CN104851072A - Robust watermarking method for medical image in cloud environment based on DFT encryption - Google Patents

Abstract

The invention discloses a robust watermarking method for a medical image in a cloud environment based on DFT encryption, belongs to the field of multimedia signal processing, and achieves the embedding of a robust watermark in an encrypted image. The method comprises the following steps: carrying out DFT transformation of an original medical image firstly, achieving the quick encryption of the image through the point multiplication with a two-value chaotic sequence in a transform domain, and carrying out the encryption of a two-value text watermark; secondly carrying out the DFT transformation of the encrypted original image, extracting a characteristic vector of the encrypted image, carrying out exclusive-or operation through the characteristic vector of the encrypted image and the encrypted watermark, embedding the watermark, and obtaining a two-value logic sequence; thirdly extracting and restoring the watermark; finally carrying out DFT calculation of the encrypted image, and decrypting the encrypted image into the original image. The method is the zero watermark technology. Because the watermark is embedded into the encrypted image, the watermark is protected, and the original image is also protected.

Description

A kind of based on DFT encrypted medical image robust watermarking method under cloud environment

Technical field

The invention belongs to field of multimedia signal processing, relate to a kind of medical image digital watermark technology based on DFT conversion, chaotic maps (Logistic Map) and image feature vector, specifically under cloud environment one based on DFT encrypted medical image robust watermarking method.

Background technology

The widespread use of the modern medicine image techniques such as X-ray, CT, MRI, ultrasound wave and PET, makes researchist carry out deep research to the confidentiality of medical image, integrality and authenticity.The development of internet cloud, bringing for medical skill easily simultaneously, also proposes challenge to the safety of data.Under cloud environment, in order to use cloud service not leak data privacy, user needs to be encrypted data, and hope can at ciphertext domain analysis data.Medical image contains the important information of patient, when they carry out storing and transmitting on the internet or otherwise, is easy to face security threat.Therefore, in the urgent need to taking safeguard measure to the medical image of patient.Adopt suitable Information hiding means just can prevent unauthorized user from conducting interviews to relevant information.Some common encryption methods are seen in practice, such as DES, IDEA, AES etc.But, when using these methods to be encrypted, computationally requiring a great deal of time, being also just applicable to the encryption to text data simultaneously.Because the data volume that medical image has is large, redundance is high, the feature that between neighbor, correlativity is strong, makes traditional encryption method and is not suitable for for being encrypted image.Therefore, some new encryption methods are applied to image encryption, such as famous chaos system.The ergodicity had due to chaos system, the feature such as high susceptibility, pseudo-randomness, data mixing to initial value/controling parameters, make it be well suited for for design safety and resume image efficiently.

According to the action scope of cryptographic algorithm, the resume image based on chaos can be divided into spatial domain cryptographic algorithm and frequency domain encryption algorithm.The cipher mode of spatial domain does not generally need the conversion of using from spatial domain to frequency domain, and calculated amount is relatively less, but its local random scrambling effect is not fine.The advantage of frequency domain algorithm is, the change of every bit all can produce certain impact to whole data acquisition in a frequency domain.Relative to spatial-domain algorithm, frequency domain algorithm encryption efficiency is higher.In medical image using personal information as digital watermark embedding in medical image, just can solve the Information hiding problem of patient preferably.Classic method be directly by the watermark embedment after scramble in original image, do like this and can protect watermark information, but still can not eliminate the potential safety hazard that carrier information is revealed when medical image suffers unauthorized access.Just can be addressed this problem well by embed digital watermark in encrypted image, but correlative study yet there are no any report.

Based on this, the present invention is a kind of based on DFT encrypted medical image robust watermarking method under cloud environment.First at DFT transform domain, original medical image is encrypted, then using the important information of patient as watermark signal, is embedded into after scramble in encrypted medical image.So, even if the information of patient is subject to having a mind to or attacking unintentionally, as long as unauthorized user does not know key, just cannot cracks out medical image and watermark, the personal information of patient is really protected.

Summary of the invention

The present invention is a kind of based on DFT encrypted medical image robust watermarking method under cloud environment; by by the proper vector of encrypted medical image, cryptography and zero watermarking combine with technique; compensate for the shortcoming that traditional digital watermark method can not be protected medical image itself; there is very strong robustness and invisibility, the privacy information of patient and the data security of medical image can be protected simultaneously.

To achieve these goals, the present invention is performed such: convert based on full figure DFT, in DFT conversion coefficient, extract the medical image visual feature vector of a resist geometric attacks, and digital watermark and chaos encryption, Hash function and " third party's concept " are combined, achieve resist geometric attacks and the conventional attack of digital watermarking.The method applied in the present invention comprises original image encryption, watermark scramble, watermark embedment, watermark extracting, watermark reduction and encrypted medical image restoring six major part.

Now be described in detail as follows to method of the present invention:

Select a significant two-value text image as the watermark embedding encrypted medical image, be designated as W={w (i, j) | w (i, j)=0,1; .Meanwhile, we choose the tenth an of medical volume data and are designated as I (i, j) as original medical image, and W (i, j) and I (i, j) represents the grey scale pixel value of watermark and original medical image respectively.

Part I: at transform domain to original medical image I (i, j) encryption, generate encrypted image EI (i, j)

1) full figure DFT conversion is carried out to original medical image I (i, j), obtain matrix of coefficients D (i, j);

D(i,j)＝DFT(I(i,j))

2) according to initial value y ₀, application Logistic Map generates chaos sequence L (j);

3) obtain two-dimensional matrix by rising maintenance and operation calculation to L (j), then chaos matrix is passed through symbolic operation, will the number of 0.5 be more than or equal to, assignment is " 1 ", all the other assignment are "-1 ", with obtain two-value chaos Matrix C ' (i, j));

4) by DFT matrix of coefficients D (i, j) of former figure and two-value chaos Matrix C ' (i, j) carry out point multiplication operation, obtains DFT matrix of coefficients ED (i, j) after encryption;

ED(i,j)＝D(i,j).*C'(i,j)

5) DFT inverse transformation is carried out to matrix of coefficients ED (i, j), obtain encrypted medical image EI (i, j);

EI(i,j)＝IDFT(ED(i,j))

Part II: the encryption of watermark

6) two-value chaos matrix is obtained

First according to initial value x ₀generating one dimension chaos sequence X (j), obtaining two-dimensional matrix by rising maintenance and operation calculation; Then, by chaos sequence X (j) by symbolic operation, will be more than or equal to the element assignment of 0.5 for " 1 ", all the other assignment are " 0 ", to obtain two-value chaos Matrix C (i, j).

7) watermark of chaos encryption is obtained

By binary watermarking W ^k(i, j) and two-value chaos Matrix C (i, j) obtain the multi-watermarking EW (i, j) encrypted through XOR;

$EW(i,j)=W\left(i,j\right)\⊕C(i,j)$

Part III: the embedding of watermark EW (i, j)

8) proper vector of encrypted medical image EI (i, j) is extracted

To encrypted medical image EI (i, j) DFT conversion is carried out, obtain DFT matrix of coefficients ED (i, j), choose L before in coefficient individual, by symbolic operation obtain visual feature vector EV (j) of encrypted image=ev (j) | ev (j)=0,1; 1≤j≤L}, L is the number of got DFT conversion coefficient, is expressed as follows:

ED(i,j)＝DFT2(EI(i,j))

EV(j)＝sign(ED(i,j))

9) embed watermark obtain logical key

Watermark EW (i, j) after proper vector EV (j) and encryption is carried out XOR by turn, just by watermark embedment in encrypted image, logical key Key (i, j) can be obtained simultaneously;

$Key(i,j)=EW(i,j)\⊕EV\left(j\right)$

Preserve Key (i, j), will use this extracts watermark during below.By Key (i, j) is applied for third party as key, entitlement and the right to use of original medical image can be obtained, thus reach the object of conservation medicine image;

Part IV: the extraction of watermark

10) encrypted medical image EI'(i, j to be measured) proper vector

DFT conversion is carried out to encrypted medical image to be measured, obtain DFT matrix of coefficients ED'(i, j), choose L before in coefficient individual, the visual feature vector EV'(j of encrypted image to be measured is obtained by symbolic operation)=ev'(j) | ev (j)=0,1≤j≤L}.L is the number of got DFT conversion coefficient, is 32 herein;

ED'(i,j)＝DFT2(EI'(i,j))

EV'(j)＝sign(ED'(i,j))

11) watermark EW'(i, j is extracted)

Proper vector EV'(j by encrypted image to be measured) and logical key Key (i, j) carry out XOR, just extract the watermark EW of encryption " (i, j);

${\mathrm{EW}}^{\′}(i,j)=Key(i,j)\⊕{\mathrm{EV}}^{\′}\left(j\right)$

This algorithm only needs key K ey (i, j) when extracting watermark, and not needing original image to participate in, is a kind of blind watermatking extraction algorithm;

Part V: the reduction of watermark

12) two-value chaos encryption Matrix C (i, j) is obtained

Utilize and the same method of watermark encrypting, obtain identical two-value chaos Matrix C (i, j);

13) encrypted watermark that extracts is reduced

Two-value chaos Matrix C (i, j) and the encrypted watermark EW (i, j) that extracts just are obtained the watermark W'(i, the j that reduce through XOR);

$W^{\′}(i,j)=C(i,j)\⊕EW(i,j)$

By calculate W (i, j) and W'(i, j) related coefficient NC, determine the entitlement of medical image and the watermark information of embedding;

Part VI: the reduction of encryption original medical image EI (i, j)

14) two-value scrambled matrix is generated

Utilize and original image encrypt same method, generate identical two-value chaos Matrix C ' (i, j);

15) the DFT matrix of coefficients of encrypted medical image is obtained

DFT conversion is carried out to encrypted medical image, obtains DFT matrix of coefficients ED'(i, j);

ED(i,j)＝DFT(EI(i,j))

16) the original medical image I (i, j) deciphered is obtained

By the DCT coefficient matrix ED (i of encrypted image, j) and two-value chaos encryption Matrix C ' (i, j) point multiplication operation is carried out, obtain the DFT matrix of coefficients D'(i deciphered, j), again DFT inverse transformation is carried out to it, just can obtain the medical image I (i, j) to be measured deciphering reduction;

D(i,j)＝ED(i,j).*C'(i,j)

I(i,j)＝IDFT(D(i,j))

This algorithm is based on DFT and Logistic chaotic maps, take into account DFT computing velocity fast, precision is high, compatible good advantage and chaos system ergodicity, the feature such as high susceptibility, pseudo-randomness, data mixing to initial value, watermark and medical image are all encrypted, just the technology that watermark information is encrypted is compared relative to traditional, for medical image itself, there is higher security.Medical image, as a class particular image, requires that raw data has integrality.This algorithm, owing to adopting zero watermarking embedded technology, solves traditional watermark embedding technique well and artwork data is revised to the defect caused, ensure that the quality of medical image.Utilize third-party concept, adapt to the practical of network today technology and standardization.

Illustrate from theoretical foundation and experimental data below:

1) discrete Fourier transformation

Two-dimensional discrete Fourier direct transform (DFT) formula is as follows:

$F(u,v)=\underset{x=0}{\overset{M-1}{\Σ}}\underset{y=0}{\overset{N-1}{\Σ}}f(x,y)\·e^{-j2\mathrm{\π}xu/M}{e}^{-j2\mathrm{\π}yv/N}$

u＝0,1,…,M-1；v＝0,1,…,N-1；

Two-dimensional discrete Fourier inverse transformation (IDFT) formula is as follows:

$f(x,y)=\frac{1}{MN}\underset{u=0}{\overset{M-1}{\Σ}}\underset{v=0}{\overset{N-1}{\Σ}}F\left(u,v\right)e^{j2\mathrm{\π}\left(\frac{ux}{M}+\frac{vy}{N}\right)}$

x＝0,1,…,M-1；y＝0,1,…,N-1

Wherein x, y are spatial domain sampled value; U, v are frequency field sampled value, and usual digital picture pixel square formation represents, i.e. M=N.

From formula above, the coefficient symbols of DFT is relevant with the phase place of component.

2)Logistic Map

Chaos is one random motion seemingly, refers to the similar random process occurred in deterministic system.Therefore, had its initial value and parameter, we just can generate this chaos system.Foremost a kind of chaos system is Logistic Map, and it is the Nonlinear Mapping be given by the following formula:

x _k+1＝μx _k(1-x _k)

Wherein, 0≤μ≤4 are called growth parameter, x _k∈ (0,1) is system variable, and k is iterations.The research work of Chaos dynamic system is pointed out, when growth parameter 3.569945≤μ≤4, Logistic Map works in chaos state.Can see that initial value has a slight difference will cause the significant difference of chaos sequence.Therefore, above sequence is a desirable key sequence.Set μ=4 herein, chaos sequence is by different initial value x ₀produce.

3) choosing method of encrypted medical image feature vector

The main cause of current most of medical image watermarking algorithm resist geometric attacks ability is: people by digital watermark embedding in pixel or conversion coefficient, the slight geometric transformation of encrypted medical image, usually cause pixel value or transform coefficient values to have larger change, the watermark of embedding so just can be made under attack very easily.If can find the proper vector of reflection encrypted medical image geometry feature, so when little geometric transformation occurs image, can not there is obvious sudden change in the eigenwert of this image substantially.Through observing the full figure DFT data (Low Medium Frequency) of a large amount of encrypted images, we find when to undertaken by method of the present invention medical image encrypt time, when common geometric transformation is carried out to an encrypted medical image, may be there are some changes in the size of Low Medium Frequency coefficient, but its coefficient symbols remains unchanged substantially.According to human vision property (HVS), low intermediate frequency signal is comparatively large to the visual impact of people, and represent the principal character of image, therefore the Low Medium Frequency coefficient symbols sequence of encrypted medical image selected by us is as visual feature vector.

We choose some experimental datas and are shown in Table 1, and the encrypted image being used as the original medical image of test is Fig. 9, and it is the frequency domain encryption image of sectioning image Fig. 1 of a width brain.In table, the 1st row display is encrypted medical image type under attack, 4th row are to the 8th row, the FF (1 got in DFT matrix of coefficients, 1) ~ FF (1,5), 5x2=10 Low Medium Frequency coefficient (here a plural number, regarding real part and imaginary part two coefficients as) altogether.Wherein coefficient F (1,1) represents the direct current component value of encrypted medical image.As shown in table 1, for conventional attack, these Low Medium Frequency coefficient values remain unchanged substantially, and original medical image value approximately equal; For geometric attack, part coefficient has larger change, but we can find, medical image is when being subject to geometric attack, and the size of part DFT Low Medium Frequency coefficient there occurs change but its symbol does not change substantially.We are by DFT coefficient (plural number regards real part and imaginary part two coefficient values as here), on the occasion of representing with small incidental expenses " 1 ", negative value represents with " 0 ", so for original medical image, FF (1 in DFT matrix of coefficients, 1) ~ FF (1,5) coefficient, corresponding coefficient symbols sequence is: " 1001000000 ".Observe these row can find, no matter conventional attack or this symbol sebolic addressing of geometric attack can keep similar with original encryption image, with the normalized correlation coefficient all comparatively large (normalized correlation coefficient NC value is all 1) of original encryption medical image.

Table 1. encrypted medical image DFT converts medium and low frequency part index variation value (DFT encryption)

Coefficient unit: 1.0e+005

Proper vector facies relationship (128bit) between table 2. different encrypted image

	V1	V2	V3	V4	V5	V6	V7	V8
									V1	1.00	0.14	0.03	0.01	0.01	0.09	-0.22	0.03
V2	0.14	1.00	0.14	-0.01	-0.03	0.14	0.01	0.14
									V3	0.03	0.14	1.00	0.08	0.12	0.03	-0.18	0.13
V4	0.01	-0.01	0.08	1.00	0.17	0.10	0.02	0.07
									V5	0.01	-0.03	0.12	0.17	1.00	0.05	-0.03	0.08
V6	0.09	0.14	0.03	0.10	0.05	1.00	0.03	0.00
									V7	-0.22	0.01	-0.18	0.02	-0.03	0.03	1.00	-0.10
V8	0.03	0.14	0.13	0.07	0.08	0.00	-0.10	1.00

In order to verify the image encryption undertaken by the present invention further, full figure DFT conversion coefficient symbol sebolic addressing after encryption can as of an encrypted image proper vector, we are to different 8 original medical image, see Fig. 1-Fig. 8, the chaos encrypting method carried out based on DFT transform domain proposed by the present invention is encrypted, medical image after encryption, as Fig. 9-Figure 16, carry out full figure DFT conversion, obtain corresponding DFT coefficient F (1,1) ~ F (8,8), calculate the normalized correlation coefficient between encrypted image, result is as shown in table 2.As can be seen from Table 2, between different encrypted medical images, the degree of correlation is less, is less than 0.3.

In sum, we are by the analysis to the full figure DFT coefficient of encrypted image, find to utilize the symbol sebolic addressing of DFT Low Medium Frequency coefficient as the proper vector of encrypted medical medical image.And utilize the proper vector of this encrypted image to carry out embedding and the extraction of zero watermarking.

Accompanying drawing explanation

Fig. 1 is the original medical image standard drawing (mri-1) of test

Fig. 2 is the original medical image standard drawing (mri-2) of test

Fig. 3 is the original medical image standard drawing (mri-3) of test

Fig. 4 is the original medical image standard drawing (engine) of test

Fig. 5 is the original medical image standard drawing (head) of test

Fig. 6 is the original medical image standard drawing (teddy bear) of test

Fig. 7 is the original medical image standard drawing (mri1-1 back1) of test

Fig. 8 is the original medical image standard drawing (mri1-1 back2) of test

Fig. 9 is the image after medical image (mri-1) chaos encryption

Figure 10 is the image after medical image (mri-2) chaos encryption

Figure 11 is the image after medical image (mri-3) chaos encryption

Figure 12 is the image after medical image (engine) chaos encryption

Figure 13 is the image after medical image (head) chaos encryption

Figure 14 is the image after medical image (teddy bear) chaos encryption

Figure 15 is the image after medical image (mri1-1 back1) chaos encryption

Figure 16 is the image after medical image (mri1-1 back2) chaos encryption

Figure 17 is original medical image (mri1-1)

Figure 18 is the image after original medical image (mri1-1) chaos encryption

Figure 19 is the medical image (mri1-1) using correct secret key deciphering

Figure 20 is the medical image (mri1-1) of the secret key deciphering of mistake in

Figure 21 is the original watermark for embedding

Figure 22 is the watermark after Logistic scramble

Figure 23 is the watermark using correct secret key deciphering

Figure 24 is the watermark of the secret key deciphering of mistake in

Figure 25 is the encrypted image after Gaussian noise 3% is attacked

Figure 26 be extract after Gaussian noise 3% watermark picture

Figure 27 is the encrypted image after JPEG compression 20% is attacked

Figure 28 be JPEG compression 20% after extract watermark picture

Figure 29 is medium filtering [3,3], the encrypted image after attacking for 1 time

Figure 30 is medium filtering [3,3], extract after 1 time watermark picture

Figure 31 is the encrypted image after attacking that turns clockwise 2 degree

Figure 32 be turn clockwise 2 degree attack after extract watermark picture

Figure 33 is the encrypted image after zoom factor 0.5 is attacked

Figure 34 be zoom factor 0.5 attack after extract watermark picture

Figure 35 is the encrypted image after vertically moving down (1%) attack

Figure 36 be vertically move down (1%) attack after extract watermark picture

Figure 37 is the encrypted image after Y-direction shears 2% attack

Figure 38 be Y-direction shear 2% attack after extract watermark picture

Embodiment

Below in conjunction with accompanying drawing, the invention will be further described, and Figure 17 is original medical image,

Figure 18 is original medical image by the image after the encryption of this method after DFT encryption;

Figure 19 is the medical image using correct secret key deciphering; Figure 20 is the medical image of the secret key deciphering of mistake in; Select a significant bianry image as original watermark, be designated as: W={w (i, j) | w (i, j)=0,1; 1≤i≤M1,1≤j≤M2}, is shown in Figure 21, and watermark size is 32 × 32.See Figure 22 by the watermark after Logistic Map Chaotic Scrambling, can obviously see that watermark has a very large change, security improves.The original medical image Figure 18 tested after encryption used is the brain three-dimensional imaging of a width after CT scan, choose the tenth sectioning image of this medical volume data, see Figure 17 as original image, be designated as: F={f (i, j) f (i, j) ∈ R; 1≤i≤N1,1≤j≤N2}.Consider the capacity of robustness and disposable embed watermark, we select the 4x4=16 of medium and low frequency plural coefficient to do proper vector (plural number being regarded as real part and imaginary part two coefficients here), then total 16x2=32 Low Medium Frequency coefficient, i.e. L=32.The DFT matrix of coefficients chosen is D (i, j), 1≤i≤4,1≤j≤4.Watermark is extracted and for W ' (i after reduction by watermark extraction algorithm, j) after, calculate W (i again, j) with W ' (i, j) normalized correlation coefficient NC (Normalized Cross Correlation), has judged whether watermark embedment.

The watermark that Figure 23 extracts when being and not adding interference, can see NC=1.00, the existence of watermark obviously be detected.

Figure 24 is the watermark of the secret key deciphering of mistake in.

For checking the robustness of watermarking algorithm of encrypted image, below we judge anti-conventional attack ability and the resist geometric attacks ability of this digital watermark method by specific experiment.

1. the anti-conventional attack ability of watermark:

(1) Gaussian noise is added

The watermarking images that Figure 25 is Gaussian noise intensity when being 3% is visually very fuzzy;

Figure 26 is the watermark of extracting, and NC=0.9388, obviously can detect the existence of watermark.

Table 3 is watermark anti-Gauss detection data when disturbing.Can see from experimental data, when Gaussian noise intensity up to for 25% time, the PSNR of image is down to 9.20dB, and the watermark related coefficient NC=0.82 at this moment extracted, still can detect the existence of watermark.This illustrates the anti-Gaussian noise ability adopting this invention to have.

The data that the anti-Gaussian noise of table 3. is attacked

Noise intensity (%)	1	3	5	10	15	20	25
								PSNR(dB)	20.84	16.43	14.53	12.06	10.62	9.81	9.20
NC	0.96	0.93	0.84	0.62	0.78	0.81	0.82

(2) JPEG compression process

Image compression quality percentage is adopted to carry out JPEG compression as parameter to watermarking images;

Figure 27 to be compression quality be 20% image, there is blocking artifact in this figure;

Figure 28 is the watermark of extracting, NC=0.9056.

Table 4 is the experimental data of the anti-JPEG compression of watermarking images.When compression quality is 2%, the existence of watermark still can be detected, NC=0.90, this illustrates and adopts this invention to have very strong anti-JPEG compressed capability.

The experimental data of the anti-JPEG compression of table 4 watermark

Compression quality (%)	2	4	8	10	20	40	60	80	100
										PSNR(dB)	20.10	20.13	20.19	20.25	20.42	20.52	20.60	20.65	20.67
NC	0.90	0.84	0.84	0.84	0.90	0.90	0.90	0.90	0.90

(3) medium filtering process

Figure 29 is medium filtering parameter is [3x3], the medical image that filter times is 1 time, and image has occurred fuzzy; Figure 30 is the watermark of extracting, and NC=1, Detection results is obvious.

Table 5 is the anti-medium filtering data of watermarking images, it can be seen from the table, when medium filtering parameter is [7x7], when filtering multiplicity is 20, still can record the existence of watermark, NC=0.69.

The anti-medium filtering experimental data of table 5 watermark

2. watermark resist geometric attacks ability:

(1) rotational transform

Figure 31 is the encrypted medical image that watermarking images rotates 2 °, PSNR=21.5776dB, and signal to noise ratio (S/N ratio) is very low; Figure 32 is the watermark of extracting, and the existence of watermark obviously can be detected, NC=0.84444.

Table 6 is watermark anti-rotation attack experimental data.Can see that, when watermarking images rotates 10 °, NC=0.52, still can detect the existence of watermark from table; The resist geometric attacks algorithm that the people such as Pitas propose, in the annulus of watermark embedment DFT amplitude spectrum, can only resist the rotation being not more than 3 degree.

Table 6 watermark anti-rotation attacks experimental data

Rotate the number of degrees (degree)	-1	-2	-3	-4	-5	-10
							PSNR(dB)	25.79	21.57	19.69	18.61	17.95	16.33
NC	0.90	0.84	0.75	0.71	0.65	0.52

(2) scale transformation

Figure 33 to be zoom factor be 0.5 watermarking images, at this moment center image is less than former figure;

Figure 34 is the watermark of extracting, and NC=0.9056, obviously can detect the existence of watermark.

Table 7 is the nonshrink attack experimental data of putting of watermark, as can be seen from Table 7, when zoom factor little to 0.2 time, related coefficient NC=0.67, still can record watermark, illustrates that this invention has stronger anti-zoom capabilities.

Table 7 watermark is nonshrink puts attack experimental data

Zoom factor	0.2	0.5	0.8	1	1.2	2	4	6
									NC	0.67	0.90	0.90	1.00	1.00	1.00	1.00	0.96

(3) translation transformation

Figure 35 is that watermarking images vertically moves down 1%, at this moment PSNR=24.1747dB, and signal to noise ratio (S/N ratio) is very low; Figure 36 is the watermark of extracting, and NC=1, obviously can detect the existence of watermark.Table 8 is the anti-translation transformation experimental datas of watermark.Learning when vertically moving down 10% from table, the existence of watermark still can be detected, therefore this invention having stronger anti-translation capability.

The anti-translation transformation experimental data of table 8 watermark

Vertically move down number percent (%)	1	2	4	6	8	10
							PSNR(dB)	24.17	19.96	16.79	16.07	15.34	14.87
NC	1	0.80	0.80	0.76	0.54	0.49

(4) shearing attack

Figure 37 is that watermarking images shears the situation of 2% by Y direction, and at this moment bottom is relative to original medical image, has been sheared a part; Figure 38 is the watermark of extracting, and NC=0.74585, obviously can detect the existence of watermark.The experimental data that table 9 is the anti-shearing attack of watermark, from table, experimental data is known, and this algorithm has certain anti-shear ability.

The anti-shearing attack experimental data of table 9 watermark (shearing by Y direction)

Shear number percent (%)	1	2	4	6	8	10	20	30
									NC	1.00	0.74	0.74	0.65	0.65	0.59	0.64	0.50

By above description of test, this watermark embedding method has stronger anti-conventional attack and geometric attack ability, and the embedding of watermark does not affect former medical image, is a kind of zero watermarking algorithm.

Claims (1)

1. under cloud environment, one is characterized in that based on DFT encrypted medical image robust watermarking method: convert based on full figure DFT, in DFT conversion coefficient, extracts the medical image visual feature vector of a resist geometric attacks; Again by encryption after watermark embedment in encrypted medical image, not only common digital watermark and chaos encryption, " third party's concept " are combined, achieve anti-geometry and the conventional attack of digital watermarking, and make original image also be provided with good confidentiality; The method applied in the present invention comprises original image encryption, watermark scramble, watermark embedment, watermark extracting, watermark reduction and encrypted image reduction six major part;

Part I: at transform domain to original medical image I (i, j) encryption, generate encrypted image EI (i, j)

1) full figure DFT conversion is carried out to original medical image I (i, j), obtain matrix of coefficients D (i, j);

D(i,j)＝DFT(I(i,j))

2) according to initial value y ₀, application Logistic Map generates chaos sequence L (j);

3) obtain two-dimensional matrix by rising maintenance and operation calculation to L (j), then chaos matrix is passed through symbolic operation, will the number of 0.5 be more than or equal to, assignment is " 1 ", all the other assignment are "-1 ", with obtain two-value chaos Matrix C ' (i, j));

4) by DFT matrix of coefficients D (i, j) of former figure and two-value chaos Matrix C ' (i, j) carry out point multiplication operation, obtains DFT matrix of coefficients ED (i, j) after encryption;

ED(i,j)＝D(i,j).*C'(i,j)

5) DFT inverse transformation is carried out to matrix of coefficients ED (i, j), obtain encrypted medical image EI (i, j);

EI(i,j)＝IDFT(ED(i,j))

Part II: the encryption of watermark

6) two-value chaos matrix is obtained

First according to initial value x ₀generating one dimension chaos sequence X (j), obtaining two-dimensional matrix by rising maintenance and operation calculation; Then, by chaos sequence X (j) by symbolic operation, will be more than or equal to the element assignment of 0.5 for " 1 ", all the other assignment are " 0 ", to obtain two-value chaos Matrix C (i, j);

7) watermark of chaos encryption is obtained

By binary watermarking W ^k(i, j) and two-value chaos Matrix C (i, j) obtain the multi-watermarking EW (i, j) encrypted through XOR;

$EW(i,j)=W(i,j)\⊕C(i,j)$

Part III: the embedding of watermark EW (i, j)

8) proper vector of encrypted medical image EI (i, j) is extracted

To encrypted medical image EI (i, j) DFT conversion is carried out, obtain DFT matrix of coefficients ED (i, j), choose L before in coefficient individual, by symbolic operation obtain visual feature vector EV (j) of encrypted image=ev (j) | ev (j)=0,1; 1≤j≤L}, L is the number of got DFT conversion coefficient, is expressed as follows:

ED(i,j)＝DFT2(EI(i,j))

EV(j)＝sign(ED(i,j))

9) embed watermark obtain logical key

Watermark EW (i, j) after proper vector EV (j) and encryption is carried out XOR by turn, just by watermark embedment in encrypted image, logical key Key (i, j) can be obtained simultaneously;

$Key(i,j)=EW(i,j)\⊕EV\left(j\right)$

Preserve Key (i, j), will use this extracts watermark during below.By Key (i, j) is applied for third party as key, entitlement and the right to use of original medical image can be obtained, thus reach the object of conservation medicine image;

Part IV: the extraction of watermark

10) encrypted medical image EI'(i, j to be measured) proper vector

DFT conversion is carried out to encrypted medical image to be measured, obtain DFT matrix of coefficients ED'(i, j), choose L before in coefficient individual, the visual feature vector EV'(j of encrypted image to be measured is obtained by symbolic operation)=ev'(j) | ev (j)=0,1≤j≤L}.L is the number of got DFT conversion coefficient, is 32 herein;

ED'(i,j)＝DFT2(EI'(i,j))

EV'(j)＝sign(ED'(i,j))

11) watermark EW'(i, j is extracted)

Proper vector EV'(j by encrypted image to be measured) and logical key Key (i, j) carry out XOR, just extract the watermark EW of encryption " (i, j);

${\mathrm{EW}}^{\′}(i,j)=Key(i,j)\⊕{\mathrm{EV}}^{\′}\left(j\right)$

This algorithm only needs key K ey (i, j) when extracting watermark, and not needing original image to participate in, is a kind of blind watermatking extraction algorithm;

Part V: the reduction of watermark

12) two-value chaos encryption Matrix C (i, j) is obtained

Utilize and the same method of watermark encrypting, obtain identical two-value chaos Matrix C (i, j);

13) encrypted watermark that extracts is reduced

Two-value chaos Matrix C (i, j) and the encrypted watermark EW (i, j) that extracts just are obtained the watermark W'(i, the j that reduce through XOR);

$W^{\′}(i,j)=C(i,j)\⊕EW(i,j)$

By calculate W (i, j) and W'(i, j) related coefficient NC, determine the entitlement of medical image and the watermark information of embedding;

Part VI: the reduction of encryption original medical image EI (i, j)

14) two-value scrambled matrix is generated

Utilize and original image encrypt same method, generate identical two-value chaos Matrix C ' (i, j);

15) the DFT matrix of coefficients of encrypted medical image is obtained;

DFT conversion is carried out to encrypted medical image, obtains DFT matrix of coefficients ED'(i, j);

ED(i,j)＝DFT(EI(i,j))

16) the original medical image I (i, j) deciphered is obtained

By the DCT coefficient matrix ED (i of encrypted image, j) and two-value chaos encryption Matrix C ' (i, j) point multiplication operation is carried out, obtain the DFT matrix of coefficients D'(i deciphered, j), again DFT inverse transformation is carried out to it, just can obtain the medical image I (i, j) to be measured deciphering reduction;

D(i,j)＝ED(i,j).*C'(i,j)

I(i,j)＝IDFT(D(i,j))

This algorithm is based on DFT and Logistic chaotic maps, take into account DFT computing velocity fast, precision is high, compatible good advantage and chaos system ergodicity, the feature such as high susceptibility, pseudo-randomness, data mixing to initial value, watermark and medical image are all encrypted, just the technology that watermark information is encrypted is compared relative to traditional, for medical image itself, there is higher security.