CN1885342A - Print domain trademark anti-counterfeit method based on digital watermark technology - Google Patents

Abstract

The invention relates to a print field mark false-proof method, based on digit watermark technique, wherein the mark image is embedded with watermark false-proof information, which comprises: 1, using self-adaptable watermark strength to embed two-dimension random copy watermark into the idle field of original mark image, to generate false-proof mark image; 2, using false-proof mark image and one-dimension Ladon transformation to generate self-registration template, to be used to calculate self-registration geometrical deformation parameter at the mark false check step; 3, generating two-dimension random copy watermark which is same as in step 1; the image collector collects tested mark image, extracting self-registration parameter; based on self-registration parameter, synchronizing the two-dimension random copy watermark; based on synchronized watermark and tested mark image, using self-related judgment, to realize the recognition of real and false mark.

Description

Print domain trademark anti-counterfeit method based on digital watermark technology

Technical field

The invention belongs to field of anti-counterfeit technology, particularly be used for print domain trademark anti-counterfeit, realize the genuine piece of printed matter trade mark and a kind of print domain trademark anti-counterfeit method of fakement identification based on digital watermark technology.

Background technology

Digital watermarking is as a kind of Information Hiding Techniques of novelty, for solve copyright protection, the source authentication on the open network, a series of problems such as authentication, distribution on the net, usertracking and authentication of distorting provide brand-new solution thinking.Be the extensive concern that the digital watermark technology of research object has obtained domestic and international academia with the image, and obtain to a certain degree application on the net at Internet, mainly the image copyright to digital form plays the better protect effect.The digital figure watermark technology tool transparency can not change visual quality of images, and human eye is difficult for discovering, and robustness is better.Make common digital figure watermark technology obtain further key in application point and difficult point is how to resist geometric attack, how to resist digital-to-analogue/modulus and attack the papery protect image that is applied to analog domain with good characteristic the digital figure watermark technology.

On the other hand, fake and inferior commodities are spread unchecked on the Chinese circulation market, because trade mark is easy to but be difficult to take the effective precautionary measures by printing personation in enormous quantities.At present the anti-counterfeiting technology that generally adopts comprises that barcode technology, laser printing and photoetching, laser hologram, laser anti-counterfeit, special antiforge, fluorescence falsification preventing, magnetic anti-counterfeit and various printing ink are false proof etc., these anti-counterfeiting technologies still exist many drawbacks in the circulation of commodities process, poor as technical monopoly and uniqueness; Easily forge and reveal core technology; The regular fee of anti-counterfeit recognition has been imputed to consumer etc.

General Electronic Coding technology and most of digital watermark method only provide the antiforge function of numeric field to image, can not resist geometric attack and similar print scanned modulus digital-to-analogue and attack, and can not play false proof function at analog domain.If the image of embed watermark is transmitted at analog domain with paper-based form, and detect when having the watermark of anti-counterfeiting information, the process of printing/printing (digital to analog conversion) and scanning/shooting (analog to digital conversion) has been gone through in watermark together in company with image, this process has been introduced much noise (as thermonoise, scanning jittering noise, white balance noise, geometric deformation etc.), watermark information has been subjected to destruction to a great extent, brings big difficulty for the design of robust watermarking algorithm.Digital watermark technology is used for trademark anti-counterfeit, not only require algorithm can resist the information loss that one-step print/print scanned process is brought, also will reach the secondary printing/printing, the no longer valid antifalse effect of scanning/shooting back watermark that republish/print, scan, this makes the application of digital watermark technology in the print domain trademark anti-counterfeit field become more challenging.

Summary of the invention

The objective of the invention is to realize that for solution utilizes digital watermark technology the anti-fraud functional technical matters of trade mark proposes a kind of method of the print domain trademark anti-counterfeit based on digital watermark technology, promptly in the trademark image of numeric field behind the embed watermark anti-counterfeiting information, trademark image is through the first impression or the printing field that circulates, after scanning or shooting, watermark information when discerning the false from the genuine in the image is effectively, i.e. a robust; The watermark information that scanning or the trademark image that obtains of shooting are republished or print, scan or make a video recording after duplicating is invalid, and is fragile, i.e. the secondary fragility.Above-mentioned antiforge function abbreviates robust, secondary fragility as.

The present invention proposes a kind of print domain trademark anti-counterfeit method based on digital watermark technology: at first propose a kind of two random orders and duplicate the production method of watermark, the production process of watermark has increased the difficulty that gangs up against, and has overcome the high frequency noise that printing or printing and scanning or shooting process bring and the influence of registration error simultaneously; Secondly combine with human visual system's feature, adopt and improve Harris (Harris) Corner Detection device calculating self-adapting watermarking intensity; Utilize orthogonal experiment design, on the basis of experiment, carry out statistical study, optimize the parameter of control self-adapting watermarking intensity in the watermarking algorithm, make the performance of watermarking algorithm reach robust one time, the false proof requirement of secondary fragility, utilize trademark image to resist rotation/convergent-divergent/translation (RST:Rotation as the self registration template at last at one dimension Laden (Radon) conversion coefficient of certain given angle, Scaling and Translation) attack, realization does not have the self registration of original image to guarantee the correct detection of watermark; To the trademark image of regular shape, directly utilize the geometric properties of regular shape to carry out registration, realize the watermark synchronization pre-service of watermark detection; Calculate the related coefficient of trademark image to be measured and synchronized watermark information, judge the true and false of trade mark to be measured according to the related coefficient size.

A kind of method of the print domain trademark anti-counterfeit based on digital watermark technology may further comprise the steps:

A) generative process of anti-false trademark is carried out the embedding that watermark is duplicated in the two-dimensional random position with self-adapting watermarking intensity in the spatial domain of original trademark image, obtains the anti-false trademark image.B) self registration template generative process.Utilize the anti-false trademark image in the step a) to generate the self registration template that is used for trade mark authenticity process self registration geometric deformation calculation of parameter; Anti-false trademark image for having regular geometric shapes need not to generate the self registration template.C) trade mark authenticity process, utilize image capture device to carry out the trademark image collection, the image that collects is a trademark image to be measured, after trademark image to be measured carried out the self registration parameter extraction, watermark is duplicated in the two-dimensional random position carry out synchronous processing, utilize the correlativity judgement that the judgement of the trade mark true and false is carried out in watermark after the synchronous processing and trademark image to be measured.

The method based on the print domain trademark anti-counterfeit of digital watermark technology that the present invention proposes is a kind of spatial domain watermarking algorithm, promptly carries out the embedding of watermark at the brightness space of original trademark image.If f (x, y) the original trademark image of representative coordinate (x, the pixel value of y) locating, f ' (x, y) represent behind the embed watermark information the anti-false trademark image coordinate (then the telescopiny of watermark can be expressed as for x, the pixel value of y) locating:

f′(x，y)＝f(x，y)+W(x，y)β(x，y) (1)

Wherein (x, y) ((x is that self-adapting watermarking intensity β is in coordinate (x, the value of y) locating y) to β to W for x, the value of y) locating at coordinate for watermark W is duplicated in the two-dimensional random position.

The generative process of anti-false trademark:

According to original trademark image size M * N and position frequency reproduction f, determine that original two dimensional random watermark L size is (M/f) * (N/f).Utilize key key (user ID or the numeral that is generated by non-reversible algorithm) to produce an one dimension pseudo-random sequence, its length equates with the line number of original two dimensional random watermark L, is M/f, and its element is a natural number; Claim that this sequence is random row offset sequence O.Utilizing same key to produce a length is the pseudo-random sequence w of (1,1) greater than max (O)+(N/f), element; Element with random row offset sequence O is a side-play amount, respectively w is carried out M/f displacement to the right, obtains the new sequence of M/f length more than or equal to (N/f); Each new sequence of brachymemma is to length N/f, and producing size is the original two dimensional random watermark L of (M/f) * (N/f); It is that duplicate the position of f that original two dimensional random watermark L is carried out frequency, generates size for duplicating watermark W in the two-dimensional random position of M * N, i.e. the watermark anti-counterfeiting information.The position replicate run that the original two dimensional random watermark is carried out is handled, can be overcome the high frequency noise that the watermark trademark image introduces and the influence of registration error in transmission course.

The present invention utilize a kind of improvement Harris Corner Detection device come in the calculating formula (1) self-adapting watermarking intensity β (x, y).If G _x, G _y(x, pixel y) is along the gradient of x and y direction at coordinate to represent original trademark image respectively.With each pixel of original trademark image is research object generator matrix M:

$M(x,y)=\left[\begin{array}{cc}\underset{m,n}{\mathrm{\Σ}}{\left(G_x(m,n)\right)}^2& \underset{m,n}{\mathrm{\Σ}}{G}_x(m,n)G_y(m,n)\\ \underset{m,n}{\mathrm{\Σ}}{G}_x(m,n)G_y(m,n)& \underset{m,n}{\mathrm{\Σ}}{\left(G_y(m,n)\right)}^2\end{array}\right]---\left(2\right)$

Then Corner Detection device output valve is:

H(x，y)＝det(M(x，y))-0.04×trace(M(x，y)) (3)

Obtain self-adapting watermarking intensity β (x, y):

$\mathrm{\β}(x,y)=\frac{H(x,y)}{\max \left(H(m,n)\right)}l+g---\left(4\right)$

Wherein, (m is n) so that (x is a value in the circular coordinates regional of radius for the center of circle, r y) for coordinate; R is a value greater than 1 integer, and l is the local strength factor of watermark, and g is the overall strength factor of watermark.L, g value can be controlled self-adapting watermarking intensity for just by choosing of l, g.

The present invention adopts the orthogonal method design that experimentizes, and (x, controlled variable l y), g realize the antiforge function of trademark image as previously described, reach the vision invisibility to optimize self-adapting watermarking intensity β.Orthogonal experiment divides two stages to carry out, the phase one, under the requirement of a robust, experimentize, and determine the variation range of parameter r, l and g, and provide the size of three parameters the watermark embed strength influence; Subordinate phase experimentizes under the requirement of secondary fragility, will further be dwindled by the parameter r that obtains previous stage, l and g variation range, thereby obtain making the parameter value interval of a robust secondary fragility of watermark anti-counterfeiting information.

The present invention can effectively control anti-counterfeiting performance by controlled self-adapting watermarking intensity.Overall situation strength factor is big more, and the robustness of method for anti-counterfeit of the present invention is just strong more; Utilize Corner Detection device and local strength factor to portray the local characteristic of original trademark image, fully taken into account HVS (human visual system) characteristic, at the less watermark anti-counterfeiting information of the low frequency part embedment strength of original trademark image, at the bigger watermark anti-counterfeiting information of the HFS embedment strength of original trademark image.

The generative process of self registration template:

Anti-false trademark image behind the embed watermark is in printing or printing, can be subjected to geometric deformation when scanning or shooting, the present invention adopts following self registration method opposing printing or prints, the geometric deformation that scanning or shooting process are brought is attacked: select a special angle θ (horizontal direction is rotated counterclockwise angle), make original trademark image in the single Radon conversion of this angle, to find Euclidean distance a little bigger greater than two local pole of threshold value D, these two points still exist after the process low-pass filtering, we claim that above-described content is self registration condition (SRC), and wherein the value of D is a positive integer; This one dimension Radon conversion coefficient as geometric templates, is used for the image self registration geometric deformation CALCULATION OF PARAMETERS of trade mark authenticity process.

Being calculated as follows between a little bigger along two local pole of x ' direction apart from d:

$d=\left|\underset{x^{\′}}{\arg }\left(\mathrm{extrem}{e}_0\left(R_{f,\mathrm{\θ}}\left(x^{\′}\right)\right)\right)-\underset{x^{\′}}{\arg }\left(\mathrm{extrem}{e}_1\left(R_{f,\mathrm{\θ}}\left(x^{\′}\right)\right)\right)\right|---\left(5\right)$

Radon conversion R wherein _{F, θ}(x ') defined by (6) formula.

Above-mentioned self registration process has mainly been utilized the rotation/convergent-divergent/translation feature of Radon conversion:

The Radon conversion represents the image as along the set of different directions projection.Image f promptly is a line integral on certain direction in the projection on certain direction.If coordinate axis x ', y ' are coordinate axis x (level), y (vertically) is through being rotated counterclockwise the coordinate axis behind the angle θ, then along the Radon conversion of x ' direction be exactly f along being parallel to the axial line integral of y ':

$R_{f,\mathrm{\θ}}\left(x^{\′}\right)={\∫}_{-K}^{K}f(x^{\′}\cos \mathrm{\θ}-y^{\′}\sin \mathrm{\θ},x^{\′}\sin \mathrm{\θ}+y^{\′}\cos \mathrm{\θ}){\mathrm{dy}}^{\′}---\left(6\right)$

Wherein the span of y ' is [K, K], and has

$\left[\begin{array}{c}{x}^{\′}\\ y^{\′}\end{array}\right]=\left[\begin{array}{cc}\cos \mathrm{\θ}& \sin \mathrm{\θ}\\ -\sin \mathrm{\θ}& \cos \mathrm{\θ}\end{array}\right]\left[\begin{array}{c}x\\ y\end{array}\right]=T\left[\begin{array}{c}x\\ y\end{array}\right]---\left(7\right)$

Matrix T is the unit orthogonal matrix.

1) zoom feature: the Radon conversion through trademark image to be measured behind the convergent-divergent is the convergent-divergent of Radon conversion same ratio on x ' direction and amplitude of original trademark image; With f (x y) represents original trademark image, q (x y) represents trademark image to be measured, when with positive zoom factor s simultaneously when level and vertical direction act on original image, have q (a, b)=f (a/s, b/s), a=xs, b=ys; Then q (a, b) satisfy in the Radon of given angle θ conversion:

R _q，θ(a′)＝sR _f，θ(sx′) (8)

2) revolving property: when image anglec of rotation α, q (a, Radon conversion b) only is the translation of original image Radon conversion on the angle axle:

R _q，θ(x′)＝R _f，θ+α(x′) (9)

3) translation feature: suffered the translation of (Δ x, Δ y) when image after, it is the simple translation of original image Radon conversion along x ' axle in the Radon of given angle θ conversion.Definition Δ x " be image during in angle θ+90 ° the Radon conversion then have along the axial translation distance of Δ x ':

$\left[\begin{array}{c}\mathrm{\Δx}\\ \mathrm{\Δy}\end{array}\right]=T^{-1}\left[\begin{array}{c}{\mathrm{\Δx}}^{\′}\\ \mathrm{\Δ}{x}^{\′\′}\end{array}\right]=T\left[\begin{array}{c}{\mathrm{\Δx}}^{\′}\\ \mathrm{\Δ}{x}^{\′\′}\end{array}\right]---\left(10\right)$

If θ ₀The one dimension Radon conversion coefficient R of direction _{F, θ 0}(x ') is exactly the self registration template that needs in the trade mark authenticity process.

Regular shape trademark image to be measured such as circular, oval, square has radius/line of apsides/standard geometric properties such as the length of side, and the present invention no longer calculates its geometric templates, but carries out registration according to the geometric properties of regular shape.

Trade mark authenticity process:

When carrying out the authenticity of trademark image to be measured, the present invention at first calculates the single Radon conversion of trademark image to be measured in each angle, and find out wherein all one dimension Radon conversion of satisfying the SRC condition and form a vector, the element number of establishing this vector is P, establishing testing image is q:

$R_q=\{R_{q,{\mathrm{\θ}}_1},R_{q,{\mathrm{\θ}}_2},...,R_{q,{\mathrm{\θ}}_P}\}---\left(11\right)$

Calculate the limit distance of each element, use symbol d _iExpression.By d ₀And d _iCalculating is with reference to yardstick s _i:

s _i＝d _i/d (12)

According to yardstick s _iConvergent-divergent x ' axle, extraction or interpolation R _{Q, θ i}(x ').

With lowest mean square root error (RMSE) is foundation, determines RMSE anglec of rotation θ hour _rWith flexible yardstick s _r, estimate actual geometric deformation parameter.Calculate R _qEach element and self registration template R _{F, θ 0}The RMSE of (x '):

$e\left({\mathrm{\θ}}_i\right)=\sqrt{\frac{1}{L}\underset{x^{\′}}{\mathrm{\Σ}}{(R_{f,{\mathrm{\θ}}_0}\left(x^{\′}\right)-R_{q,{\mathrm{\θ}}_i})}^2}---\left(13\right)$

θ _r＝arg(min(e(θ _i))) (14)

Then can obtain anglec of rotation α=θ _r-θ ₀With scale factor s=s _rThe accuracy of simultaneously available following formula conclusive evidence scale factor:

$S_r=\mathrm{extrem}{e}_i\left(R_{q,{\mathrm{\θ}}_r}\right)/\mathrm{extreme}\left(R_{f,{\mathrm{\θ}}_0}\right)---\left(15\right)$

At last, calculate (Δ x, Δ y), finish trademark image self registration geometric deformation CALCULATION OF PARAMETERS to be measured according to formula (9).

As previously mentioned, for the trademark image to be measured that has such as regular shape such as circular, oval, square, the present invention carries out registration by the geometric properties of extracting rule shape trademark image to be measured, obtains self registration geometric deformation parameter.

The present invention estimates the size of original trademark image after finishing self registration geometric deformation CALCULATION OF PARAMETERS: [M, N]=[M ', N ']/s _r, wherein [M ', N '] be the physical dimension of trademark image to be measured; To produce size be that watermark W is duplicated in the two-dimensional random position of M * N with embedding identical method of stage and key with watermark, according to the anglec of rotation that obtains previously, the peaceful amount of moving of convergent-divergent yardstick, adopt bilinear interpolation, the synchronous processing to W is rotated the convergent-divergent translation obtains synchronized watermark information W '.

If m (x, y) be trademark image q to be measured one with (x, y) for the center of circle, R are average in the border circular areas of radius, F ^-1Be that average is that 0 variance is 1 normal probability paper distribution inverse function, Thr (q) is the self-adapting watermarking decision threshold, by given probability of false detection P _FdWith local average m (x, y) decision:

$\mathrm{Thr}\left(q\right)=\frac{F^{-1}(1-P_{\mathrm{fd}})}{\sqrt{\mathrm{MN}}}\sqrt{\frac{1}{\mathrm{MN}-1}\underset{x}{\mathrm{\Σ}}\underset{y}{\mathrm{\Σ}}{(q(x,y)-m(x,y))}^2}---\left(16\right)$

Calculate the related coefficient ρ of synchronized watermark information W ' and trademark image q to be measured:

$\mathrm{\ρ}=\underset{x=1}{\overset{M}{\mathrm{\Σ}}}\underset{y=1}{\overset{N}{\mathrm{\Σ}}}{W}^{\′}(x,y)(q(x,y)-m(x,y))/\mathrm{MN}$

Judge the true and false of the picture of marking on a map to be measured by top judging process.

Decision threshold Thr (q) is adaptive to different trademark images to be measured among the present invention.Stand at trademark image to be measured under the situation of RST attack, the influence of pixel outside the trademark image to be measured that increases when rotating has been eliminated in the use of local average to a great extent, has promoted the detection performance of this algorithm greatly.

Beneficial effect of the present invention is:

1, based on the antiforge technology of digital watermarking, thoroughly upgrade the traditional concept of antiforge, had the not available peculiar property of a lot of traditional anti-counterfeiting printing technologies: hidden property, determinacy, robustness, safe reliability, technical advance and device independence etc.; Have wide practical use.

2, realized special antiforge function: watermarking algorithm was a robust when watermark trademark image was carried out one-step print or printing, scanning or shooting, and watermarking algorithm was fragile when the trademark image that scanning or shooting are obtained duplicated and republishes or print, scans or make a video recording.

3, a method of utilizing the position to duplicate has been carried out down conversion process to watermark information, can resist the high frequency noise that brings in printing (printing)/scanning (shooting) process and the influence of registration error effectively.

4, in the process that the parameter to self-adapting watermarking intensity is optimized, adopt orthogonal experiment design, under the prerequisite that does not influence experiment effect, reduce experiment number.

5, single Radon conversion being used for the image self registration makes watermarking algorithm obtain ability separately anti-and that associating RST attacks.

6, use local mean value computation self-adaptive decision threshold, eliminated the influence of null value pixel outside the trademark image to be measured that increases when rotating to a great extent, promoted the detection performance of algorithm greatly.

Print domain trademark anti-counterfeit method based on the digital figure watermark technology also will bring new thinking to other paper printed matter anti-counterfeitings that comprise certificate, bill etc.

Description of drawings

Fig. 1 is the generative process of anti-false trademark of the present invention;

Fig. 2 is the transmission course of anti-false trademark image of the present invention;

Fig. 3 is a trade mark authenticity process of the present invention;

Fig. 4 is the original trademark image of embed watermark information not in the example of the present invention;

Fig. 5 is the anti-false trademark image behind the embed watermark information in the example of the present invention;

Fig. 6 is the trademark image to be measured that utilizes scanner to gather in the example of the present invention;

Embodiment

The present invention realizes that for solution utilizes digital watermark technology the anti-fraud functional technical matters of trade mark proposes a kind of method of the print domain trademark anti-counterfeit based on digital watermark technology, promptly in the trademark image of numeric field the embed watermark anti-counterfeiting information, after, trademark image is through the first impression or the printing field that circulates, after scanning or shooting, watermark information when discerning the false from the genuine in the image is effectively, i.e. a robust; Scanning or the trademark image that obtains of shooting are republished or prints, and the watermark information after scanning or shooting are duplicated is invalid, and is fragile, and anti-counterfei waterprint information promptly has the antiforge function of robust, secondary fragility.Below in conjunction with example the present invention is further described.

Implementation process based on the trademark anti-counterfeit method of digital watermark technology is divided into following step and is described: a) generative process of anti-false trademark; B) generation of self registration template; C) transmission course of anti-false trademark image; D) trademark image self registration geometric deformation CALCULATION OF PARAMETERS to be measured; E) trademark image authenticity to be measured.

Fig. 1 illustrates the generative process of (a) anti-false trademark; (b) generation of self registration template template.Fig. 2 illustrates the transmission course of anti-false trademark image, and trade mark with the paper-based form field that circulates, utilizes plain scan instrument to gather to be measured trademark image when detecting the true and false through the print production process in the present embodiment.Fig. 3 is illustrated in the trademark image self registration geometric deformation CALCULATION OF PARAMETERS to be measured in trade mark authenticity stage and the authenticity of trademark image to be measured.Fig. 4 illustrates the not original trademark image of embed watermark anti-counterfeiting information; Fig. 5 illustrates the anti-false trademark image behind the embed watermark anti-counterfeiting information; Fig. 6 illustrates the trademark image to be measured that collects with scanner.If the resolution of original trademark image is per inch 300 pixels in the present embodiment, the resolution when scanner is gathered trademark image is per inch 300 pixels or higher, and trademark image size M * N is 300 * 300.

A) generation of anti-false trademark:

Be illustrated in figure 1 as the generative process of anti-false trademark: generate the two-dimensional random position according to key key and duplicate watermark; Calculate the self-adapting watermarking intensity of original trademark image, with the two-dimensional random position duplicate watermark by formula (1) be embedded in the original trademark image with the method in spatial domain, generate the anti-false trademark behind the embed watermark.

Generate the two-dimensional random position and duplicate watermark: M=N=300 in the present embodiment, position frequency reproduction f value 2, key key is 500.According to big or small M * N of original trademark image (illustrating as Fig. 4) and position frequency reproduction f, the size of determining original two dimensional random watermark L is 150 * 150.As shown in Figure 1, at first produce the pseudorandom line displacement sequence of an one dimension O according to key key, its length equates with the line number of original two dimensional random watermark L, is 150, and the element of line displacement sequence O is a natural number.Utilizing same key to produce a length is the pseudo-random sequence w of (1,1) greater than max (O)+(N/f), element; Element with random row offset sequence O is a side-play amount, respectively w is carried out M/f=150 displacement to the right, obtains the new sequence of M/f=150 length more than or equal to (N/f=150); So far can produce sequence number is n, and sequence w (n) side-play amount is 150 new sequence w (n+O (i)) of O (i) relatively:

C(w)＝{w(n+O(1))；w(n+O(2))；w(n+O(3))，…}

Brachymemma w (n+O (i)) is to length N/f (=150), as the delegation of original two dimensional random order watermark L, so just constitutes a size and be 150 * 150 original two dimensional random order watermark L.It is that duplicate 2 position that original two dimensional random order watermark L is carried out frequency f, and promptly the bit of 1*1 → 2*2 repeats, and then obtains size for duplicating watermark W in the two-dimensional random position of M * N (300 * 300).

The process that two random seriess produce watermark has jointly strengthened the security of watermark information; Because the object of protection of present embodiment is the papery trade mark of print production, through the signal attenuation of printing/two complex processes of scanning is shown as Fig. 2, anti-false trademark in the relative numeric field, the trademark image to be measured in trade mark authenticity stage has been subjected to even or uneven high frequency noise and has polluted, and the registration error that has certain geometric deformation to cause; The original two dimensional random watermark is carried out a position down conversion process of duplicating, can effectively resist the high frequency noise introduced in printing/scanning process and the influence of registration error.

Calculate then self-adapting watermarking intensity β (x, y), the embed watermark anti-counterfeiting information, be that watermark W is duplicated in the two-dimensional random position: (x y), draws Corner Detection device output valve H (x according to formula (3) according to formula in the summary of the invention (2) generator matrix M, y), obtain at last watermark embed strength β (x, y):

$\mathrm{\β}(x,y)=\frac{H(x,y)}{\max \left(H(m,n)\right)}l+g$

Wherein, (m is n) so that (x is a value in the border circular areas of radius for the center of circle, r y); L and g are the constants by the experiment decision.Obviously the parameter that influences watermark embed strength is: r, l, g.

Watermark W is duplicated in the two-dimensional random position, and (x y) is embedded into original trademark image, obtains anti-false trademark with self-adapting watermarking intensity β.

The parameters optimization that obtains in this example is: l=3, g=8, r=7; Watermark embed process is:

$f^{\′}(x,y)=f(x,y)+W(x,y)[\frac{H(x,y)}{\max \left(H(m,n)\right)}\×3+8]$

Table 1

Adopt orthogonal experiment to determine r, l and g among the embodiment: in three parameters, r be one greater than 1 integer variable, the sensitive range of portrayal Corner Detection device, the general stability of decision Corner Detection, variation range is less; L, g are continuous variable, directly control the embedment strength of watermark, and variation range is bigger.With the orthogonal arrage design that experimentizes, the collocation of each each level of factor is balanced, and experiment number is few, but stronger representativeness is arranged; Like this, under the prerequisite that does not influence experiment effect, reduce experiment number, and can analyze and there emerged a the influence degree of factor test findings.

In the phase one experiment, carry out one-step print/scanning detection, utilize the orthogonal method design that experimentizes; In the subordinate phase experiment, the trademark image of one-step print, scanning is republished/scan, the mode that experiment adopts parameter to travel through is carried out.

Phase one, utilize the orthogonal method design that experimentizes: utilize L ₉(3 ⁴) orthogonal arrage (table 1) designs experiment: the index of experiment is related coefficient ρ, and its numerical value is the bigger the better.This experiment has three factors (l, g, r), and wherein l is 3 levels: l=2, and 3,4; G is 3 levels: g=6,7,8; R is 2 levels: r=5,7; The number of levels of three factors is inequality in this experiment, it is the experiment of many factor problem of a mixed-level, we adopt the pseudo level method, be in the 1st, the 2 two level of 2 factor r promptly at number of levels, select a level to allow it repeat once as the 3rd level, in this example the 2nd level being repeated is the 3rd level, as testing 4 in the table 2, testing shown in 5.Table 2 has write down the The whole analytical process in this stage: the size of extreme difference has represented that g is a principal element; Excellent scheme (the A of last column ₂, B ₃, C ₂) selection represent to allow the l value near 3, the g value is near 8, the effect of r value 7 is better than 5.

Table 2

Subordinate phase, near the optimized parameter of determining previous stage, do the traversal experiment with certain step-length (this example is with being set at 0.5), the criterion of experiment is that anti-counterfei waterprint information satisfies a robust secondary fragility, thereby gets the parameter value that meets false proof requirement to the end.R=7 in addition in this example, l={3,3.5}, g={7.5,8,8.5} gets three samples under each group parameter, carry out that one-step print scanning detects and the experiment of secondary printing scanning detection, and the experimental result that obtains is listed in table 2.Can determine that the parameter value that meets the demands is: l=3, g=7.5}{l=3, g=8}{l=3, g=8.5}, obviously, the value that g and l can be continuous so we have reason to draw following conclusion, is worked as l=3,7＜g＜8.5 o'clock, algorithm satisfies false proof requirement.

Table 3

B) generation of self registration template

Illustrate as Fig. 1, after anti-false trademark generates, calculate the self registration template that is used for the trade mark authenticity.

This example d ₀Expression is corresponding to angle θ ₀The limit of single Radon conversion between distance, the anti-false trademark image is at angle θ ₀Single Radon conversion coefficient have two distances a little bigger greater than the local pole of threshold value D, these two points are through still existing after the low-pass filtering; d ₀Pressing summary of the invention Chinese style (10) calculates.With the anti-false trademark image at angle θ ₀One dimension Radon conversion coefficient as the self registration template, only according to the self registration template, can derive d at detection-phase ₀

The present invention only utilizes the one dimension Radon conversion coefficient of anti-false trademark image as the self registration template, does not need original image as the reference image.

Present embodiment trademark image (illustrating as Fig. 4) has regular figure, can carry out registration by the geometric properties of extracting rule shape trademark image to be measured, obtains self registration geometric deformation parameter, walks around self registration template calculation procedure (b).

C) transmission course of anti-false trademark

Fig. 4 illustrates the original trademark image in the present embodiment, generates anti-false trademark image as shown in Figure 5 behind the embed watermark anti-counterfeiting information, but vision acceptance height.

Illustrate as Fig. 2, after the process of the anti-false trademark behind embed watermark anti-counterfeiting information print production, transmit with the circulation of papery trade mark form at analog domain; In the time of need verifying the trade mark true or false papery trade mark is converted to again the trademark image to be detected of digital form by scanning process.

D) trademark image self registration geometric deformation CALCULATION OF PARAMETERS to be measured

As described in summary of the invention, self registration geometric deformation parameter according to self registration template extraction trademark image to be measured calculates anglec of rotation α and scale factor s, calculates translation (Δ x then, Δ y), finish trademark image self registration geometric deformation CALCULATION OF PARAMETERS to be measured.

Single Radon conversion is used for image self registration geometric deformation CALCULATION OF PARAMETERS, makes watermarking algorithm obtain ability separately anti-and that associating RST attacks.

The regular shape of original trademark image in the present embodiment need not to utilize geometric templates to calculate self registration geometric deformation parameter, directly utilizes the geometric properties of regular figure to carry out registration.Obtain the anglec of rotation and the length on limit by the detection on square shaped limit, thereby obtain the anglec of rotation of square trademark image to be measured and the information of size and displacement.

F) trade mark authenticity

As shown in Figure 3, at first duplicate watermark according to key key and step a) generation two-dimensional random position; Calculate the self registration geometric deformation parameter of trademark image to be measured according to self registration template and trademark image to be measured; According to the self registration geometric deformation parameter that obtains to the two-dimensional random position duplicate that watermark is rotated, the synchronous processing of convergent-divergent, translation, obtain synchronized watermark information W '; According to synchronized watermark and trademark image to be measured, adopt the auto-correlation judgement to identify the true and false of trade mark to be measured.

In the auto-correlation judging process, at first according to formula in the summary of the invention (16) calculated threshold Thr; Set probability of false detection P in this example _Fd=0.05, calculated threshold Thr then:

$\mathrm{Thr}\left(q\right)=\frac{F^{-1}\left(0.95\right)}{\sqrt{300\×300}}\sqrt{\frac{1}{300\×300-1}\underset{x}{\mathrm{\Σ}}\underset{y}{\mathrm{\Σ}}{(q(x,y)-m(x,y))}^2}$

Calculate the related coefficient ρ of watermark information W ' and trademark image q to be measured, the size by comparison related coefficient ρ and threshold value Thr provides court verdict: as related coefficient ρ during greater than threshold value Thr, adjudicating trademark image to be measured is non-forged trade mark; As related coefficient ρ during smaller or equal to threshold value Thr, adjudicating trademark image to be measured is forged trade mark.

Claims (10)

1. the print domain trademark anti-counterfeit method based on digital watermark technology is characterized in that, may further comprise the steps:

A) generation of anti-false trademark

Generate the two-dimensional random position according to a key and duplicate watermark, form the watermark anti-counterfeiting information;

Utilize and improve Harris's Corner Detection device calculating self-adapting watermarking intensity;

Adopt orthogonal experiment design that the parameter of decision self-adapting watermarking intensity is optimized, realize sightless watermark antiforge function, reach watermarking algorithm robust of printing/scanning process, secondary fragility;

Carry out the embedding of watermark anti-counterfeiting information with self-adapting watermarking intensity in the spatial domain of original trademark image, obtain anti-false trademark;

B) generation of self registration template

After the watermark that will comprise anti-counterfeiting information embeds original trademark image, generate the self registration template, be used for detection-phase and calculate trademark image self registration geometric deformation parameter to be measured according to anti-false trademark image and one dimension Laden conversion;

Anti-false trademark image for having regular geometric shapes need not to generate the self registration template;

C) trade mark authenticity process

Utilize image capture device to carry out the trademark image collection, the image that collects is a trademark image to be measured;

According to the self registration template, trademark image to be measured is carried out pre-service, extract self registration geometric deformation parameter,

For trademark image to be measured with regular geometric shapes, utilize its geometric properties to realize the registration of trademark image to be measured, obtain self registration geometric deformation parameter,

Adopt the key identical to generate identical watermark anti-counterfeiting information, the watermark anti-counterfeiting information is carried out synchronous processing with trademark image to be measured according to self registration geometric deformation parameter with the embedding stage, i.e. rotation, convergent-divergent and translation,

The watermark information after utilizing synchronously and the correlativity of trademark image to be measured are carried out the judgement of the trademark image true and false to be measured, whether have antiforge function.

2. the print domain trademark anti-counterfeit method based on digital watermark technology according to claim 1, it is characterized in that: watermarking algorithm was a robust when antiforge function of described watermark was meant the anti-false trademark image carried out one-step print or printing/scanning or shooting, i.e. robust, the watermark anti-counterfeiting information can be resisted the signal attenuation that one-step print or printing/scanning or shooting bring, and judges that this trademark image is for true; The scanning or the trademark image that obtains of making a video recording are duplicated, republish or print/scan or when making a video recording watermarking algorithm be fragile, be the secondary fragility, the watermark anti-counterfeiting information decays to and can't detect in secondary printing or printing/scanning or shooting process, judges that this trademark image is for false; Judge that all trademark images that do not contain the watermark anti-counterfeiting information are for false.

3. the print domain trademark anti-counterfeit method based on digital watermark technology according to claim 1, it is characterized in that: a described key is meant a numeral corresponding with the anti-false trademark image, can be user ID or the numeral that generates by non-reversible algorithm, the blanket brand image of same batch can use same key, and key is held by the grantee of trade-mark.

The generation of watermark is duplicated in described two-dimensional random position:

In the process that produces the watermark anti-counterfeiting information, for overcoming the influence of high frequency noise and registration error, watermark information has been carried out the down conversion process that duplicate the position, the position frequency reproduction is f,

According to original trademark image size M * N and position frequency reproduction f, determine that original two dimensional random watermark L size is (M/f) * (N/f),

The key key of the numeral of utilizing user ID or being generated by non-reversible algorithm produces an one dimension pseudo-random sequence, and its length equates with the line number of original two dimensional random watermark L, is made as M/f, and its element is a natural number; Claim that this sequence is random row offset sequence O,

Utilizing same key to produce a length is the pseudo-random sequence w of (1,1) greater than max (O)+(N/f), element;

Element with random row offset sequence O is a side-play amount, respectively w is carried out M/f displacement to the right, obtains the new sequence of M/f length more than or equal to (N/f); Each new sequence of brachymemma is to length N/f, and producing size is the original two dimensional random watermark L of (M/f) * (N/f);

It is that duplicate the position of f that original two dimensional random watermark L is carried out frequency, generates size for duplicating watermark W in the two-dimensional random position of M * N, i.e. the watermark anti-counterfeiting information.

4. the print domain trademark anti-counterfeit method based on digital watermark technology according to claim 1, it is characterized in that: described utilization improves Harris's Corner Detection device and calculates self-adapting watermarking intensity, if (x, y) be a pixel coordinate point of anti-false trademark image, utilize a kind of improvement Harris Corner Detection device H to calculate self-adapting watermarking intensity at (x, y) value of locating: (x, y)=H (x, y) * and l/max (H (m, n))+g, wherein, coordinate (m, n) so that (x y) is the center of circle, and r is a value in the circular coordinates regional of radius; The value of r is the integer greater than 1; L is the local strength factor of watermark, and value is a positive number; G is the overall strength factor of watermark, and value is a positive number, by l, g control self-adapting watermarking intensity.

5. according to claim 1 or 4 described print domain trademark anti-counterfeit methods based on digital watermark technology, it is characterized in that: described parameter to decision self-adapting watermarking intensity is optimized and is meant that the employing orthogonal experiment design method carries out test design, parameter l, the g of optimal control self-adapting watermarking intensity reach the false proof false proof purpose of described watermark.

6. according to claim 1 or 5 based on the print domain trademark anti-counterfeit method of digital watermark technology, it is characterized in that: the described embedding of carrying out the watermark anti-counterfeiting information with self-adapting watermarking intensity in the spatial domain of original trademark image, obtain anti-false trademark, if f is (x, y) the original trademark image of representative is at coordinate (x, the pixel value of y) locating, f ' (x, y) represent anti-false trademark image behind the embed watermark information at coordinate (x, y) pixel value of locating, (x y) duplicates watermark W at coordinate (x for the two-dimensional random position to W, y) value of locating, x, y are self-adapting watermarking intensity, at coordinate (x, y) value of locating, then the telescopiny of watermark anti-counterfeiting information can be expressed as:

f′(x，y)＝f(x，y)+W(x，y)β(x，y)

7. the print domain trademark anti-counterfeit method based on digital watermark technology according to claim 1, it is characterized in that: a special angle θ is selected in being generated as of described self registration template, make original trademark image in single Laden's conversion of this angle, to find Euclidean distance a little bigger greater than two local pole of threshold value D, these two points still exist after the process low-pass filtering, wherein, threshold value D value is a positive number; This one dimension Laden conversion coefficient as geometric templates, is used for the image self registration geometric deformation CALCULATION OF PARAMETERS of trade mark authenticity process; The geometric deformation parameter of trademark image to be measured comprises: the anglec of rotation of the original relatively trademark image of trademark image to be measured, the peaceful distance of moving of convergent-divergent yardstick.

8. the print domain trademark anti-counterfeit method based on digital watermark technology according to claim 1, it is characterized in that: described anti-false trademark image with regular geometric shapes, need not to generate the self registration template, such as circular, oval, square regular shape trademark image to be measured, have radius/line of apsides/standard geometric properties such as the length of side, self registration CALCULATION OF PARAMETERS process is except the embedding of carrying out the watermark anti-counterfeiting information by the spatial domain of original trademark image realizes, also the geometric properties by extracting rule shape trademark image to be measured carries out.

9. the print domain trademark anti-counterfeit method based on digital watermark technology according to claim 1, it is characterized in that: described the watermark anti-counterfeiting information is carried out synchronous processing with trademark image to be measured, utilize self registration geometric deformation parameter, adopt bilinear interpolation to the two-dimensional random position duplicate that watermark W is rotated, the synchronous processing of convergent-divergent, translation.

10. the print domain trademark anti-counterfeit method based on digital watermark technology according to claim 1, it is characterized in that: the watermark information after the described utilization synchronously and the correlativity of trademark image to be measured are carried out the judgement of the trademark image true and false to be measured, if w ' is the watermark information after synchronously, trademark image to be measured is q, m (x, y) be q one with (x, y) for the center of circle, R are average in the border circular areas of radius, F ¹Be that average is that 0 variance is 1 normal probability paper distribution inverse function, Thr (q) is the self-adapting watermarking decision threshold, by given probability of false detection P _FdWith local average m (x, y) decision:

$\mathrm{Thr}\left(q\right)=\frac{F^{-1}(1-P_{\mathrm{fd}})}{\sqrt{\mathrm{MN}}}\sqrt{\frac{1}{\mathrm{MN}-1}\underset{x}{\mathrm{\Σ}}\underset{y}{\mathrm{\Σ}}{(q(x,y)-m(x,y))}^2}$

Calculate the related coefficient of w ' and q:

$\mathrm{\ρ}=\underset{x=1}{\overset{M}{\mathrm{\Σ}}}\underset{y=1}{\overset{N}{\mathrm{\Σ}}}{W}^{\′}(x,y)(q(x,y)-m(x,y))/\mathrm{MN}$

Make judgement: ρ≤Thr (q), the watermark anti-counterfeiting information does not exist, and trade mark is false; ρ＞Thr (q), the watermark anti-counterfeiting information exists, and trade mark is true.

Images