked after oval feature region;

(7) characteristic area replace: remember watermarked after characteristic area R ₂with primitive character region R ₁difference be R _d, by R _dbe added in original video frame.

2. the video watermark method for anti-counterfeit of resist geometric attacks according to claim 1, is characterized in that, described method for anti-counterfeit also comprises watermark extraction step below:

(1) to the frame of video being embedded with watermark, extract according to the method for the first step maximum stable extremal region MSER being embedded with the frame of video of watermark;

(2) carry out elliptic region screening according to the method for second step, obtain the elliptic region that may comprise watermark;

(3) to the elliptic region that may comprise watermark, (1), (2) and (3) of the 3rd step is performed successively;

(4) watermark bit blind Detecting: note w'(x, y) be watermark bit detected at (x, y), then the mode of watermark bit blind Detecting is as follows:

$W^{\&prime;}(x,y)=\left\{\begin{array}{ccc}1,& \mathrm{if}& D^{\&prime;}(x,y)\&GreaterEqual;0\\ 0,& \mathrm{if}& D^{\&prime;}(x,y)<0\end{array}\right.,$ Wherein, D'(x, y)=HL'(x, y)-LH'(x, y), HL'(x, y) and LH'(x, y) wavelet coefficient of expression at position (x, y) place;

(5) recovery rate is calculated, if recovery rate is less than certain threshold tau _nc, be considered as in this region not with watermarked information, get τ _nc=0.68;

(6) final watermark w* (x, y) ballot obtains according to following ballot formula:

$w^{*}(x,y)=\left\{\begin{array}{ccc}1,& \mathrm{if}& {\mathrm{Num}}_1(x,y)\&GreaterEqual;{\mathrm{Num}}_0(x,y)\\ 0,& \mathrm{if}& {\mathrm{Num}}_0(x,y)<{\mathrm{Num}}_1(x,y)\end{array}\right.$

Wherein, w* (x, y) represents the watermark bit voting results at position (x, y) place, Num ₁(x, y) and Num ₀(x, y) is illustrated respectively in the number of " 1 " that position (x, y) detects and the number of " 0 ".

3. the video watermark method for anti-counterfeit of resist geometric attacks according to claim 1, is characterized in that, it is 0.09 that robustness puts threshold alpha.