CN1333371C - Digital watermark method capable of resisting geometric attack and conventional attack - Google Patents

Abstract

The present invention discloses a digital watermark technology based on full figure DCT conversion. The present invention belongs to the field of multimedia signal treatment. The present invention first embeds a watermark, which comprises the following steps: step 1, the full figure DCT conversion of an original figure is performed, and one vector capable of repressing the important visual characteristic of the original figure from a conversion coefficient is abstracted; step 2, through the characteristic vector and the watermark which needs to be embedded, one binary value logic sequence is obtained by a Hash function. The present invention then extracts the watermark, which comprises the following steps: step 3, the full figure DCT conversion is carried out for an image which needs to be measured, and one visual characteristic vector of the image which needs to be measured is found; step 4, the watermark is abstracted with the property of the Hash function and the binary value logic sequence which is obtained when the watermark is embedded. The present invention is the digital watermark technology based on full figure DCT conversion. Experiments show that the present invention has the advantages of powerful geometric attack ability and conventional attack resisting ability.

Description

But the digital watermark method of a kind of resist geometric attacks and conventional attack

Technical field

The present invention relates to a kind of digital watermark technology based on dct transform and Image Visual Feature, is a kind of multimedia data protection method, belongs to field of multimedia signal processing.

Technical background

Digital watermarking at present is one of effective ways of protection Digital Media copyright.Digital watermarking has invisibility and robustness.Aspect robustness, present most of watermarks can only be resisted conventional attack, for common or even small geometric attack is all powerless.In recent years, abroad some scholar's Against Geometrical Attacks watermarking algorithms have been done some theoretical researches, but effect is unsatisfactory.O ' Ruanaidh etc. are based on Fourier-Mellin conversion (containing twice DFT conversion and the conversion of a LPM logarithm polar coordinates), utilize this conversion to have the anti-RST (Rotate that the RST unchangeability realizes watermark, Scalling, Transtation) geometric attack ability, but this method makes the watermarking images quality become relatively poor.Pereira etc. have used the template insertion, but the embedding of template and extraction are complicated, and the information capacity that embeds is little.Propositions such as Pitas embed watermark information on DFT territory annular region or the circumference, but it can only provide less anti-rotation performance and embed less information.Though the localization watermark of propositions such as Kutter has stronger robustness to shearing, can't resist the combination attacks of geometric transformation, the himself has also proposed the notion of second generation digital watermarking, promptly watermark information is embedded in the characteristics of image vector.But all there are some problems, and

Above major part is that what to adopt is the DFT conversion, and present main flow conversion is DCT and DWT, DCT be present JPEG compression, H.261/263 with the standard of MPEG-1/2, its fast operation, the precision height, be celebrated with the ability of extracting characteristic component and the optimum balance between the arithmetic speed, DWT is the JPEG2000 standard.

In the digital watermarking research field, geometric attack is still a more insoluble problem up to now, and the Image Watermarking as for being highly resistant to these two kinds of attack types simultaneously has not yet to see report, still belongs to blank.And in the practical application, digital watermarking image usually is subjected to this two kinds of attacks simultaneously.

Summary of the invention

The objective of the invention is to propose a kind ofly can resist digital watermarking embedding and the extracting method that geometric attack can be resisted conventional attack again, it has higher robustness, with the copyright of protection Digital Media.

To achieve these goals, the present invention is performed such: based on full figure DCT and visual feature vector, and digital watermark and cryptography combined, realized the anti-geometry and the conventional attack of digital watermarking.The method applied in the present invention comprises watermark embedding and watermark extracting two large divisions, first is that watermark embeds, comprise: (1) is by carrying out the full figure dct transform, obtain a visual feature vector V (j) of image, (2) generate a two-valued function sequence key (j) according to watermark W (j) and visual feature of image vector V (j).Second portion is a watermark extracting, comprising: the visual feature vector V ' that (3) obtain testing image (j), (4) utilize two-valued function sequence key (j) and testing image visual feature vector V ' (j), extract watermark W ' (j).

Now be elaborated as follows to method of the present invention:

First:, obtain two-valued function sequence key (j) by the embedding operation of watermark

At first with one group of binary pseudo-random W that can represent copyright information,

W={w (j) | w (j)=0,1; 1≤j≤L} is as digital watermarking, original image be designated as F={f (i, j) | f (i, j) ∈ R; 1≤i≤N1,1≤j≤N2) }, wherein, w (j) and f (i j) represents the grey scale pixel value of watermark sequence and original image respectively, establishes N1=N2=N, and the embedding of watermark is as follows:

1), obtains visual feature of image vector V (j) by carrying out the full figure dct transform

(i j) carries out the full figure dct transform, obtains DCT matrix of coefficients FD (i to former figure F earlier, j), (i j), carries out Zig-Zag scanning to DCT matrix of coefficients FD again, obtain frequency DCT coefficient sequence Y (j) from low to high, get preceding L value, and obtain this visual feature of image vector V (j) by the computing of DCT coefficient symbols, specific practice is to work as the DCT coefficient for we represent with " 1 " just, with " 0 " expression (reason is part as follows), program description was as follows when coefficient was negative or zero:

FD(i，j)＝DCT2(F(i，j))

Y(j)＝Zig_Zag(FD(i，j))

V(j)＝-Sign(Y(j))

2) generate a two-valued function sequence key (j) according to watermark W (j) and visual feature of image vector V (j)

key(j)＝V(j)W(j)

Key (j) is by visual feature of image vector V (j) and watermark W (j), and the Hash function commonly used by cryptography generates.Preserve key (i), need use when extracting watermark afterwards.By key (j) is applied for to the third party as key,, reach the purpose of copyright protection to obtain the entitlement of original works.

Second portion: the extraction of watermark

3) the visual feature vector V ' that obtains testing image (j)

If testing image be F ' (i, j), through obtain behind the full figure dct transform DCT matrix of coefficients be FD ' (i, j), by above-mentioned Stepl method, the visual feature vector V ' that tries to achieve testing image (j);

FD’(i，j)＝DCT2(F’(i，j))

Y’(j)＝Zig_Zag(FD’(i，j))

V’(j)＝-Sign(Y’(j))

4) in image to be measured, extract watermark W ' (j)

According to the visual feature vector V ' of key (j) that generates when the embed watermark and testing image (j), utilize watermark W ' that Hash character can extract testing image (j)

W’(j)＝key(j)V’(j)

Differentiate the owner of testing image again according to W (j) and W ' degree of correlation (j).

The present invention and existing digital watermark relatively have following advantage:

Because the present invention is based on the digital watermark technology of dct transform, it is fast to have computing velocity, and the precision height has compatibility preferably, and stronger resist geometric attacks ability and anti-conventional attack ability are arranged; Embed watermark does not influence original image quality, is a kind of harmless watermark, has practical value at aspects such as medical treatment, remote measurement, aviation shootings, and usable range is wide, and can realize the embedding and the extraction of many watermarks and big watermark easily.

Below we from the explanation of theoretical foundation and test figure:

1) discrete cosine transform

DCT is used for the standard that picture coding is present widely used JPEG compression and MPEG-1/2.DCT is the suboptimum orthogonal transformation that is only second to Karhunen-Loeve transformation that draws for a short time in the Minimum Mean Square Error condition, is a kind of harmless chief of a tribe's conversion.Its fast operation, the precision height is celebrated with the ability of extracting characteristic component and the optimum balance between the arithmetic speed.

2-D discrete cosine direct transform (DCT) formula is as follows:

$F(u,v)=c\left(u\right)c\left(v\right)\underset{x=0}{\overset{M-1}{\mathrm{\Σ}}}\underset{y=0}{\overset{N-1}{\mathrm{\Σ}}}f(x,y)\cos \frac{\mathrm{\π}(2x+1)u}{2M}\cos \frac{\mathrm{\π}(2y+1)v}{2N}$

u＝0，1，…，M-1； v＝0，1，…，N-1；

In the formula $c\left(u\right)=\left\{\begin{array}{cc}\sqrt{1/M}& u=0\\ \sqrt{2/M}& u=\mathrm{1,2},...,M-1\end{array}\right.$

$c\left(v\right)=\left\{\begin{array}{cc}\sqrt{1/N}& v=0\\ \sqrt{2/N}& v=\mathrm{1,2},...,N-1\end{array}\right.$

2-D discrete cosine inverse transformation (IDCT) formula is as follows:

$f(x,y)=\underset{u=0}{\overset{M-1}{\mathrm{\Σ}}}\underset{v=0}{\overset{N-1}{\mathrm{\Σ}}}c\left(u\right)c\left(v\right)F(u,v)\cos \frac{\mathrm{\π}(2x+1)u}{2M}\cos \frac{\mathrm{\π}(2y+1)v}{2N}$

x＝0，1，…，M-1； y＝0，1，…，N-1

X wherein, y is the spatial domain sampled value; U, v are the frequency field sampled value, and digital picture is represented with the pixel square formation usually, i.e. M=N

From top formula as can be known, the coefficient symbols of DCT is relevant with the phase place of component.

2) a kind of choosing method of image vision principal character vector:

The main cause of present most of watermarking algorithm resist geometric attacks ability is: people are embedded in digital watermarking in pixel or the conversion coefficient, and the slight geometric transformation of image usually causes the bigger suddenly variation of pixel value or transform coefficient values.The watermark of Qian Ruing is just attacked easily like this.If can find the visual feature vector of reflection image geometry characteristics, when little geometric transformation took place image, tangible sudden change can not take place in this visual feature of image value so.Hayes studies show that for characteristics of image, phase place is more important than amplitude.We find through observing a large amount of full figure DCT data (Low Medium Frequency), when an image is carried out common geometric transformation, some variations may take place in the Low Medium Frequency coefficient magnitude, but its coefficient symbols remains unchanged substantially, and we choose some test figures and are shown in Table 1.The former figure that is used as test in the table 1 is Fig. 1 (a), is the lena512 of a band black surround frame.What the 1st row showed in the table is image type under attack, and the image that is subjected to behind the conventional attack is seen Fig. 1 (b)-(d), and the image that is subjected to behind the geometric attack is seen Fig. 2 (a)-(e).The 2nd is listed as the 11st row, and this is F (1,1)-ten Low Medium Frequency coefficients of F (1,10) of getting in the DCT matrix of coefficients, wherein the DC component value of coefficient F (1,1) presentation video.For conventional attack, these Low Medium Frequency coefficient values F (1,1)-F (1,10) remains unchanged and former figure value approximately equal substantially; For geometric attack, the part coefficient has bigger variation, but we can find that image is when being subjected to geometric attack, and the size of part DCT Low Medium Frequency coefficient has taken place to change but its symbol does not change substantially.We use positive DCT coefficient " 1 " expression, negative coefficient is used " 0 " expression (containing value is zero coefficient), so for former figure, F (1 in the DCT matrix of coefficients, 1)-F (1,10) coefficient, corresponding coefficient symbols sequence is: " 1001000110 ", see Table 1 the 12nd row, we observe these row and can find, no matter conventional attack still is this symbol sebolic addressing of geometric attack can keep similar with former figure, with former figure normalized correlation coefficient all big (seeing the 13rd row), (having got 10 DCT coefficient symbols here for the purpose of convenient).But for the dct transform coefficient symbol sebolic addressing that further proves full figure is a vision key character that belongs to this figure, we are again different test patterns (seeing Fig. 3 (a)-(f)), carry out the full figure dct transform, obtain corresponding DCT coefficient F (1,1)-F (1,10), be shown in Table 2, the 12nd classifies DCT coefficient symbols sequence as, the 13rd classifies the degree of correlation with former schematic symbol series as, can find different test patterns, their degrees of correlation are very little, and this can illustrate that also the symbol sebolic addressing of DCT coefficient can reflect the main visual signature of this image.After watermarking images was subjected to a certain degree conventional attack and geometric attack, this vector was constant substantially, and this also meets the DCT ability that " very strong extraction characteristics of image arranged ".

Table 1 image full figure dct transform Low Medium Frequency part coefficient and be subjected to different the attack after changing value

	The 1st row	The 2nd row	The 3rd row	The 4th row	The 5th row	The 6th row	The 7th row	The 8th row	The 9th row	The 10th row	The 11st row	The 12nd row	The 13rd row
															Image manipulation	F(1，1)	F(1，2)	F(1，3)	F(1，4)	F(1，5)	f(1，6)	F(1，7)	F(1，8)	F(1，9)	F(1，10)	The coefficient symbols sequence	The degree of correlation
Conventional attack	Former figure (frame is arranged)	2.219	-0.164	-1.381	0.325	-0.532	-0.204	-0.107	0.053	0.491	-0.077	1001000110	1.0
														The jpeg compression	2.221	-0.162	-1.380	0.324	-0.534	-0.205	-0.110	0.052	0.490	-0.076	1001000110	1.0
	Gauss disturbs	2.226	-0.181	-1.378	0.326	-0.542	-0.189	-0.107	0.076	0.487	-0.083	1001000110	1.0
														Medium filtering	2.208	-0.169	-1.375	0.335	-0.526	-0.210	-0.116	0.052	0.497	-0.075	1001000110	1.0
	Geometric attack	Move horizontally	2.219	-0.341	-1.331	0.342	-0.590	-0.054	-0.054	0.283	0.398	-0.128	1001000110														1.0
Vertical moving														2.219	-0.180	-1.378	0.327	-0.538	-0.192	-0.103	0.075	0.488	-0.083	1001000110	1.0
		Rotation	2.219	-0.131	-1.362	0.260	-0.507	-0.138	-0.055	0.077	0.349	-0.150	1001000110													1.0
Lena amplifies														3.203	-0.275	-1.207	0.452	-0.709	-0.162	-0.795	0.076	-0.200	-0.207	1001000100	0.75
		Lena dwindles	1.427	-0.081	-1.236	0.194	-0.047	-0.191	0.312	0.096	0.205	-0.021	1001001110													0.83

* coefficient unit 1.0e+004

The Low Medium Frequency part coefficient of the test pattern full figure dct transform that table 2 is different

The 1st row	The 2nd row	The 3rd row	The 4th row	The 5th row	The 6th row	The 7th row	The 8th row	The 9th row	The 10th row	The 11st row	The 12nd row	The 13rd row
													Image name	F(1，1)	F(1，2)	F(1，3)	F(1，4)	F(1，5)	F(1，6)	F(1，7)	F(1，8)	F(1，9)	F(1，10)	The coefficient symbols sequence	The degree of correlation
Former figure lena512	5.071	-0.658	-0.094	0.503	0.420	-0.233	-0.033	0.218	-0.106	-0.213	1001100100	1.00
													baboo512	6.597	-0.051	0.189	0.221	-0.181	0.039	0.494	0.218	-0.456	-0.080	1011011100	0.25
peppers512	6.091	-0.006	0.021	-0.184	-0.415	-0.315	0.187	0.187	0.454	0.116	1010001111	-0.17
													crowd512	4.497	0.347	0.177	-0.063	0.015	-0.059	-0.036	0.167	0.183	0.137	1110100111	0.083
harbour512	6.430	-0.267	-0.073	0.079	0.023	0.032	0.030	-0.021	-0.035	-0.071	1001111000	0.01
													woman512	5.521	-0.545	0.482	-1.132	0.858	0.241	0.257	0.158	0.086	0.010	1010111111	-0.08

* coefficient unit: 1.0e+004

3) position of watermark embedding and the length of disposable embedding

According to human visual system (HVS), the Low Medium Frequency signal is bigger to people's visual impact, the principal character of representative's image.Therefore our selected visual feature of image vector is the symbol (realizing the ordering of frequency by Zig_zag scanning) of Low Medium Frequency coefficient, it is relevant with the robustness of the quantity of information of the size of the original image that carries out the full figure dct transform and disposable embedding and requirement that the number of Low Medium Frequency coefficient is selected, the L value is more little, the quantity of information of disposable embedding is few more, but robustness is high more.In the test of back, the length that we choose L is 32.

4) for the embedding and the extraction of many watermarks

This invention can be carried out the embedding of many watermarks.If n watermark arranged, we can try to achieve g watermark W by Hash function and visual feature of image vector V (j) when watermark embedded ^g(j) Dui Ying two-valued function array key ^g(j), key ^g(j)=W ^g(j)  V (j), g=1,2 ... n; Preserve these key ^g(j), when the extraction of watermark, use.When extracting watermark, utilize different key ^g(j) and the visual feature vector V ' of testing image (j), can extract pairing a plurality of watermark W ' ^g(j)=key ^g(j)  V (j), g=1,2 ... n.If embed a big watermark, then can be divided into a plurality of little watermarks, embed and extract by multi-watermarking again.

In sum, we utilize the symbol sebolic addressing of Low Medium Frequency coefficient to obtain a kind of method that obtains the Image Visual Feature vector by the analysis to full figure DCT coefficient; And realized the embedding and the extracting method of many watermarks with big watermark.

Description of drawings

Fig. 1 (a) is the original image of band edge frame.

Fig. 1 (b) is the image that Fig. 1 (a) attacks through JPEG.

Fig. 1 (c) is the image that Fig. 1 (a) disturbs through Gauss.

Fig. 1 (d) is the image of Fig. 1 (a) through medium filtering.

Fig. 2 (a) is the image through moving horizontally.

Fig. 2 (b) is the image through vertical moving.

Fig. 2 (c) is the image through rotational transform.

Fig. 2 (d) is the image through convergent-divergent 1.2.

Fig. 2 (e) is the image through convergent-divergent 0.8.

Fig. 3 (a) is standardized test chart Lena512.

Fig. 3 (b) is standardized test chart Baboo512.

Fig. 3 (c) is standardized test chart Peppers512.

Fig. 3 (d) is standardized test chart Crowd512.

Fig. 3 (e) is standardized test chart Harbour512.

Fig. 3 (f) is standardized test chart Woman512.

Fig. 4 (a) is the watermarking images when not disturbing.

Fig. 4 (b) is the watermarking detecting results of Fig. 4 (a) when not disturbing.

Fig. 5 (a) is that Gauss's interference strength is 3% o'clock a watermarking images.

Fig. 5 (b) is the watermarking detecting results of Fig. 5 (a).

Fig. 6 (a) is that compression quality is 4% watermarking images.

Fig. 6 (b) is the watermarking detecting results of Fig. 6 (a).

Fig. 7 (a) is the watermarking images behind rotation 30 degree.

Fig. 7 (b) is the watermarking detecting results of Fig. 7 (a).

Fig. 8 (a) is the image that moves horizontally behind the 30pix.

Fig. 8 (b) is the watermarking detecting results of Fig. 8 (a).

Fig. 9 (a) is that zoom factor is 0.6 watermarking images.

Fig. 9 (b) is the image watermark testing result of Fig. 9 (a).

Figure 10 (a) is that zoom factor is 1.2 watermarking images.

Figure 10 (b) is the image watermark testing result of Figure 10 (a).

The watermarking images of Figure 11 (a) when being length breadth ratio 0.4.

Figure 11 (b) is the image watermark testing result of Figure 11 (a).

Figure 12 (a) is that length breadth ratio becomes 1.4 watermarking images.

Figure 12 (b) is the image watermark testing result of Figure 12 (a).

Figure 13 (a) shears 10% watermarking images.

Figure 13 (b) is the image watermark testing result of Figure 13 (a).

Embodiment

The invention will be further described uses 1000 groups of independently binary pseudo-random (value is+1 or-1) below in conjunction with accompanying drawing, every group of sequence length is 32bit, in these 1000 groups of data, appoint and extract one group (selecting the 500th group here), as the watermark sequence that embeds.Original image is seen Fig. 4 (a), be one to have the Lena512_square (512 * 512 * 8) of black surround, add dark border and be in order to guarantee energy conservation when the geometric transformations such as image rotation, if former figure is expressed as F (i, j), 1≤i≤512 wherein, the full figure DCT matrix of coefficients of 1≤j≤512 correspondences is FD (i, j), 1≤i≤512,1≤j≤512 wherein, is Y (j) to it through the DCT ordinal number that obtains frequency after the Zig_Zag scanning and sort from low to high, 1≤j≤L, the DC component of first value Y (1) representative image, frequency order is from low to high arranged then.Consider that we get preceding 32 coefficients, i.e. L=32 for the capacity of robustness and disposable embed watermark here.After detecting W ', we have judged whether that by calculating normalized correlation coefficient NC (Normalized CrossCorrelation) watermark embeds.

Fig. 4 (a) is the watermarking images that does not add when disturbing;

Fig. 4 (b) does not add when disturbing, and the output of watermark detector can be seen NC=1.0, obviously detects the existence of watermark.

Below we judge the anti-conventional attack ability and the resist geometric attacks ability robustness of this digital watermark method by concrete test

The ability (1) of the anti-conventional attack of this watermarking algorithm of test adds Gaussian noise earlier

Use imnoise () function in watermarking images, to add gaussian noise.

Table 3 is the anti-Gauss of watermark detection data when disturbing.Can see from experimental data, when Gaussian noise intensity when being 60%, watermarking images PSNR reduces to 10.82, at this moment detects watermark, related coefficient NC=0.75 still can detect the existence of watermark. this explanation adopts this invention that good anti-Gaussian noise ability is arranged.

Fig. 5 (a) is for the watermarking images when Gaussian noise intensity is 3%, and is visually very fuzzy;

The output of Fig. 5 (b) watermark detector can clearly detect the existence of watermark, NC=1.0.

The anti-Gaussian noise interference test of table 3 watermark data

Noise intensity (%)	1	3	5	10	20	40	60
								PSNR(dB)	21.48	18.31	16.73	14.59	12.81	11.40	10.82
NC	1.0	1.0	1.0	1.0	1.0	0.96	0.75

(2) the JPEG compression is handled

Adopt image compression quality percentage watermarking images to be carried out the JPEG compression as parameter;

Table 4 is the test figure of the anti-JPEG of watermarking images.When compression quality is very poor, compression quality is 2% o'clock, still can record the existence of watermark, NC=0.96.

Fig. 6 (a) is that compression quality is 4% image, and blocking artifact has appearred in this figure;

Fig. 6 (b) is the response of watermark detector, NC=0.96, and it is obvious to detect effect.

The anti-JPEG test figure of table 4 watermark

Compression quality (%)	2	4	8	10	20	40
							PSNR(dB)	26.31	27.89	30.67	31.97	34.66	35.37
NC	0.96	0.96	1.00	1.00	1.00	1.00

Watermark resist geometric attacks ability

(1) rotational transform

Table 6 is the anti-rotation of watermark challenge trial data.Can see in the table that when watermarking images rotates 45 ° NC=0.60 still can detect watermark and exist; The resist geometric attacks algorithm that people such as Pitas propose embeds watermark in the garden ring of DFT amplitude spectrum, can only resist the rotation that is not more than 3 degree.

Fig. 7 (a) is 30 ° of watermarking images rotations, the PSNR=13.36dB of watermarking images at this moment, and signal to noise ratio (S/N ratio) is very low;

Fig. 7 (b) is the watermarking images of detection, can obviously detect the NC=0.81 that exists of watermark.

The anti-rotation of table 6 watermark challenge trial data

The rotation number of degrees	0°	5°	10°	15°	20°	25°	30°	35°	40°	45°
											PSNR(dB)		17.51	15.27	14.40	14.03	13.66	13.36	13.22	13.22	13.16
NC	1.00	0.96	0.92	0.92	0.85	0.81	0.81	0.81	0.64	0.60

(2) translation transformation

Table 7 is the anti-translation challenge trial of watermark data.From table, learn, still can detect the existence of watermark, so this digital watermarking has stronger anti-translation capability as level or vertical moving 30pix.

Fig. 8 (a) is the move to right situation of 30pix of image level, PSNR=13.44dB at this moment, and signal to noise ratio (S/N ratio) is very low;

Fig. 8 (b) is watermark detector output, can obviously detect the NC=0.81 that exists of watermark.

The anti-translation test figure of table 7 watermark

	Move horizontally				Vertical moving
									Distance (pix)	5pix	10pix	20pix	30pix	5pix	10pix	20pix	30pix
PSNR(dB)	18.39	16.30	14.50	13.44	21.81	18.94	16.40	15.03
									NC	0.96	0.96	0.85	0.81	1.0	1.0	0.96	0.79

(3) scale transformation

Table 8 is watermark convergent-divergent challenge trial data, as can be seen from Table 8 when the watermarking images zoom factor little to 0.5 the time, related coefficient NC=0.6 still can record watermark.Pereira etc. ^[6]The method of inserting template in DFT that adopts can only be resisted zoom factor and be not less than 0.65 convergent-divergent, illustrates that this invention has stronger nonshrink exoergic power.

Fig. 9 (a) is 0.6 watermarking images for zoom factor, PSNR=11.82dB;

Fig. 9 (b) is a watermarking detecting results, can obviously detect the NC=0.73 that exists of watermark.

Figure 10 (a) is the watermarking images when zoom factor 1.2, at this moment center image big than former figure;

Figure 10 (b) is a watermarking detecting results, can detect the existence of watermark, NC=0.72.

Table 8 watermark convergent-divergent challenge trial data

Zoom factor	0.5	0.6	0.7	0.8	0.9	1.1	1.2	1.3
									PSNR(dB)	11.41	11.82	12.38	13.39	15.54	15.02	12.33	10.63
NC	0.60	0.73	0.73	0.92	0.92	0.72	0.72	0.56

(4) length breadth ratio changes

Table 9 is the anti-length breadth ratio change trial of watermark data, can see when length breadth ratio is 0.2 from table, still can detect the existence of watermark, and NC=0.88 illustrates the ability that this has stronger anti-length breadth ratio to change.

Figure 11 (a) becomes 0.4 watermarking images for length breadth ratio, and at this moment center image obviously narrows down, and PSNR=12.26dB is worth less.

Figure 11 (b) is a watermarking detecting results, and NC=0.88 can obviously detect the existence of watermark.

Figure 12 (a) is for becoming greatly to 1.4 when length breadth ratio, and center image sees that image obviously broadens, PSNR=12.29dB;

Figure 12 (b) is a watermarking detecting results, can obviously detect the NC=0.92 that exists of watermark.

The anti-length breadth ratio change trial of table 9 watermark data

Aspect Ratio	0.2	0.4	0.6	0.8	1.2	1.4	1.5
								PSNR(dB)	11.23	12.26	13.11	14.28	14.06	12.09	11.64
NC	0.88	0.88	0.85	0.92	0.92	0.92	0.77

(5) shear test

Table 10 is watermark cut-through resistance test data, and test figure can learn that this algorithm has certain anti-shear ability from table.

Figure 13 (a) is for to shear 10% situation to watermarking images, at this moment PSNR=15.73dB;

Figure 13 (b) is its watermark detection situation, can obviously detect the existence of watermark, NC=0.64.

Table 10 watermark cut-through resistance test data

The cutting ratio	5％	10％	15％
				PSNR(dB)	22.24	15.73	17.17
NC	0.884	0.64	0.60

Claims (3)

1. But 1, the digital watermark method of a kind of resist geometric attacks and conventional attack, it is characterized in that: based on full figure DCT and visual feature vector, and digital watermark and cryptography combined, the anti-geometry and the conventional attack of digital watermarking have been realized, this digital watermark method amounts to four steps altogether in two sub-sections:

   First is that watermark embeds: by the embedding operation to watermark, obtain corresponding two-valued function sequence key (j);

   1) former figure is carried out the full figure dct transform, from the DCT coefficient, obtain the visual feature vector V (j) of this figure according to the symbol sebolic addressing of Low Medium Frequency coefficient;

   2) the watermark W (j) that utilizes the Hash function and will embed obtains a two-valued function sequence key (j), key (j)=V (j)  W (j);

   Preserve key (j), will use when extracting watermark below; By key (j) is applied for to the third party as key, to obtain entitlement to former figure;

   Second portion is a watermark extracting: the visual feature vector V ' by two-valued function sequence key (j) and testing image (j) extracts watermark W ' (j);

   3) testing image is carried out the full figure dct transform; In the DCT coefficient, the visual feature vector V ' that goes out testing image according to the symbol extraction of Low Medium Frequency coefficient (j);

   4) utilize Hash function character to extract watermark, W ' (j)=key (j)  V ' (j);

   W (j) and W ' (j) are carried out degree of correlation test, determine the entitlement of image.
2. 2, digital watermark method according to claim 1 is characterized in that: when carrying out the embedding of a plurality of watermarks, and the corresponding two-valued function sequence key (j) of each watermark; When the extraction of watermark, each key (j) extracts a corresponding watermark, with embedding and the extraction that realizes many watermarks.
3. 3, digital watermark method according to claim 1 is characterized in that: when embedding a big watermark, big watermark can be divided into a plurality of little watermarks, embed and extract by multi-watermarking again.

Images