CN104217389B - A kind of method and apparatus based on the image watermark insertion, the extraction that improve Arnold conversion - Google Patents
A kind of method and apparatus based on the image watermark insertion, the extraction that improve Arnold conversion Download PDFInfo
- Publication number
- CN104217389B CN104217389B CN201410029947.5A CN201410029947A CN104217389B CN 104217389 B CN104217389 B CN 104217389B CN 201410029947 A CN201410029947 A CN 201410029947A CN 104217389 B CN104217389 B CN 104217389B
- Authority
- CN
- China
- Prior art keywords
- mtd
- watermark
- mtr
- mrow
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Abstract
The present invention relates to a kind of based on the image watermark insertion for improving Arnold conversion, the method and apparatus extracted, embedding grammar includes scramble step：Using Arnold conversion is improved to watermarking images W processing, the embedded location in carrier image I is obtained；Training step：In embedded coordinate position, k sample point is chosen, k sample point is trained by Smooth Support Vector Machines；Embedded step；Improve Arnold conversion and add parameter beta, be expanded to the coordinate value after original watermark scramble in the space of carrier image in proportion in embedded watermark, narrow down to some of carrier image coordinate value in proportion when extracting watermark in the space of original watermark.
Description
Technical field
The present invention relates to the digital image watermarking technology in a kind of information security field, it is embedded in more particularly, to image watermark
With extracting method.
Background technology
Digital figure watermark is hidden in numeral as effective means of supplementing out economy of conventional encryption methods using data embedding method
In image product, to prove ownership of the creator to its works, and the foundation illegally encroached right as identification, prosecution, simultaneously
Ensure the complete reliability of digital information by the detection to watermark and analysis, so that as intellectual property protection and digital many matchmakers
The false proof effective means of body, causes the great attention of people in recent years, has also turned into the focus that international academic community is studied.
Image watermark will play due effect, it is necessary to possess two fundamentals of robustness and imperceptible.Watermark robustness is
Refer to Digital Media after the signal transacting by routine or external attack, embedded image watermark, which still has, preferably may be used
Detection property.Watermark imperceptible refers to that the insertion of watermark can not have influence on the visual quality of original figure media.
Image watermark can be divided into copyright protection watermark, bill anti-counterfeit watermark by purposes, distort prompting watermark and hidden identification
Watermark.It can be divided into blind watermatking and plaintext watermark by extraction process.It can be divided into multipurpose watermarking and fragile watermarking by attacking ability,
Wherein multipurpose watermarking is mainly used in digital copyright protection, and change of the fragile watermarking requirement to signal is sensitive, mainly
Applied to integrity protection.Watermarking algorithm can be divided into according to watermark embedded location by two classes：Based on transform-domain algorithm and
Based on spatial-domain algorithm.With widely using for JPEG compression and JPEG2000, up to the present, there are many based on transform domain
Watermarking algorithm.According to the difference for using conversion, transform domain watermarking algorithm can be divided into following several classes：Based on dct transform
Watermarking algorithm, the watermarking algorithm based on wavelet transformation, the Robust Digital Watermarking Algorithm based on DFT transform.But these
Algorithm comparison is complicated, and, it is necessary to consider the null tone domain conversion process of complexity, efficiency is low, can embedding information amount it is less.Space area image
Digital watermark is because its algorithm is simple, fireballing advantage and turn into new study hotspot, it is by directly changing original image
Pixel value reaches the purpose of embedded watermark, but space watermark algorithm classical at present is highly susceptible to compression of images conversion etc.
The interference of common image procossing, substantially can not be to water after the basic handlings such as geometry rotation, compression are carried out to image
Print is correctly extracted, and experiment simulation shows that the attack tolerant of such algorithm is not strong, and robustness is relatively low.But with nerve net
The introducing of the machine learning methods such as network, SVMs, the insertion of watermark and detection process can make full use of one in image
A little physical features, the watermark insertion and robust detection effect that can so cause spatial domain obtains certain raising.Although machine
Device learns and the combination of various image areas conversion has preferable performance for the insertion and extraction of specific watermark, but still deposits
In many problems.Such as image watermark method based on SVMs etc. typically can not all realize Blind extracting, the guarantor of watermark
Also there is certain hidden danger in close property；The computation complexity as the insertion based on empty frequency-domain transform with extracting method is higher, and resistance is attacked
The ability of hitting also has to be strengthened etc..Sum up and still suffer from some following subject matters：1. SVMs is mainly used to determine
Optimum position and suitable strength in watermark insertion digital picture, and Arnold conversion only plays scramble in watermark insertion,
Application study is relatively simple.2. the SVMs scheme proposed at present is essentially all the SVMs of normative reference,
The speed and precision of sample training are not very high, and this causes the watermarking images distortion finally extracted than more serious.3. watermark system
The robustness of system is to assess watermaking system to carry processed conventionally ability, and this is particularly important for watermark；Existing digital picture water
Notice is placed on confrontation normal signal processing by print detection method（Such as lossy compression method, LPF, noise jamming）Research
On, and rotate, scale, translating, the resistance effect of geometric attack such as ranks are removed, shearing is not fine；SVMs
It is combined with transform domain, although the robustness of watermark detection can be effectively improved, but these algorithms are to attacks such as shearing, rotations
Robustness still has several drawbacks.4. blind Detecting necessarily requires the insertion of watermark and extraction algorithm well to balance and can not feel
Intellectual and robustness, are introduced after SVMs, although existing many Image Watermarking Technique robusts based on SVMs
Property be improved, but sentience declines, therefore does not typically possess blind Detecting characteristic, and this is also a weight urgently to be resolved hurrily
Want problem.5. existing watermark insertion and extraction algorithm based on Arnold conversion all only possess substance key feature, that is, pass through
A kind of characteristic information of digital media products itself constructs a watermark keys, it is clear that this feature, which is existed, is easily attacked
With the limitation cracked, digital media products by it is a certain or it is several gang up against after, the extraction that can increase image watermark is difficult
Degree；When substance key is cracked, disabled user just can delete or distort real image watermark embedded in product, make its heavy
The new state for returning to no copyright protection, this is by the interests of each side such as serious infringement copyright owner.Liu Fang, Jia Cheng, Yuan Zheng write
Write《A kind of binary image watermarking algorithm based on Arnold conversion》（Computer application, 2008,28 (6):1404-1406）Carry
Arrive using a kind of binary image watermarking insertion based on Arnold conversion and extracting method, with reference to Arnold scrambling algorithms, led to
Cross in flipped image and meet the pixel of Vision Constraints condition to reach the purpose of addition watermark.Experiment shows that the algorithm not only improves
The invisibility of watermark, improves the embedding capacity of watermark, and realize the Blind extracting of watermark.But there is also some not
Foot, such as parameter is less during Arnold space field transformations, causes the key of image very little, security is not high, Arnold
Scramble image is also only played a part of in conversion, and the algorithm is to the repellence under the conventional attack of image, especially geometric attack
It is poor, without preferably balance watermark invisibility and robustness etc..So the insertion of every kind of image watermark and extracting method all without
Method is provided simultaneously with very high not sentience and robustness, and improves security and need to increase the quantity of key, and meeting simultaneously
Improve computation complexity.Therefore according to image space characteristic of field find watermark invisibility and robustness preferably, security more
High embedded mobile GIS is the important research content of image watermark.Document《Watermarkingschemebasedonsupport
vectormachineforcolourimages》（FuY,ShenR,LuH,ElectronicsLetters,2004,40(16):
986-987）Indicate that the difficult point for being embedded in watermark using image space characteristic of field is essentially consisted in：How number of keys is effectively increased,
To improve the security of watermark；How fast and effeciently training sample, can remember local pixel after watermark experience various attacks
Relation between point, so as to realize the correct detection to watermark；The pixel value of embedded point can not change too much, otherwise can influence to carry
The not sentience of body image, but must be easy to find the change of embedded point pixel when extracting watermark.
What Fu Yonggang write《Watermarking algorithm based on broad sense Arnold conversion and SVMs》（Collects The American University is learned
Report (natural science edition), 2011,16 (1):65-70）Disclose a kind of design Arnold conversion and the image of SVMs
Watermark insertion and extracting method, but this method uses standard SVMs, training sample speed is slow, precision is low；Train sample
This is more complicated, and data volume is big；Embedded mobile GIS number of keys is few, and security is not high；It is not good using broad sense Arnold conversion performance.
The content of the invention
It is an object of the invention to provide a kind of insertion of image watermark, the method and apparatus extracted, to solve broad sense
The defect of Arnold conversion；It is perfect by further extending, additionally it is possible to solve existing insertion not sentience is poor and Shandong
The problem of rod is weak, and the problem of training sample speed is slow, precision is low；The problem of training sample is more complicated；Number of keys is few
The problem of；The problem of blind Detecting.
To achieve the above object, the solution of the present invention includes：
It is a kind of as follows based on the Image Watermarking for improving Arnold conversion, including step：
1）Scramble step：Using Arnold conversion is improved to watermarking images W processing, obtain in carrier image I
Embedded location；Arnold conversion is improved according to equation below
N iteration is carried out, with the position coordinates (x of watermarking images0,y0) as initial value, wherein 1≤x0≤ M, 1≤y0≤ K,
Obtain the embedded location (x of correspondence watermark bitn,yn), wherein 1≤xn≤ N, 1≤yn≤ N, Wherein floor
It is lower bracket function, M is the exponent number of watermarking images matrix now, and N is the exponent number of initial carrier image, and a, b and n are positive integer
And It is the bulk composition of iteration；Finally give M × K position coordinates
(xi,yi), wherein i=1,2 ..., M × K.
2）Training step：In embedded coordinate position, k sample point is chosen, by Smooth Support Vector Machines to k sample
This point is trained, and k is setting value；
3）Embedded step：The pixel value of embedded watermarking images in initial carrier image.
Step 2）The SVMs used is obscures Smooth Support Vector Machines, on the basis of standard SVMs,
Introduce fuzzy membership μiWith obfuscation training sample；Introduce a nonlinear function Ф (x, y) and sample point is mapped to higher-dimension
Feature space；It is last to carry out linear regression in high-dimensional feature space, so as to obtain in former Space Nonlinear regression effect.
Step 2）The training object of middle selection is the single order above square of sampled pixel point.
Watermark extracting method, comprises the following steps：
1）Training step：Image I' of the supporting vector machine model trained with watermark telescopiny to embedded watermark
It is trained, determines watermark embedded location；
2）Arnold inverse transformation steps：The coordinate value of watermarking images is obtained by Arnold inverse transformations；
3）Pixel value extraction step：According to the qualified insertion rule and the quantization step d amounts of pixel value during embedded watermark
Change and extract watermark pixel value；
4）Watermark recovery step：Watermarking images, weight are recovered according to the coordinate value of watermarking images and corresponding pixel value
Group original watermark image W.
A kind of image watermark flush mounting based on improvement Arnold conversion, including：
1）Scramble module：Using Arnold conversion is improved to watermarking images W processing, obtain in carrier image I
Embedded location；Arnold conversion is improved according to equation below
N iteration is carried out, with the position coordinates (x of watermarking images0,y0) as initial value, wherein 1≤x0≤ M, 1≤y0≤ K,
Obtain the embedded location (x of correspondence watermark bitn,yn), wherein 1≤xn≤ N, 1≤yn≤ N, Wherein floor
It is lower bracket function, M is the exponent number of watermarking images matrix now, and N is the exponent number of initial carrier image, and a, b and n are positive integer
And It is the bulk composition of iteration；Finally give M × K position coordinates
(xi,yi), wherein i=1,2 ..., M × K；
2）Training module：In embedded coordinate position, k sample point is chosen, by Smooth Support Vector Machines to k sample
This point is trained, and k is setting value；
3）Embedded module：The pixel value of embedded watermarking images in initial carrier image.
Module 2）The SVMs used is obscures Smooth Support Vector Machines, on the basis of standard SVMs,
Introduce fuzzy membership μiTo each training sample；Introduce a nonlinear function Ф (x, y) and sample point is mapped to higher-dimension spy
Levy space；It is last to carry out linear regression in high-dimensional feature space, so as to obtain in former Space Nonlinear regression effect.
Module 2）The training object of middle selection is the single order above square of sampled pixel point.
Watermark extraction apparatus, including：
1）Training module：Image I' of the supporting vector machine model trained with watermark telescopiny to embedded watermark
It is trained, determines watermark embedded location；
2）Arnold inverse transform blocks：The coordinate value of watermarking images is obtained by Arnold inverse transformations；
3）Pixel value extraction module：According to the qualified insertion rule and the quantization step d amounts of pixel value during embedded watermark
Change and extract watermark pixel value；
4）Watermark retrieving module：Watermarking images, weight are recovered according to the coordinate value of watermarking images and corresponding pixel value
Group original watermark image W.
Different from broad sense Arnold conversion, modified Arnold conversion of the invention is directed not only to independent parameter a, b and changed
For frequency n, oneself set in embedded watermark for copyright owner, in addition to β, in embedded watermark by original watermark scramble
Coordinate value afterwards is expanded in the space of carrier image in proportion, when extracting watermark that some of carrier image coordinate value is same
Scale smaller is into the space of original watermark.
Further, using qualified insertion, compared with traditional embedding method, this rule changes the picture of initial carrier image
The amplitude of element value is smaller, and maximum change amount is d, and the not sentience after watermark insertion can be realized by setting d values.Press
Rule can calculate M × K watermark embedded coordinate position (x in carrier image successively like thisn,yn) pixel value.By what is calculated
M × K watermark embedded coordinate position (xn,yn) pixel value original pixel value, i.e. I (x are replaced in carrier image successivelyn,
yn)→I′(xn,yn)|n=1,2,…,M×K.It can thus obtain being embedded in the digital carrier image after watermark, image size is still N × N,
Simply trickle change occurs for the pixel value of base point.
Further, the present invention is using fuzzy Smooth Support Vector Machines（FSSVM）It is trained, in standard SVMs
On the basis of, in order to improve the efficiency and precision of prediction, problem is converted into unconstrained optimization problem, and combine fuzzy mathematics
Risk function is converted into fuzzy antithesis extreme-value problem by concept, optimization object function, so as to be effectively reduced carrier image prediction
The error of pixel value and actual pixel value.
Further, the thought of the first moment and second moment in probability statistics is quoted in the selection of characteristic value, in different realities
Apply in mode, the species of characteristic value, such as third moment etc. can also be increased, this is conducive to improving the precision predicted, makes training number
It is simpler according to structure set, represent more information with smaller memory data output.
Further, a, b, n and d, the present invention have four keys, improve the security of watermark.
In watermark extracting, first with it is embedded when SVMs find out pixel value changes very big pixel, i.e. watermark
Embedded point, then recover with Arnold inverse transformations original watermarking images pixel point coordinates, the watermark extracting gone out by our backsteppings
Rule recovers the pixel value of its corresponding pixel points；So as to which watermarking images are approximately intactly recovered.
Brief description of the drawings
Fig. 1 is the insertion of the present invention with extracting flow chart；
Fig. 2-1 is initial carrier image；
Fig. 2-2 is original watermark image；
Fig. 2-3 is the carrier image after embedded watermark；
The watermarking images that Fig. 3 is extracted when being without attack；
Fig. 4-1 is the carrier image containing watermark after (+75) that brightens；
Fig. 4-2 is the watermarking images for extracting Fig. 4-1；
Fig. 4-3 is the carrier image containing watermark after dimmed (- 50)；
Fig. 4-4 is the watermarking images for extracting Fig. 4-3；
Fig. 5-1 is the carrier image containing watermark after histogram equalization；
Fig. 5-2 is the watermarking images for extracting Fig. 5-1；
Fig. 5-3 is the histogram after image equilibration；
Fig. 6-1 is to add the carrier image containing watermark after Gaussian noise (μ=0 and σ=0.02)；
Fig. 6-2 is the watermarking images for extracting Fig. 6-1；
Fig. 7-1 is the carrier image containing watermark after medium filtering (9 × 9)；
Fig. 7-2 is the watermarking images for extracting Fig. 7-1；
Fig. 8-1 is the carrier image containing watermark after JPEG compression 10%；
Fig. 8-2 is the watermarking images for extracting Fig. 8-1；
Fig. 9-1 is the carrier image containing watermark behind geometry cutting left side 100 × 300；
Fig. 9-2 is the watermarking images for extracting Fig. 9-1；
Figure 10-1 is the carrier image containing watermark after geometry rotation 10o；
Figure 10-2 is the watermarking images for extracting Figure 10-1.
Embodiment
The present invention will be further described in detail below in conjunction with the accompanying drawings.
The watermark embedding method of the present invention mainly includes：Watermark scramble, sample training and the step of insertion, inventive point are main
Be to find watermark embedded location using improved Arnold conversion and recover original watermark position, add number of keys with
The degree of safety of watermark is improved, and makes full use of the Chaotic Scrambling characteristic of Arnold conversion, watermarking images are evenly distributed by realization
Into host image.And sample training and insertion can be carried out using traditional approach.
Specifically, following present a kind of preferred embodiment, not only scramble step is using the improved of the present invention
Arnold conversion, and sample training step and Embedded step also improved, and this two step corresponds to FSSVM and quantization respectively
It is embedded.FSSVM good small sample Training Capability so that watermarking images remain to remember local pixel point after experience various attacks
Between relation, so as to realize the correct detection to watermark.It is of the invention fully to combine the characteristics of spatial domain and machine learning and excellent
Change its algorithm, obtain image watermark insertion and extracting method that the anti-normal image with excellent robust performance is attacked, very well
Ground balances the robustness and the not contradiction between sentience of image watermark, realizes the blind Detecting of watermark.
This scramble introduced below, training, embedded Arnold conversion, FSSVM, the qualified insertion of being respectively adopted are preferable to carry out
The watermark embedded mode of example, telescopiny is divided with step（1）, step（2）, step（3）, step（4）Form.
Step（1）Input picture size is N × N digital carrier image I, is used as the initial carrier figure of watermark to be embedded
Picture, then the bianry image W that input picture size is M × K, as watermarking images to be embedded, it is designated as I={ I (i, j), 1 respectively
≤ i≤N, 1≤j≤N }, W={ w (i, j), 1≤i≤M, 1≤j≤K }, wherein M, K are respectively the height and width of bianry image, I (i,
J) it is pixel value of the initial carrier image in (i, j) position, w (i, j) is pixel value of the watermarking images in (i, j) position.
Step（2）The picture element matrix of bianry image is expanded into square formation so that M=K, to meet the condition of Arnold conversion：
If M<K, then w (i, j)=1, wherein M<i≤K；
If M>K, then w (i, j)=1, wherein K<j≤M.
Now M=K, using Arnold conversion algorithm is improved, n Arnold iterated transform is done to binary marking pattern W,
I.e. with the position coordinates (x of watermarking images0,y0) as initial value, wherein 1≤x0≤ M, 1≤y0≤ K, according to equation below
I.e.
N iteration is carried out, the embedded location (x of correspondence watermark bit is obtainedn,yn), wherein 1≤xn≤ N, 1≤yn≤ N,Wherein floor is lower bracket function, and M is the exponent number of watermarking images matrix now, and N is initial carrier image
Exponent number, a, b and n be positive integer and Be iteration main body into
Point；Finally give M × K position coordinates (xi,yi), wherein i=1,2 ..., M × K.
In above-mentioned transform, independent parameter a, b and iterations n are set by copyright owner oneself, are joined as three keys
Number can be for recovery watermark signal.The general values between 1 to 20 of n, because n too conference influence program operation speeds, are calculated
Complexity strengthens, and Arnold conversion has periodically, i.e. n has periodically.β is (x0,y0) terminate in n conversion iteration
The multiplication factor of numerical value afterwards, this causes (x when being for last embedded watermarkn,yn) can adapt to whole carrier image space it is big
It is small, and be unlikely to concentrate on a certain square of image, i.e., the uniform scramble of all embedded locations is dispersed in initial carrier image.
Step（3）M × K position coordinates is embedded into the position in carrier image, i.e. watermarking images as watermark scramble
Position coordinates (x0,y0) pixel be embedded into the (x of carrier imagen,yn) position, obtain M × K embedded location (xi,yi),
Wherein i=1,2 ..., M × K, therefrom randomly choose k position coordinates (xi,yi) carry out FSSVM training, wherein i=1,2 ...,
K。
Affiliated step（3）Middle FSSVM specific training process is as follows：
Step（3.1）From M × K embedded location coordinate (xi,yi) k are randomly selected in (i=1,2 ..., M × K), it is denoted as
(xi,yi), wherein i=1,2 ..., K, pixel value of the correspondence in initial carrier image I are I (xi,yi), wherein i=1,2 ..., k.
Step（3.2）For each selected watermark reference or embedded location (xi,yi), in initial carrier image I,
With position coordinates (xi,yi) centered on, the image block that a size is 3 × 3 is chosen, k image block is so obtained altogether.
Step（3.3）For each position coordinates (xi,yi), the feature of its correspondence image block is calculated in carrier image I
Value, i.e., except central point (xi,yi) outside pixel average
With except central point (xi,yi) outside pixel variance
Such one is obtained k groups characteristic value { X (xi,yi),D(xi,yi)}|i=1,2,…,k。
Step（3.4）With each characteristic vector { X (xi,yi),D(xi,yi) it is training dataset, corresponding original load
Body image pixel value I (xi,yi) it is the desired value trained, k is constituted to training sample set { X (xi,yi),D(xi,yi)→I(xi,
yi)}|i=1,2,…,k；Training to this progress FSSVM：It is firstly introduced into fuzzy membership μiTo each training sample, obfuscation input
Sample set { X (xi,yi),D(xi,yi)→I(xi,yi), 0≤μi≤1；Secondly a nonlinear function Ф (x, y) is introduced by sample
This point is mapped to high-dimensional feature space；It is last to carry out linear regression in high-dimensional feature space, so as to obtain in former Space Nonlinear
Regression effect, its regression function f is represented by
Formula Kernel Function mainly uses Gaussian radial basis function（RBF）：
And K（X, xj)=Φ (x) Φ (xi),
X, y ∈ R in formulanRepresent input vector { X (xi,yi),D(xi,yi), αiWithFor the weight system obtained after training
Number, β ∈ R are deviation, represent inner product operation.
Step（3.5）Wherein parameter alphai、Determination with β uses structuring least risk principle, recurrence that will be original
Equation solution is converted into solution Unconstrained optimization problem, and object function is
In formula, λ is regulation parameter, C>0 be constant be used for decision model complexity and
The compromise degree of empiric risk.
FSSVM operation principle：, will in order to improve the efficiency and precision of prediction on the basis of standard SVMs
Problem is converted into unconstrained optimization problem, and combines the concept of fuzzy mathematics, and risk function is converted into mould by optimization object function
Antithesis extreme-value problem is pasted, so as to be effectively reduced carrier image predicted pixel values and the error of actual pixel value.
Step（3.6）Fuzzy membership μiBy sample input data set { X (xi,yi),D(xi,yi) and desired value I (xi,
yi) relation determine；First to all characteristic value collections { X (xi,yi),D(xi,yi) fuzzy C-means clustering is done, it is divided into two classes and looks for
Chu Lianggelei centers, then calculate each characteristic vector { X (xi,yi),D(xi,yi) to correspondence class center apart from di, then degree of membership
It is expressed asWherein dmaxRepresent characteristic vector { X (xi,yi),D(xi,yi) where characteristic point in class to class center
Ultimate range.
Step（4）Find the position coordinates (x that watermarking images are embedded into carrier imagen,yn) after, continue to determine to carry original
The pixel value I ' (x of embedded watermark location in body imagen,yn), mainly using quantizing rule（D=10 are the quantization step of pixel value
AndRound is round function）Carry out the insertion of watermark.
If watermark pixel value w (x0,y0)=1 and k=2m+1, wherein m ∈ N, i.e. k are odd numbers, then
If watermark pixel value w (x0,y0)=0 and k=2m+1, wherein m ∈ N, i.e. k are odd numbers, then
If watermark pixel value w (x0,y0)=1 and k=2m, wherein m ∈ N, i.e. k are even numbers, then
If watermark pixel value w (x0,y0)=0 and k=2m, wherein m ∈ N, i.e. k are even numbers, then
Compared with traditional embedding method, the amplitude for the pixel value that this rule changes initial carrier image is smaller, maximum
Knots modification is 10, it is achieved thereby that the not sentience after watermark insertion.M in carrier image can be calculated successively according to this rule
× K watermark embedded coordinate position (xn,yn) pixel value.
Step（5）By the M × K watermark embedded coordinate position (x calculatedn,yn) pixel value successively in carrier image
Replace original pixel value, i.e. I (xn,yn)→I′(xn,yn)|n=1,2,…,M×K.It can thus obtain being embedded in the numeral load after watermark
Body image, image size is still N × N, and simply trickle change occurs for the pixel value of base point.
Above-mentioned watermark embedding method is directed to, extraction process is as follows, also with step（1）, step（2）, step（3）Form
Divided：
Watermark extracting comprises the following steps：
Step（1）Input the digital carrier image I' of binary bitmap to be extracted, image size is N × N, I'(i, j)
Expression is embedded with pixel value of the carrier image in (i, j) position of watermark, wherein 1≤i≤N, 1≤j≤N.
Step（2）The FSSVM models trained with watermark telescopiny are trained to digital picture I'：First
Calculate each pixel (x for the carrier image I' for being embedded with watermarki,yi) characteristic value, i.e., except central point (xi,yi) outside 8 points
Pixel average
With except central point (xi,yi) outside 8 points pixel variance
Such one is obtained N × N groups characteristic value { X (xi,yi),D(xi,yi)}|i=1,2,…,N×N；With this N × N group characteristic value
{X(xi,yi),D(xi,yi)}|i=1,2,…,N×NFor the input data set x and y of FSSVM model measurement samples, model is substituted into, is passed through
Anticipation functionCalculating, wherein parameter alphai、With β via in watermark telescopiny
Training is drawn, it is possible to obtain being embedded with the predicted pixel values I' of carrier image N × N number of position of watermark0(i,j)。
The thought of the first moment and second moment in probability statistics is quoted in the selection of this characteristic value, can also increase feature in fact
Species of value, such as third moment etc., this is conducive to improving the precision predicted.Wherein n ranks square can outline for
In formulaThe average value of the pixel in selected image block in addition to central point is represented, E represents to solve average.
Step（3）The carrier image I' of watermark pixel value I'(i is would be embedded with, j) with the prediction pixel of its N × N number of position
Value I'0(i, j) is contrasted, calculate the matrix of differences d of the two=| I'(i, j)-I'0(i, j) |, then its each element is sorted, taken
Go out the corresponding preceding M × K pixel diminished successively of d, these point be watermark be embedded in carrier image position coordinates (i,
J), wherein 1≤i≤N, 1≤j≤N.Arnold is carried out with Arnold conversion is improved to this M × K pixel coordinate value (i, j)
Inverse transformation, i.e., using pixel coordinate value (i, j) as initial value, wherein 1≤i≤N, 1≤j≤N, according to equation below
I.e.
Carry out n iteration, during each iteration, x0Assignment in, y0Assignment in, (i, j) correspondence is obtained in watermarking images
In position coordinates (x0,y0), wherein 1≤x0≤ M, 1≤y0≤ K,- 1 is finding the inverse matrix, and M=K is watermark
The exponent number of image array, N be carrier image exponent number, a, b and n be positive integer and.Finally give M × K
Individual position coordinates (xi,yi), wherein i=1,2 ..., M × K, the coordinate value of these coordinates exactly original watermark image.
The value of independent parameter a, b and iterations n in transform oneself setting that is copyright owner in embedded watermark, β and
Implication during embedded watermark is the same, is now simply in order to some of carrier image coordinate value is narrowed down into initial condition in proportion
In the space of print.
Step（4）The position coordinates (i, j) that watermark is embedded into carrier image is determined, and calculates its correspondence in watermark
Position coordinates (x in image0,y0), it is further continued for calculating its pixel value of the correspondence in watermarking images, i.e., when according to embedded watermark
The pixel value of carrier image changes rule, derives that the pixel of now watermark extracting recovers rule, ifd
=10 still quantify segmentation distance for pixel, I'(i, j) for the carrier image after embedded watermark position (i, j) place pixel value,
Floor is a lower bracket function.
If λ=2m+1, wherein m ∈ N, i.e. λ are odd numbers, then the pixel point value of w (i, j)=1, i.e. watermark is 1；
If λ=2m, wherein m ∈ N, i.e. λ are even numbers, then the pixel value of w (i, j)=0, i.e. watermark is 0, simply position now
It is not also the position coordinates corresponded in watermark to put coordinate (i, j).
The recovery rule of this pixel value just embodies the advantage of binary bitmap, because its pixel value non-zero i.e. 1, institute
Only it is the pixel that can determine whether watermark by the quantized result of the carrier image pixel value containing watermark so that its pixel feature need not be relied on
Value.
Step（5）The w (i, j) that M × K pixel coordinate value (i, the j) point selected according to being sorted in carrier image is calculated
Value, then can recover the pixel value of raw water watermark image, then each position coordinates (i, j) is replaced as after Arnold inverse transformations
(x0,y0), i.e. w (i, j) → w (x0,y0), the coordinate value of such watermarking images and corresponding pixel value are determined, according to
The result of the two can recombinate out raw water watermark image W.
Describe watermark insertion and extraction process, below using the typical test image Simulation results of two width and analysis as
Example, image watermark insertion and the feasibility and validity of extracting method of embodiment proposition is better described.
Experimental verification is in PC（Winows8, Intel (R) Core (TM) i5-3337UCPU1.80GHz,
4.0GBMemory）Upper use MATLABR2013a software programmings realize that the original digital image I of image watermark to be embedded is selected
Uint8 Lena gray level images, image size is 300 × 300, as shown in Fig. 2-1；Real image watermark W to be embedded is selected
One binary sequence school badge image, image size is 150 × 150, as shown in Fig. 2-2.
In embedded watermark, determine to be embedded into the position in carrier image, key parameter a by improvement Arnold conversion first
=1, b=2 and n=1, then determine the pixel value of embedded location；And when extracting watermark, embedded location is found out by FSSVM first, then it is extensive
The pixel value of multiple embedded location, finally gains raw water watermark image by embedded location through Arnold contravariant.Both mutually auxiliary phases of process
Into.
Pass through general population（Age distribution is in the right side of fifty, normal visual acuity）Naked eyes the watermark signal of extraction is led
See and distinguish, and the bit error rate that can be also printed using the watermark extracted with raw water（BER）It is objective that the watermark that index is come to extraction is carried out
Evaluate, BER illustrates that the robustness of watermaking system is higher, anti-attack ability is stronger, its BER is expressed as follows closer to 0
Wherein M=35, K=35, w (i, j) and w'(i, j) is respectively the picture of original watermark on correspondence position with extracting watermark
Element value, ⊕ represents the XOR that step-by-step is carried out.
The quality and perceptual performance of digital picture after embedded real image watermark use Y-PSNR（PSNR）To enter
Row is judged, and it represents damaged condition of the embedded watermark information to carrier quality, and PSNR is bigger, and damaged condition is smaller, its PSNR table
Show as follows
Wherein m=300, n=300, I (i, j) and I'(i, j) is respectively initial carrier image and the carrier image added with watermark
The pixel value of each point.
The objective evaluation of image watermark testing result also can use normalizated correlation coefficient（NC）, water is embedded in by carrier image
Change to evaluate the degree of approximation of watermark before and after print, similarity NC is bigger, illustrates that the robustness of watermark is higher, its NC is represented such as
Under
Fig. 2-3 is the Lena digital pictures after the method insertion actual watermark image W according to the present invention.Can from Fig. 2-3
To see, any change does not occur for the Lena digital picture qualities after embedded watermark, and PSNR is very high, reaches
36.3697dB, it is consistent with the original Lena digital pictures shown in Fig. 2-1, fully meet the requirement of watermark imperceptibility.Figure
3 be the watermarking images extracted according to the method for the present invention, is as a result shown, the Lena after insertion actual watermark shown in Fig. 2-3
Digital picture almost can nondestructively extract embedded real image watermark when not handled by any attack, and NC=
0.9785, closely 1, BER=0.0059, is approximately equal to 0.Therefore the image extracted is exactly original watermark image substantially.
Various attacks processing is carried out to the Lena digital pictures after the insertion actual watermark shown in Fig. 2-3 below, to verify
The robustness proposed by the present invention being embedded in based on the image watermark for improving Arnold conversion and FSSVM with extracting method.
（1）Simple brightness regulation
Brightness regulation processing is carried out to the Lena digital pictures after the insertion actual watermark shown in Fig. 2-3, that is, owned
Pixel value carries out the computing Jia 75 He subtract 50 respectively, obtains the watermark Lena digital pictures as shown in Fig. 4-1 and Fig. 4-3.By
Image pixel value plus-minus processing after, visually from the point of view of, bright, the darkness of watermark Lena digital pictures there occurs obvious change,
And PSNR drops to 21.6031dB and 12.5791dB respectively.With the method for the present invention to the watermark shown in Fig. 4-1 and Fig. 4-3
Lena digital pictures carry out image watermark extraction, and the real image watermark extracted is respectively as shown in Fig. 4-2 and Fig. 4-4.As a result
Show, image watermark is not influenceed by luminance digital image substantially, with carrier image not by the watermark figure extracted when attacking
As almost consistent.BER now is respectively 0.0124 and 0.0074, is approximately equal to 0.Therefore the extraction algorithm is to carrier image
Brightness change is with very strong robustness.
（2）Histogram equalization
To after the insertion actual watermark shown in Fig. 2-3 Lena digital pictures carry out histogram equalization processing, obtain as
Watermark Lena digital pictures shown in Fig. 5-1.By histogram equalization processing, the pixel Distribution value of watermark Lena digital pictures
Obvious change is there occurs, PSNR drops to 15.9505dB.With the method for the present invention to the watermark Lena digitized maps shown in Fig. 5-1
As carrying out image watermark extraction, the real image watermark extracted is as shown in Fig. 5-2.From result, embedded real image
Watermark can be extracted more preferably, and BER=0.0088 is approximately equal to 0.Therefore the extraction algorithm is to the contrast of carrier image
Change has stronger robustness.
（3）It is superimposed Gaussian noise
Noise jamming is carried out to the Lena digital pictures after the insertion actual watermark shown in Fig. 2-3, noise is 0 from average
Gaussian noise with variance is 0.02, obtains the digital pictures of Lena containing watermark as in Figure 6-1.With the method for the present invention to figure
The digital pictures of Lena containing watermark shown in 6-1 carry out image watermark extraction, and the real image watermark extracted is as in fig. 6-2.
From Fig. 6-1, although the digital pictures of Lena containing watermark are by Gauusian noise jammer, visual quality is seriously degenerated, PSNR
23.9825dB is dropped to, but Fig. 6-2 shows, and embedded real image watermark still has good anti-noise jamming ability,
BER=0.0730, is approximately equal to 0, then the watermark extracted is relatively without result when attacking.Therefore the extraction algorithm is to making an uproar
Acoustic jamming has preferable robustness.
（4）Medium filtering
Median filter process, filtering window are carried out to the Lena digital pictures after the insertion actual watermark shown in Fig. 2-3
Size selection is [9 × 9], obtains the watermark Lena digital pictures as shown in Fig. 7-1.With the method for the present invention to shown in Fig. 7-1
Watermark Lena digital pictures carry out image watermark extraction, the real image watermark extracted is as shown in Fig. 7-2.Can by Fig. 7-1
To find out, at this moment the details of watermark Lena digital pictures is relatively fuzzyyer, and PSNR drops to 31.5116dB, but by Fig. 7-2 tables
Bright, embedded real image watermark still has more satisfactory anti-filter capacity, and BER=0.0270 is approximately equal to 0.Therefore this is carried
Take algorithm that there is preferable robustness to filtering process.
（5）JPEG compression
JPEG lossy compression method processing is carried out to the Lena digital pictures after the insertion actual watermark shown in Fig. 2-3, matter is compressed
It is 10% to measure the factor, obtains the watermark Lena digital pictures as shown in Fig. 8-1.With the method for the present invention to the watermark shown in Fig. 8-1
Lena digital pictures carry out image watermark extraction, and the real image watermark extracted is as shown in Fig. 8-2.It can be seen from Fig. 8-1
At this moment watermark Lena digital pictures show obvious blocking artifact, and visual quality there occurs serious degeneration, and PSNR is only
22.7442dB, but shown by Fig. 8-2, embedded real image watermark still has highly desirable anti-JPEG lossy compression methods processing energy
Power, BER=0.0737 is approximately equal to 0.Therefore the extraction algorithm has stronger robustness to JPEG compression processing.
（6）Geometry is cut
Geometry cutting process is carried out to the Lena digital pictures after the insertion actual watermark shown in Fig. 2-3, since left side
100 × 300 pixels are cut, watermark Lena digital pictures as shown in fig. 9-1 are obtained.With the method for the present invention to Fig. 9-1
Shown watermark Lena digital pictures carry out image watermark extraction, and the real image watermark extracted is as shown in Fig. 9-2.By Fig. 9-
1 as can be seen that at this moment watermark Lena digital pictures but shown by larger destruction, PSNR=12.5184dB by Fig. 9-2, this
Inventive method has relatively good robustness for geometry cutting, and embedded real image watermark remains to be extracted well
Come, BER=0.0126 is approximately equal to 0.Therefore the extraction algorithm has very strong robustness to geometry cutting process.
（7）Geometry rotates
Lena digital pictures after insertion actual watermark shown in Fig. 2-3 are rotated clockwise, angle is
10o, obtains the watermark Lena digital pictures as shown in Figure 10-1, PSNR=22.0357dB.With the method for the present invention to Figure 10-1
Shown watermark Lena digital pictures carry out image watermark extraction, it is not necessary to which first reversely rotating postrotational image again can be direct
Real image watermark is extracted, as shown in Figure 10-2.Shown by Figure 10-2, the inventive method still has very for geometry rotation attack
Strong robustness, embedded real image watermark can be extracted well, and BER=0.0142 is approximately equal to 0.Therefore this is carried
Take algorithm that there is very strong robustness to geometry rotation processing.
In summary, utilization space area image digital watermark of the present invention, determines that watermark is embedding based on Arnold conversion is improved
Enter the position into carrier image, not only key parameter increase, security improve, and realize watermark not sentience and
Its Blind extracting.Wherein, the scramble characteristic of Arnold conversion is 1. taken full advantage of, watermark is uniformly dispersedly distributed in carrier image
Whole space in；2. change the pixel value of embedded location using the thought for quantifying in mathematics and rounding up, not only realize
The not sentience of watermark, and counter can derive the quantizing rule of extraction when extracting watermark, it is not necessary to rely on original graph
Picture, realizes the Blind extracting of watermark；3. the original pixel value of carrier image is predicted using FSSVM when extracting watermark, by with adding
Watermark embedded location is found out in the contrast of carrier image pixel value after watermark, and this takes full advantage of the space domain characteristic of image, enter
And improve the precision and efficiency predicted the outcome.The method of the present invention is different from traditional image watermark and is embedded in and extracting method,
Its essence is utilization space area image digital watermark, organically combined based on Arnold conversion and FSSVM, overcome the former security not
The not strong shortcoming of high, robustness, also overcome the latter can not Blind extracting watermark defect, reach effect of mutual supplement with each other's advantages, be applicable
In the occasion of a variety of copyrights under fire.
It is given above it is a kind of specific preferred embodiment and specific experiment checking, but the present invention be not limited to it is described
Embodiment, as long as scrambled fashion is using improving Arnold conversion, and training method and embedded use traditional smooth support
Vector machine and qualified insertion rule, either FSSVM is with traditional embedding grammar or using traditional Smooth Support Vector Machines and biography
System embedding grammar can realize the present invention, because traditional embedded mode and traditional Smooth Support Vector Machines belong to existing skill
Art, will not be repeated here.
The basic ideas of the present invention are such scheme, for those of ordinary skill in the art, according to the religion of the present invention
Lead, design the models of various modifications, formula, parameter and creative work need not be spent.The principle of the present invention is not being departed from
Still fallen within the change carried out in the case of spirit to embodiment, modification, replacement and deformation in protection scope of the present invention.
Claims (6)
1. it is a kind of based on the Image Watermarking for improving Arnold conversion, it is characterised in that as follows including step：
1) scramble step：Using Arnold conversion is improved to watermarking images W processing, the insertion in carrier image I is obtained
Position；
2) training step：In embedded coordinate position, k sample point is chosen, by Smooth Support Vector Machines to k sample point
It is trained, k is setting value；
3) Embedded step：The pixel value of embedded watermarking images in initial carrier image；
Step 1) Arnold conversion is improved according to equation below
<mrow>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mi>n</mi>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>y</mi>
<mi>n</mi>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<mrow>
<mo>(</mo>
<mo>(</mo>
<mrow>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mi>a</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>b</mi>
</mtd>
<mtd>
<mrow>
<mi>a</mi>
<mi>b</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mn>0</mn>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>y</mi>
<mn>0</mn>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
<mo>)</mo>
<mi>mod</mi>
<mi> </mi>
<mi>M</mi>
<mo>)</mo>
</mrow>
<mi>&beta;</mi>
<mo>,</mo>
</mrow>
N iteration is carried out, with the position coordinates (x of watermarking images0,y0) as initial value, wherein 1≤x0≤ M, 1≤y0≤ K, is obtained
Embedded location (the x of correspondence watermark bitn,yn), wherein 1≤xn≤ N, 1≤yn≤ N, Wherein floor takes under being
Integral function, M be the exponent number of watermarking images matrix now, N be initial carrier image exponent number, a, b and n be positive integer andIt is the bulk composition of iteration；Finally give M × K position coordinates (xi,
yi), wherein i=1,2 ..., M × K；Improve Arnold conversion and with the addition of factor beta, different size watermark is adapted to carrying for adjusting
The matching of body image.
2. it is according to claim 1 a kind of based on the Image Watermarking for improving Arnold conversion, it is characterised in that
Step 2) in choose training object be sampled pixel point single order above square.
3. the watermark extracting method of watermark embedding method described in corresponding claims 1, it is characterised in that comprise the following steps：
1) training step：The supporting vector machine model trained with watermark telescopiny is carried out to the image I' of embedded watermark
Training, determines watermark embedded location；
2) Arnold inverse transformations step：The coordinate value of watermarking images is obtained by Arnold inverse transformations；
3) pixel value extraction step：Extracting rule is gone out according to embedding method backstepping during embedded watermark, so as to extract watermark pixel
Value；
4) watermark recovery step：Watermarking images are recovered according to the coordinate value of watermarking images and corresponding pixel value, restructuring is former
Beginning watermarking images W；
Step 2) in Arnold inverse transformations according to equation below：
<mrow>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mn>0</mn>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>y</mi>
<mn>0</mn>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msup>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mi>a</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>b</mi>
</mtd>
<mtd>
<mrow>
<mi>a</mi>
<mi>b</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msup>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mfrac>
<mi>i</mi>
<mi>&beta;</mi>
</mfrac>
</mtd>
</mtr>
<mtr>
<mtd>
<mfrac>
<mi>j</mi>
<mi>&beta;</mi>
</mfrac>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>)</mo>
</mrow>
<mi>mod</mi>
<mi> </mi>
<mi>M</mi>
<mo>,</mo>
</mrow>
Carry out n iteration, during each iteration, x0Assignment iny0Assignment inObtain position of (i, the j) correspondence in watermarking images
Put coordinate (x0,y0), wherein 1≤x0≤ M, 1≤y0≤ K,It is finding the inverse matrix, M=K is watermarking images
Order of matrix number, N be carrier image exponent number, a, b and n be positive integer andFinally give M × K
Position coordinates (xi,yi), wherein i=1,2 ..., M × K.
4. it is a kind of based on the image watermark flush mounting for improving Arnold conversion, it is characterised in that including：
1) scramble module：Using Arnold conversion is improved to watermarking images W processing, the insertion in carrier image I is obtained
Position；
2) training module：In embedded coordinate position, k sample point is chosen, by Smooth Support Vector Machines to k sample point
It is trained, k is setting value；
3) embedded module：The pixel value of embedded watermarking images in initial carrier image；
Module 1) Arnold conversion is improved according to equation below
<mrow>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mi>n</mi>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>y</mi>
<mi>n</mi>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<mrow>
<mo>(</mo>
<mo>(</mo>
<mrow>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mi>a</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>b</mi>
</mtd>
<mtd>
<mrow>
<mi>a</mi>
<mi>b</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mn>0</mn>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>y</mi>
<mn>0</mn>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
<mo>)</mo>
<mi>mod</mi>
<mi> </mi>
<mi>M</mi>
<mo>)</mo>
</mrow>
<mi>&beta;</mi>
<mo>,</mo>
</mrow>
N iteration is carried out, with the position coordinates (x of watermarking images0,y0) as initial value, wherein 1≤x0≤ M, 1≤y0≤ K, is obtained
Embedded location (the x of correspondence watermark bitn,yn), wherein 1≤xn≤ N, 1≤yn≤ N, Wherein floor takes under being
Integral function, M be the exponent number of watermarking images matrix now, N be initial carrier image exponent number, a, b and n be positive integer andIt is the bulk composition of iteration；Finally give M × K position coordinates (xi,
yi), wherein i=1,2 ..., M × K；Improve Arnold conversion and with the addition of factor beta, different size watermark is adapted to carrying for adjusting
The matching of body image.
5. it is according to claim 4 a kind of based on the image watermark flush mounting for improving Arnold conversion, it is characterised in that
Module 2) in choose training object be sampled pixel point single order above square.
6. the watermark extraction apparatus of watermark embedding device described in corresponding claims 4, it is characterised in that including：
1) training module：The supporting vector machine model trained with watermark telescopiny is carried out to the image I' of embedded watermark
Training, determines watermark embedded location；
2) Arnold inverse transform blocks：The coordinate value of watermarking images is obtained by Arnold inverse transformations；
3) pixel value extraction module：Extracting rule is gone out according to embedding method backstepping during embedded watermark, so as to extract watermark pixel
Value；
4) watermark retrieving module：Watermarking images are recovered according to the coordinate value of watermarking images and corresponding pixel value, restructuring is former
Beginning watermarking images W；
Module 2) in Arnold inverse transformations according to equation below：
<mrow>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<msub>
<mi>x</mi>
<mn>0</mn>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>y</mi>
<mn>0</mn>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msup>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mi>a</mi>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>b</mi>
</mtd>
<mtd>
<mrow>
<mi>a</mi>
<mi>b</mi>
<mo>+</mo>
<mn>1</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mrow>
<mo>-</mo>
<mn>1</mn>
</mrow>
</msup>
<mfenced open = "[" close = "]">
<mtable>
<mtr>
<mtd>
<mfrac>
<mi>i</mi>
<mi>&beta;</mi>
</mfrac>
</mtd>
</mtr>
<mtr>
<mtd>
<mfrac>
<mi>j</mi>
<mi>&beta;</mi>
</mfrac>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>)</mo>
</mrow>
<mi>mod</mi>
<mi> </mi>
<mi>M</mi>
<mo>,</mo>
</mrow>
Carry out n iteration, during each iteration, x0Assignment iny0Assignment inObtain position of (i, the j) correspondence in watermarking images
Put coordinate (x0,y0), wherein 1≤x0≤ M, 1≤y0≤ K,- 1 is finding the inverse matrix, and M=K is watermarking images square
Battle array exponent number, N be carrier image exponent number, a, b and n be positive integer andFinally give M × K position
Put coordinate (xi,yi), wherein i=1,2 ..., M × K.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410029947.5A CN104217389B (en) | 2014-01-22 | 2014-01-22 | A kind of method and apparatus based on the image watermark insertion, the extraction that improve Arnold conversion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410029947.5A CN104217389B (en) | 2014-01-22 | 2014-01-22 | A kind of method and apparatus based on the image watermark insertion, the extraction that improve Arnold conversion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104217389A CN104217389A (en) | 2014-12-17 |
| CN104217389B true CN104217389B (en) | 2017-09-05 |
Family
ID=52098841
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410029947.5A Expired - Fee Related CN104217389B (en) | 2014-01-22 | 2014-01-22 | A kind of method and apparatus based on the image watermark insertion, the extraction that improve Arnold conversion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104217389B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105205778B (en) * | 2015-11-09 | 2018-10-09 | 河南师范大学 | Image watermark insertion, extracting method and device based on RBF neural |
| CN107103577A (en) * | 2017-04-11 | 2017-08-29 | 中国科学院信息工程研究所 | Half fragile uniqueness image watermark insertion and extracting method for integrity protection |
| CN107256529B (en) * | 2017-05-16 | 2020-07-28 | 哈尔滨工程大学 | Wavelet digital watermark generation method of multi-target quantum bat evolution mechanism |
| CN107240061B (en) * | 2017-06-09 | 2021-01-12 | 河南师范大学 | Watermark embedding and extracting method and device based on dynamic BP neural network |
| CN109376793B (en) * | 2018-11-08 | 2022-03-15 | 深圳信息职业技术学院 | Image encryption method based on four-dimensional Chen hyper-chaotic system and K-means clustering |
| CN109712204B (en) * | 2018-11-30 | 2021-02-26 | 大连理工大学 | Image encryption method based on FCM clustering |
| CN115187443B (en) * | 2022-07-05 | 2023-04-07 | 海南大学 | Watermark embedding and detecting method and device based on spatial domain residual error feature fusion |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101038661A (en) * | 2007-04-26 | 2007-09-19 | 付永钢 | Blind watermark embedding and abstracting method based on regression support vector machine |
| CN102254295A (en) * | 2011-07-13 | 2011-11-23 | 西安电子科技大学 | Color halftoning image watermarking algorithm based on support vector machine |
-
2014
- 2014-01-22 CN CN201410029947.5A patent/CN104217389B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101038661A (en) * | 2007-04-26 | 2007-09-19 | 付永钢 | Blind watermark embedding and abstracting method based on regression support vector machine |
| CN102254295A (en) * | 2011-07-13 | 2011-11-23 | 西安电子科技大学 | Color halftoning image watermarking algorithm based on support vector machine |
Non-Patent Citations (2)
| Title |
|---|
| 《光滑支持向量机的原理和进展》;熊金志，胡金莲，袁华强;《计算机工程》;20080731;第34卷(第13期);第172-173页 * |
| 《基于广义Arnold变换与支持向量机的图像水印算法》;付永钢;《集美大学学报（自然科学版）》;20110131;第16卷(第1期);第66页第13行至第67页第35行 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104217389A (en) | 2014-12-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104217388B (en) | A kind of image watermark insertion based on FSSVM, the method and apparatus extracted | |
| CN104217389B (en) | A kind of method and apparatus based on the image watermark insertion, the extraction that improve Arnold conversion | |
| CN104217387B (en) | A kind of image watermark based on qualified insertion is embedded, the method and apparatus of extraction | |
| CN104361548B (en) | Image watermark insertion and extracting method based on BP neural network digital image compression | |
| CN104616244B (en) | Image watermark insertion and extracting method based on BP neutral net compression domains | |
| Sahu et al. | Digital image steganography using bit flipping | |
| CN106023056B (en) | Zero watermarking insertion, extracting method and the device compressed based on DWT and principal component analysis | |
| CN107240061B (en) | Watermark embedding and extracting method and device based on dynamic BP neural network | |
| CN104599225B (en) | Based on singular value decomposition and the insertion of the image watermark of principal component analysis and extracting method | |
| Ahmad et al. | Data hiding based on improved exploiting modification direction method and Huffman coding | |
| Wang et al. | Robust image watermarking approach using polar harmonic transforms based geometric correction | |
| Fatahbeygi et al. | A highly robust and secure image watermarking based on classification and visual cryptography | |
| Xia et al. | Geometrically invariant color medical image null-watermarking based on precise quaternion polar harmonic Fourier moments | |
| Murali et al. | An efficient ROI based copyright protection scheme for digital images with SVD and orthogonal polynomials transformation | |
| Xia et al. | Color image triple zero-watermarking using decimal-order polar harmonic transforms and chaotic system | |
| Wu et al. | SVD-based self-embedding image authentication scheme using quick response code features | |
| Wang et al. | A robust image watermarking algorithm using SVR detection | |
| Sisaudia et al. | A secure gray-scale image watermarking technique in fractional DCT domain using zig-zag scrambling | |
| CN104217390B (en) | A kind of Zero watermarking method, device and watermark extracting method, device | |
| Chhajed et al. | Applying decision tree for hiding data in binary images for secure and secret information flow | |
| CN114998080A (en) | Face tamper-proof watermark generation method, tamper detection method and attribute detection method | |
| Su | Color image watermarking: algorithms and technologies | |
| Yadav et al. | Lagrangian twin support vector regression and genetic algorithm based robust grayscale image watermarking | |
| Barani et al. | Image forgery detection in contourlet transform domain based on new chaotic cellular automata | |
| Gavini et al. | Lossless watermarking technique for copyright protection of high resolution images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170905 Termination date: 20210122 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |