CN105719226B - A kind of combination non-overlap piecemeal and the watermark insertion and extracting method for waiting frequency bands to merge - Google Patents
A kind of combination non-overlap piecemeal and the watermark insertion and extracting method for waiting frequency bands to merge Download PDFInfo
- Publication number
- CN105719226B CN105719226B CN201610052872.1A CN201610052872A CN105719226B CN 105719226 B CN105719226 B CN 105719226B CN 201610052872 A CN201610052872 A CN 201610052872A CN 105719226 B CN105719226 B CN 105719226B
- Authority
- CN
- China
- Prior art keywords
- formula
- watermark
- piecemeal
- random
- overlap
- 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.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/0021—Image watermarking
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Editing Of Facsimile Originals (AREA)
- Image Processing (AREA)
Abstract
The watermark insertion and extracting method that the present invention provides a kind of combination non-overlap piecemeal and frequency bands is waited to merge, the key that mentioned method is given by the SHA-1 value of cover-image and user generates watermark encrypting link and the initial value and parameter that are embedded in link respectively, encryption is carried out to 2 value watermarks for driving chaotic maps to generate 2 value pseudo random matrixs and generates the random coordinates position of not overlap partition to insertion carrier image, then it carries out l grades of wavelet coefficient pyramids to each random not overlap partition to decompose, by HLlAnd LHlFrequency band coefficient carries out equal weight and merges to form fusion coefficients, is embedded in watermark by adjusting fusion coefficients quantized value.The insertion of institute's water lift print and extract strategy can full disclosure, compared with tradition can disclose sensitive watermark, mentioned method, which is avoided, is fused together brought short -board effect for wavelet coefficient not at the same level, it also avoids filtering out strategy to the isolated point for extracting watermark 2 adjustment of progress simultaneously, but enough insertion safeties can be still provided.
Description
Technical field
The invention belongs to information security and digital image signal process crossing research field, it is related to a kind of watermark insertion and mentions
Take method, in particular to a kind of combination non-overlap piecemeal and the watermark insertion and extracting method that wait frequency bands to merge.
Background technique
The development of multimedia and internet enables people to more easily obtain various digital image informations.Number
Image has many advantages, such as easy storage, easy transporting and ease for operation, but simultaneously but also bootlegger can carry out not it
The duplication and propagation of authorization;Malicious attacker easily can be distorted or be forged to it, to easily cause serious consequence and lead
Cause heavy economic losses.
For image information safety, people have proposed a variety of methods, such as: 1. plaintext image is converted to the number of ciphertext image
Word image encryption technology;2. by digital picture steganography of the secret image hiding in irrelevant carrier;3. in conjunction with existing skill
The digital image blind forensic technologies that art feature screens the authenticity and integrity of image;4. secret image is split as shadow
Subgraph, using part distribution shadow image reconstruct Secret Image digital image sharing technology and 5. give image add copyright
Certification mark, to the authenticity and integrity of image, source person and user identify the digital image watermarking technology of certification.
Wherein digital image watermarking technology, it has also become the research hotspot of image information safety.
Van Schyndel etc. is relatively early to propose digital watermarking (A digital watermark.Proceedings of the
IEEE International Conference on Image Processing,1994,2:86-90.)。
What traditional digital watermark was generally concerned with is how watermark is embedded in, to reduce to blindage visual quality or acoustical quality
Influence generally use the fixation for embedding watermark data into frequency-domain transform and to the safety shorter mention of watermark embedded location
Position.Such as: watermark is divided into two parts by Lai, passes through the LH and HL of modification 1 grade of Haar wavelet decomposition of cover-image respectively
The singular value of band is embedded in watermark (Digital image watermarking using discrete wavelet
transform and singular value decomposition. IEEE Transactions on
Instrumentation and Measurement,2010,59(11): 3060-3063.);Tao etc. will pass through affine transformation
Watermark be embedded into the HL of 3 grades of wavelet decompositions of cover-image3In each piecemeal of sub-band coefficients, and utilize particle swarm optimization algorithm
To optimize watermark embedment strength (A wavelet-based particle swarm optimization algorithm for
digital image watermarking. Integrated Computer-Aided Engineering,2012,19(1):
81-91.);The watermark that Arnold conversion encrypts is embedded into the LL of 3 grades of wavelet decompositions of cover-image by Zope etc.3And HH3Frequency band
(Robust copyright protection of raster images using wavelet based in wavelet coefficient
digital watermarking.2014IEEE International Geoscience and Remote Sensing
Symposium (IGARSS),2014:3129-3132.);The LL that Preda etc. is decomposed with 2 grades of integer wavelets2The coefficient of frequency band is raw
At watermark, HL is embedded in after carrying out random permutation to it1、LH1And HH1(Self-recovery of in the wavelet coefficient of frequency band
unauthentic images using a new digital watermarking approach in the wavelet
domain.2014 10th IEEE International Conference on Communications, 2014:1-4.)。
Above method, which is concerned with, to be embedded watermark data into carrier, it is made to meet visual invisibility, and to watermark
The safety of embedding strategy using fixed frequency-domain transform without reference to causing embedded location of the watermark on blindage to be fixed.One
A little methods are while try to the watermark enhanced by carrying out scramble to watermark to insertion is protected, but still uses fixed frequency
Domain converts to carry out watermark insertion, and thus the embedded location on blindage is also still fixed.Fixed embedded location leads to watermark
Once open, embedded watermark can be rejected or be distorted easily embedding strategy.
This defect of wavelet transformation is realized using fixed wavelet basis to overcome traditional digital watermark. Meerwald
It is applied to watermaking system Deng by Parameterization of Wavelet, proposes the safe watermark method based on parametrization wavelet transform
(Watermark security via wavelet filter parametrization. Proceedings of the
IEEE International Conference on Image Processing, 2001,3:1027-1030.), basic
Thought is to attempt to realize that different wavelet parameters corresponds to different transform domains by constructing variable filter, and by ginseng
Number is maintained secrecy and is protected the embedded location of watermark, to improve the safety of digital watermarking.On this basis, Wu etc. proposes one kind
Semi-fragile watermarking image authentication method (A secure semi-fragile watermarking based on parameter integer small echo
for image authentication based on integer wavelet transform with
parameters.Proceedings of the Australasian workshop on Grid computing and e-
Research, 2005,44:75-80.), it is while obtains safety and lower computational complexity, is constructed using boosting algorithm
Parameter integer wavelet transformation.Cover-image is carried out 3 grades of parameter integer wavelet decompositions by this method, then passes through modification low frequency
Subband LL3Minimum five insertions watermark of middle coefficient.Chamlawi etc. is added to the method (Self of Piva etc. on the basis of Wu
recovery authentication of images in the DWT domain.International Journal of
Image and Graphics, 2005,5 (1): 149-165.) propose image authentication and recovery side based on parametrization small echo
Method (Wavelet based image authentication and recovery.Journal of Computer
Science and Technology, 2007,22 (6): 795-804.), reach image authentication and extensive by being embedded in doubly time series model
Multiple function.
But parametrization small echo is used to determine and influence the embedded location of watermark to improve the safety of watermark embedding strategy,
Above method all without carrying out theoretic analysis, based on only test an example.It is whole using parameterizing through experimental check
Number small echo is difficult to meet actual demand for the sensibility that parameter value changes, when the parameter for parameterizing small echo changes, according to
The watermark being embedded in approximate can be so extracted, therefore the safety for the watermark being embedded in still is difficult to ensure.
In applied patent of invention: digital watermark embedding and extracting method based on the quantization of multilevel wavelet coefficient weighted sum
(a kind of digital watermark embedding and extracting method [P] China based on the quantization of multilevel wavelet coefficient weighted sum of Shao Liping, Zhu Ying,
State Intellectual Property Office of the People's Republic of China, 201510145048.6.), by cover-image MD5 value, user key and initial
Parameter and Logistic map bindings are randomly choosed for the multilevel wavelet coefficient to watermark encrypting and insertion link, then
Watermark is embedded in using the multilevel wavelet coefficient weighted sum quantization of selection.In conjunction with multilevel wavelet coefficient be embedded in watermark, obtain compared with
There are the following problems: the influence to wavelet coefficient not at the same level is different and not at the same level when image is by transformation while good performance
When wavelet coefficient variable quantity is identical, the influence to image array variable quantity is also different, although we pass through weight and adjustment small echo
Coefficient weighted mean balances the influence of wavelet coefficients at different levels, balances the invisibility and robustness of watermarking algorithm, but at different levels
The performance difference of wavelet coefficient leads to short -board effect, makes the watermark quality extracted decline, and anti-attack ability weakens.
Summary of the invention
The invention reside in prior art defect is overcome, a kind of combination non-overlap piecemeal and equal frequency bands wavelet coefficient fusion are provided
Sensitive watermark insertion and extracting method are disclosed, by the way that the SHA-1 value of cover-image and user key are added with generation watermark
The initial value and Wire Parameters of close and random site selection course improve the safety and effectively of watermark insertion and watermark extracting
It avoids and wavelet coefficient not at the same level is fused together brought short -board effect, avoid and 2 adjustment are carried out to extraction watermark
Isolated point filter out strategy, but can still provide enough insertion safeties.
To achieve the above object, the invention adopts the following technical scheme:
A kind of watermark embedding method of combination non-overlap piecemeal and equal frequency bands fusion, comprising the following steps:
Step 1: note cover-image A=(Pi,j)M×NCorresponding 160 SHA-1 values are Asha-1, by Asha-1Be converted to 40
16 system numbers constitute 16 system sequences, are denoted as
Step 2: by user key xinit∈ (0,1) and μinit∈ [3.57,4] is as the initial value x and parameter μ in formula (1)
And drive-type (1) generates the pseudo-random sequence that length is 40It is temporary to eliminate
State effect abandons count >=0 time result before iteration, and S is generated from count+1 timesb, count gives by user;;
X=μ x (1-x) (1);
Step 3: by SaAnd SbSubstitution formula (2) and formula (3) calculate the intermediate initial value x for watermark protection0And parameter μ0;
Step 4: 2 value watermark figures of note are W=(wii,jj)m×n, by x0And μ0Substitution formula (1) generate length be m × n it is pseudo- with
Machine sequence Sc=< c0,…,ck,…,cm×n-1>, ck∈ (0,1), is similarly elimination transient effect, by count >=0 time before iteration
As a result it abandons, S is generated from count+1c, count gives by user;
Step 5: by pseudo-random sequence Sc2 values are carried out, are scanned as 2 value matrix E=(eii,jj)m×n, W encrypt
To encrypted watermark figure W '=(w 'ii,jj)m×n;
Step 6: S is utilizedaAnd SbThe random coordinates of [0, M-1] × [0, N-1] range are generated, and by the coordinate in blindage
Image A=(Pi,j)M×NM × n size of upper interception for el × el not overlap partition, meet el be even number andWherein M × N is cover-image A=(Pi,j)M×NSize;
Step 7: note block0,block1,…,blockm×n-1For each piecemeal of generation, L successively is carried out to each piecemeal
Grade wavelet pyramid converts to form 3L+1 subband, and meets
Step 8: note blockk, k=0,1 ..., the HL of m × n-1LAnd LHLAll coefficients of frequency band areWherein num isWithThe number of frequency band coefficient, byFusion coefficients weight is calculated by formula (15)k;
Step 9: fusion coefficients variable quantity is calculated using formula (16) and formula (17), is modified by formula (18)Watermark is embedded in,
Wherein q is quantization step:
Δk=weight 'k-weightk (17)
Step 10: each piecemeal is subjected to L grades of wavelet pyramid inverse transformations, reconstructs the A ' containing watermarking images=(P 'i,j)M×N。
Further, step 1 is by Asha-1Be converted to 40 16 system numbers, constitute 16 system sequences specific method be from
Asha-1Highest order from left to right start A successivelysha-1Be converted to 40 16 system numbers.
Further, step 5 is by pseudo-random sequence Sc2 values are carried out, are scanned as 2 value matrix E=(eii,jj)m×nSpecific side
Method is by formula (4) by pseudo-random sequence ScIt is converted into 2 value matrix E=(eii,jj)m×n:
Step 5 encrypts W to obtain encrypted watermark figure W '=(w 'ii,jj)m×nMethod particularly includes:
Step 6 utilizes SaAnd SbThe random coordinates of [0, M-1] × [0, N-1] range are generated in cover-image A=
(Pi,j)M×NThe specific method for the not overlap partition that m × n size of upper interception is el × el is by SaAnd SbIt is produced by formula (6)~(9)
Raw Logistic chaotic maps initial value x1,y1And parameter μ1,μ2, then by initial value x1,y1And parameter μ1,μ2By formula (10) and
Formula (11) generate pseudo random number, then by its by formula (12) and formula (13) generation random coordinates, judge its whether with all to have produced
Raw random coordinates are that the piecemeal of starting point is overlapped, if so, iterative (10)~formula (13) continues to judge, otherwise, at random with this
Coordinate is the piecemeal that starting point takes size to be el × el, until generating m × n piecemeal:
yt+1=μ1xt(1-xt) (10)
xt+1=μ2yt(1-yt) (11)
The random coordinates point that note generates is (X, Y), and the piecemeal that the size of selection is el × el is Bk=(bkr,s)el×el,
It is then that starting point chooses piecemeal Bk=(bk with random coordinates (X, Y) in step 6r,s)el×elSpecific method is bkr,s=
P(X+r)modM,(Y+s)modN, wherein P(X+r)modM,(Y+s)modNFor the pixel on cover-image A.
Further, the specific method that step 7 successively carries out the transformation of L grades of wavelet pyramids to each piecemeal is small by Haar
Wave conversion carries out L grades of wavelet pyramid transformation, shown in the specific method of Haar wavelet transformation such as formula (14):
In formula (14), f0,f1,f2,f3It is 4 elements that 2 × 2 fritter of matrix is adjacent up and down before converting, f0′,f1′,
f2′,f3' it is transformed coefficient, transformed all coefficients are put together by locating frequency band, that is, are constituted to original matrix
The level-one pyramid decomposition of element, wherein f0′,f1′,f2′,f3' respectively correspond as LL, HL, LH, HH subband.
Further, step 10 by each piecemeal carry out L grade wavelet pyramid inverse transformations, reconstruct the A ' containing watermarking images=
(P′i,j)M×NSpecific method be the inverse transformation converted using formula (19) Haar small echo pyramid:
In formula (19), f0,f1,f2,f3It is 4 elements that 2 × 2 fritter of matrix is adjacent up and down after converting, f0′,f1′,
f2′,f3' it is LL, HL, LH before converting, HH subband corresponding position element.
A kind of watermark extracting method of combination non-overlap piecemeal and equal frequency bands fusion, comprising the following steps:
Step 1: the A ' containing watermarking images=(P ' of certification to be extracted is inputtedi,j)M×NIt is corresponding with cover-image before insertion watermark
160 SHA-1 values be Asha-1, by Asha-140 16 system numbers are converted to, 16 system sequences are constituted
Step 2: by user key xinit∈ (0,1) and μinit∈ [3.57,4] is as the initial value x and parameter μ in formula (1)
And drive-type (1) generates the pseudo-random sequence that length is 40It is temporary to eliminate
State effect abandons count >=0 time result before iteration, and S is generated from count+1 timesb, count gives by user;
X=μ x (1-x) (1);
Step 3: by SaAnd SbSubstitution formula (2) and formula (3) calculate the intermediate initial value x for being protected to watermark0With
Parameter μ0;
Step 4: S is utilizedaAnd SbGenerate [0, M-1] × [0, N-1] range random coordinates the A ' containing watermarking images=
(P′i,j)M×NM × n size of upper interception for el × el not overlap partition, meet el be even number and
Step 5: successively to each piecemeal block of generation0,block1,…,blockm×n-1Carry out L grades of wavelet pyramids
Transformation forms 3L+1 subband, and meets
Step 6: note blockk, k=0,1 ..., the HL of m × n-1LAnd LHLAll coefficients of frequency band areWherein num isWithThe number of frequency band coefficient, byFusion coefficients weight is calculated by formula (15)k;
Step 7: watermark figure EW '=(ew ' of encryption is extractedii,jj)m×n;
Step 8: for by x0And μ0Substitution formula (1) generates the pseudo-random sequence S that length is m × nc=< c0,…,ck,…,
cm×n-1>, ck∈ (0,1), is similarly elimination transient effect, and count >=0 time result before iteration is abandoned, and produces from count+1
Raw Sc, count inputs by user;
X=μ x (1-x) (1);
Step 9: by pseudo-random sequence Sc2 values are carried out, then scanning is 2 value matrix E=(eii,jj)m×n, to EW ' progress
Watermark figure EW=(ew after being decryptedii,jj)m×n, and by decryption watermark EW=(ewii,jj)m×nW is printed with raw water
=(wii,jj)m×nComparison, obtain authentication result figure.
Further, step 1 is by Asha-140 16 system numbers are converted to, 16 system sequences are constitutedSpecific method be from Asha-1Highest order from left to right start successively
By Asha-1Be converted to 40 16 system numbers.
Further, step 4 utilizes SaAnd SbThe random coordinates of [0, M-1] × [0, N-1] range are generated in the A ' containing watermarking images
=(P 'i,j)M×NThe not overlap partition that upper m × n size of interception is el × el method particularly includes:
By SaAnd SbLogistic chaotic maps initial value x is generated by formula (6)~(9)1,y1And parameter μ1,μ2, then will be first
Initial value x1,y1And parameter μ1,μ2Generate pseudo random number by formula (10) and formula (11), then by its by formula (12) and formula (13) generation with
Machine coordinate, judge its whether with it is all Chong Die as the piecemeal of starting point using the random coordinates generated, if so, iterative (10)~
Formula (13) continues to judge, otherwise, size is carried out by starting point of the random coordinates for the piecemeal of el × el, until generating m × n
Piecemeal:
yt+1=μ1xt(1-xt) (10)
xt+1=μ2yt(1-yt) (11)
The random coordinates point that note generates is (X, Y), and the piecemeal that the size of selection is el × el is Bk=(bkr,s)el×el,
It is then that starting point chooses piecemeal Bk=(bk with random coordinates (X, Y) in step 4r,s)el×elSpecific method is bkr,s=
P′(X+r)modM,(Y+s)modN, wherein P '(X+r)modM,(Y+s)modNFor the A ' containing watermarking images=(P 'i,j)M×NOn pixel.
Further, it is characterised in that: step 5 carries out L grades of wavelet pyramids and converts the specific method to form 3L+1 subband
To carry out l grades of wavelet pyramid transformation by Haar wavelet transformation, shown in the specific method of Haar wavelet transformation such as formula (14):
Step 7 extracts encrypted watermark figure EW '=(ew 'ii,jj)m×nSpecific method be extracted by formula (20)
Watermark figure EW '=(ew ' of encryptionii,jj)m×n, q is quantization step;
(20)。
Further, step 8 is by x0And μ0Generate the pseudo-random sequence S that length is m × nc=< c0,…,ck,…,cm×n-1
>, ckThe specific method of ∈ (0,1) is by x0And μ0Substitution formula (1) generates the pseudo-random sequence S that length is m × nc=<
c0,…,ck,…,cm×n-1>, ck∈ (0,1), is similarly elimination transient effect, and count >=0 time result before iteration is abandoned, from
Count+1, which rises, generates Sc, count inputs by user;
X=μ x (1-x) (1);
Step 9 is by pseudo-random sequence ScThen binaryzation scans 2 value matrix E=(eii,jj)m×nSpecific method be by formula
(4) by ScIt is converted into matrix E=(eii,jj)m×n;
The specific method that EW ' is decrypted is by formula (21) to EW ' carry out exclusive or decryption by step 9;
Step 9 is to decryption watermark EW=(ewii,jj)m×nW=(w is printed with raw waterii,jj)m×nComparison obtain authentication result figure
Specific method be if ewii,jj=wii,jj, then blockk, k=ii × n+jj is by certification, otherwise, in authentication result figure
By block labeled as white.
Compared with the existing technology advantage is as follows by the present invention:
What the present invention provided is a kind of insertion and the ostensible sensitive watermark building method of extraction strategy, in the method,
Key bindings that the SHA-1 value of cover-image and user give are generated watermark encrypting link and be embedded in link initial value and
Parameter carries out encryption to 2 value watermarks for driving chaotic maps to generate 2 value pseudo random matrixs and generates to insertion carrier image
The not random coordinates position of overlap partition, and different cover-image SHA-1 value, that user key will lead to different watermarks is embedding
It is different with the parameter to watermark encrypting to enter position, to guarantee the safety of watermark encrypting and embedding strategy.Not with conventional watermark
Together, even if insertion and extraction algorithm full disclosure all will lead to the water of insertion using different cover-images or different keys
Chine goes through different watermark encrypting and insertion link, so that attacker attempts by selecting special cover-image to guess
Watermark encrypting parameter is correspondingly embedded in the tactful utter failure that position is distorted watermark and modified in turn with watermark is found.
What the present invention provided is a kind of sensibility watermark building method of real meaning, the key and blindage provided user
Characteristics of image extreme sensitivity, and this sensibility theoretically can confirm and guarantee.And use parameter integer small echo
The sensibility changed for parameter value is difficult to meet actual demand, through practice examining, when the parameter of parametrization small echo changes
When, the watermark being embedded in approximate can be still extracted, therefore the safety for being embedded in watermark is still difficult to ensure.
What the present invention provided is a kind of combination non-overlap piecemeal openly sensitive watermark with equal frequency bands wavelet coefficient fusion
Insertion and extracting method, with the next item up patent applied for method (Shao Liping, Zhu Ying one kind be based on multilevel wavelet coefficient weighting
Chinese, the State Intellectual Property Office of the People's Republic of China with digital watermark embedding and extracting method [P] of quantization,
201510145048.6) it compares, mentioned method, which is avoided, is fused together brought short -board effect for wavelet coefficient not at the same level,
It also avoids filtering out strategy to the isolated point for extracting watermark 2 adjustment of progress simultaneously, but enough insertion safety can be still provided
Property, and using original watermark to insertion cover-image carry out certification and to JPEG compression, cut out, noise, covering, scribble have
Certain robustness, therefore apply for a patent relative to traditional robust watermarking method and the next item up has higher safety and more
Good practical application value.
Detailed description of the invention
Fig. 1 is watermark embedding method flow chart;
Fig. 2 is watermark extracting method flow chart;
Fig. 3 is embodiment: cover-image, is 8 normal grayscale image man of 1024 × 1024 resolution ratio;
Fig. 4 is embodiment: certification to be extracted contains watermarking images, is 8 gray level images of 1024 × 1024 resolution ratio, phase
To the PSNR=50.31dB of Fig. 3;
Fig. 5 is embodiment: watermarking images, is 82 value images of 64 × 64 resolution ratio;
Fig. 6 is embodiment: the watermarking images extracted from Fig. 4, is 82 value images of 64 × 64 resolution ratio;
Fig. 7 is embodiment: the SHA-1 value for providing mistake (is indicated are as follows: 0x0bc0c0444a806971bf with 16 system numbers
E7bbc6f4821926e80e5847) the watermarking images extracted are 82 value images of 64 × 64 resolution ratio, the NC of opposite Fig. 5
=0.50;
Fig. 8 is embodiment: 1bit (is indicated are as follows: 0x 7167aff13fa6fe8685 in modification SHA-1 value with 16 system numbers
249ec02c595ba46b6c0238) the watermarking images extracted, 82 value images that resolution ratio is 64 × 64, the NC of opposite Fig. 5
=0.51;
Fig. 9 is embodiment: providing the x of mistakeinitValue (is revised as 0.7635891325647968
0.7635891325647967) watermarking images extracted are 82 value images of 64 × 64 resolution ratio, the NC=of opposite Fig. 5
0.50;
Figure 10 is embodiment: providing the x of mistakeinitValue (is revised as 0.7635891325647968
0.7635891325647969) watermarking images extracted are 82 value images of 64 × 64 resolution ratio, the NC=of opposite Fig. 5
0.51;
Figure 11 is embodiment: providing the μ of mistakeinitValue (is revised as 3.9836387425412795
3.9836387425412794) watermarking images extracted are 82 value images of 64 × 64 resolution ratio, the NC=of opposite Fig. 5
0.50;
Figure 12 is embodiment: providing the μ of mistakeinitValue (is revised as 3.9836387425412795
3.9836387425412796) watermarking images extracted are 82 value images of 64 × 64 resolution ratio, the NC=of opposite Fig. 5
0.51;
Figure 13 is embodiment: provide mistake count value (being revised as 2999 for 3000) extract watermarking images, be 64 ×
82 value images of 64 resolution ratio, the NC=0.51 of opposite Fig. 5;
Figure 14 is embodiment: provide mistake count value (being revised as 3001 for 3000) extract watermarking images, be 64 ×
82 value images of 64 resolution ratio, the NC=0.50 of opposite Fig. 5;
Figure 15 is embodiment: Fig. 6 extracts the authentication image of watermark, is 82 value images of 1024 × 1024 resolution ratio;
Figure 16 is embodiment: providing wrong SHA-1 value, (SHA-1 value is 0x0bc0c0444a806971bfe7bbc6f
4821926e80e5847) corresponding authentication image is 82 value images of 1024 × 1024 resolution ratio;
Figure 17 is embodiment: SHA-1 value (is revised as 0x 7167aff13fa6fe8685249ec02 by modification SHA-1 value
C595ba46b6c0238 the authentication image of 1bit in) is 82 value images of 1024 × 1024 resolution ratio;
Figure 18 is embodiment: providing the x of mistakeinitValue (is revised as 0.7635891325647968
0.7635891325647967) authentication image is 82 value images of 1024 × 1024 resolution ratio;
Figure 19 is embodiment: providing the x of mistakeinitValue (is revised as 0.7635891325647968
0.7635891325647969) authentication image is 82 value images of 1024 × 1024 resolution ratio;
Figure 20 is embodiment: providing the μ of mistakeinitValue (is revised as 3.9836387425412795
3.9836387425412794) authentication image, be 1024 × 1024 resolution ratio 82 value images;
Figure 21 is embodiment: providing the μ of mistakeinitValue (is revised as 3.9836387425412795
3.9836387425412796) authentication image, be 1024 × 1024 resolution ratio 82 value images;
Figure 22 is embodiment: provide mistake count value (being revised as 2999 for 3000) authentication image, be 1024 ×
82 value images of 1024 resolution ratio;
Figure 23 is embodiment: provide mistake count value (being revised as 3001 for 3000) authentication image, be 1024 ×
82 value images of 1024 resolution ratio;
Figure 24 is embodiment: containing watermarking images by after the JPEG compression of quality factor Q=80, is 1024 × 1024 points
8 gray level images of resolution, the PSNR=36.61dB of opposite Fig. 3;
Figure 25 is embodiment: containing watermarking images after cutting out, is 8 gray level images of 1024 × 1024 resolution ratio, relatively
The PSNR=13.91dB of Fig. 3;
Figure 26 is embodiment: containing watermarking images after 0.5% white noise of addition, is 8 ashes of 1024 × 1024 resolution ratio
Spend image, the PSNR=26.31dB of opposite Fig. 3;
Figure 27 is embodiment: containing watermarking images after covering, is 8 gray level images of 1024 × 1024 resolution ratio, relatively
The PSNR=21.78dB of Fig. 3;
Figure 28 is embodiment: containing watermarking images after scribble, is 8 gray level images of 1024 × 1024 resolution ratio, relatively
The PSNR=23.55dB of Fig. 3;
Figure 29 is embodiment: by after the JPEG compression of quality factor Q=80 containing the watermarking images extracted in watermarking images,
For 82 value images of 64 × 64 resolution ratio, the NC ≈ 1.00 of opposite Fig. 5;
Figure 30 is embodiment: after cutting out containing the watermarking images extracted in watermarking images, be 82 of 64 × 64 resolution ratio
It is worth image, the NC=0.95 of opposite Fig. 5;
Figure 31 is embodiment: after 0.5% white noise of addition containing the watermarking images extracted in watermarking images, be 64 × 64
82 value images of resolution ratio, the NC=0.92 of opposite Fig. 5;
Figure 32 is embodiment: after covering containing the watermarking images extracted in watermarking images, be 82 of 64 × 64 resolution ratio
It is worth image, the NC=0.96 of opposite Fig. 5;
Figure 33 is embodiment: after scribble containing the watermarking images extracted in watermarking images, be 82 of 64 × 64 resolution ratio
It is worth image, the NC=0.99 of opposite Fig. 5;
Figure 34 is embodiment: by the authentication image after the JPEG compression of quality factor Q=80, being differentiated for 1024 × 1024
82 value images of rate;
Figure 35 is embodiment: the authentication image containing watermarking images after cutting out, and is 82 values of 1024 × 1024 resolution ratio
Image;
Figure 36 is embodiment: the authentication image containing watermarking images after 0.5% white noise of addition is 1024 × 1024 points
82 value images of resolution;
Figure 37 is embodiment: the authentication image containing watermarking images after covering, is 82 values of 1024 × 1024 resolution ratio
Image;
Figure 38 is embodiment: the authentication image containing watermarking images after scribble, is 82 values of 1024 × 1024 resolution ratio
Image;
Specific embodiment
Below in conjunction with specific embodiment, the present invention will be described: implementing ring using JAVA jdk1.7.0_09 as case
Border is described in detail embodiment of the present invention in conjunction with attached drawing, but is not limited to the implementation case, and wherein Fig. 1 is that watermark is embedding
Enter method flow diagram, Fig. 2 is watermark extracting method flow chart.
A kind of watermark embedding method of combination non-overlap piecemeal and equal frequency bands fusion, comprising the following steps:
Step 1: choosing image A as shown in figure 3, being 1024 × 1,024 8 normal grayscale image man.bmp, note blindage
Image A=(Pi,j)1024×1024Corresponding 160 SHA-1 values are Asha-1=0,111,000,101,100,111,101 01111111100
010011111110100110111111101000011010000101001001001001111 011000000001011000
10110010101101110100100011011000000001000111001, from Asha-1Left side highest order starts it successively
40 16 system numbers are converted to, 16 system sequence S are constituteda=< 7,1,6,7,10,15,15,1,3,15,10,6,15,14,8,
6,8,5,2,4,9,14,12,0,2,12,5,9,5,11,10,4,6,11,6,12,0,2,3,9 >;
Step 2: by user key xinit∈ (0,1) and μinit∈ [3.57,4] as formula (1) initial value x and parameter μ simultaneously
Drive-type (1) generates the pseudo-random sequence that length is 40To eliminate transient state
Effect abandons count >=0 time result before iteration, and S is generated from count+1 timesb, count gives by user;Such as: xinit
=0.7635891325647968, μinit=3.9836387425412795, count=3000, generate length be 40 it is pseudo- with
Machine sequence b0=0.2037621195996894, b1=
0.6463179714356886,b2=0.9106241678977656, b3=0.3242195643220826, b4=
0.8728201819600013,b5=0.4422042644778347, b6=0.9826029497351249, b7=
0.06809788586437480,b8=0.2528039605978175, b9=0.7524859271167631, b10=
0.7419561282150838,b11=0.7626964470112642, b12=0.7210010734871887, b13=
0.8013428956440323,b14=0.6341652481682873, b15=0.9242029381397697, b16=
0.2790613324584196,b17=0.8014527630757558, b18=0.6339014212917692, b19=
0.9244846737878244,b20=0.2781088223085844, b21=0.7997724645639421, b22=
0.6379258439511960,b23=0.9201267809267446, b24=0.2927715059146659, b25=
0.8248377026854120,b26=0.5755579855393184, b27=0.9731670554892884, b28=
0.1040245099033162,b29=0.3712887199633916, b30=0.9299143615039163, b31=
0.2596282443581546,b32=0.7657406922318214, b33=0.7145926247730029, b34=
0.8124631430251899,b35=0.6069742248179939, b36=0.9503229763343125, b37=
0.1880644657881021,b38=0.6082865877745214, b39=0.9491975971289682;
Step 3: by SaAnd SbSubstitution formula (2) and formula (3) calculate the intermediate initial value x for being protected to watermark0And ginseng
Number μ0;Such as:
It can be calculated by formula (2) and formula (3):
Step 4: choosing image W as shown in figure 5, being 82 value images of 64 × 64 resolution ratio, by x0And μ0Substitution formula (1)
Generate the pseudo-random sequence S that length is 64 × 64c=< c0,…,ck,…,c4095>, ck∈ (0,1), by count=before iteration
3000 times result is abandoned, wherein c0=0.4993912172500605, c1=0.9764121237695018 ..., c4095=
0.4993912172500605;
Step 5: formula (4) are pressed by pseudo-random sequence ScIt is converted into 2 value matrix E=(eii,jj)64×64, W is obtained by formula (5)
The encrypted watermark figure W '=(w ' of exclusive orii,jj)64×64;Such as: w0,0=1, w0,1=1 ...,
w63,63=1,
Step 6: by SaAnd SbLogistic chaotic maps initial value x is generated by formula (6)~(9)1,y1And parameter μ1,μ2, so
Afterwards by initial value x1,y1And parameter μ1,μ2Pseudo random number is generated by formula (10) and formula (11), then it is pressed into formula (12) and formula (13)
Generate random coordinates, judge its whether with it is all Chong Die as the piecemeal of starting point using the random coordinates generated, if so, iterative
(10)~formula (13) continues to judge, otherwise, size is carried out by starting point of the random coordinates for the piecemeal of el × el, until generating 64
× 64 piecemeals:
The specific production method of each piecemeal is: the random coordinates point for remembering generation is (X, Y), the size of selection be el ×
The piecemeal of el is Bk=(bkr,s)el×el, then it is that starting point chooses piecemeal Bk=(bk with random coordinates (X, Y) in step 6r,s)el×el
Specific method is bkr,s=P(X+r)modM,(Y+s)modN, wherein P(X+r)modM,(Y+s)modNFor the pixel on image A to be encrypted.
Such as: el=8,
By taking first group of pseudo random number of generation as an example: yt+1=μ1xt(1-xt)=0.9707855550412292, xt+1=
μ2yt(1-yt)=0.02468364164788569, then random coordinates of first piecemeal starting point are as follows: Piecemeal rises
Point coordinate (648,516), bk0,0=P648,516,bk0,1=P648,517,…,bk7,7=P655,523;
Step 7: note block0,block1,…,blockm×n-1For each piecemeal of generation, l successively is carried out to each piecemeal
Grade wavelet pyramid converts to form 3L+1 subband, and l value meets
Such as: if L=3,3 grades of Haar wavelet pyramids are carried out to first piecemeal and convert to form 10 subbands, form 1
Low frequency sub-band LL3With 9 high-frequency sub-band HLL,LHLAnd HHL, L=1,2,3, it is assumed thatAfter 3 grades of Haar wavelet pyramid transformation:
Step 8: note blockk, k=0,1 ..., the HL of m × n-1LAnd LHLAll coefficients of frequency band areWherein num isWithFrequency band coefficient number, byCalculate fusion coefficients weightk;
Such as: block0HL3And LH3All coefficients of frequency band areByCalculate fusion coefficients weight0=48.5-41.5=7;
Step 9: calculating variable quantity by formula (16) and formula (17), modifies by formula (18)It is embedded in watermark, wherein q is amount
Change step-length;
Such as: q=30,w′0,0=
0, then
Δ0=weight '0-weight0=15-7=8,
Step 10: each piecemeal is subjected to L grades of wavelet pyramid inverse transformations, reconstruct contains watermarking images;
Such as: first piecemeal carries out 3 grades of wavelet pyramid inverse transformationsRepeating the above steps to each piecemeal, it is small to obtain 3 grades of Haar of progress
Wave against pyramid transform the A ' containing watermarking images, as shown in Figure 4.
A kind of openly sensitivity watermark extracting method of combination non-overlap piecemeal and equal frequency bands fusion, comprising the following steps:
Step 1: the A ' containing watermarking images=(P ' of certification to be extracted is inputtedi,j)m×nAs shown in figure 4, and being covered before insertion watermark
Body image A=(Pi,j)m×nCorresponding 160 SHA-1 values are Asha-1=01110001011001111010111111110001
001111111010011011111110100001101000 010100100100100111101100000000101100010
11001010110111010010001101100 0000001000111001, from Asha-1Left side highest order start it successively
40 16 system numbers are converted to, 16 system sequences is constituted, is denoted as Sa=< 7,1,6,7,10,15,15,1,3,15,10,6,15,
14,8,6,8,5,2,4,9,14,12,0,2,12,5,9,5,11,10,4,6,11,6,12,0,2,3,9 >;
Step 2: by user key xinit∈ (0,1) and μinit∈ [3.57,4] is as the initial value x and parameter μ in formula (1)
And drive-type (1) generates the pseudo-random sequence that length is 40It is temporary to eliminate
State effect abandons count >=0 time result before iteration, and S is generated from count+1 timesb, count gives by user;
Such as: xinit=0.7635891325647968, μinit=3.9836387425412795, count=3000 are produced
The pseudo-random sequence that growth degree is 40 b0=
0.2037621195996894,b1=0.6463179714356886, b2=0.9106241678977656, b3=
0.3242195643220826,b4=0.8728201819600013, b5=0.4422042644778347, b6=
0.9826029497351249,b7=0.06809788586437480, b8=0.2528039605978175, b9=
0.7524859271167631,b10=0.7419561282150838, b11=0.7626964470112642, b12=
0.7210010734871887,b13=0.8013428956440323, b14=0.6341652481682873, b15=
0.9242029381397697,b16=0.2790613324584196, b17=0.8014527630757558, b18=
0.6339014212917692,b19=0.9244846737878244, b20=0.2781088223085844, b21=
0.7997724645639421,b22=0.6379258439511960, b23=0.9201267809267446, b24=
0.2927715059146659,b25=0.8248377026854120, b26=0.5755579855393184, b27=
0.9731670554892884,b28=0.1040245099033162, b29=0.3712887199633916, b30=
0.9299143615039163,b31=0.2596282443581546, b32=0.7657406922318214, b33=
0.7145926247730029,b34=0.8124631430251899, b35=0.6069742248179939, b36=
0.9503229763343125,b37=0.1880644657881021, b38=0.6082865877745214, b39=
0.9491975971289682;
Step 3: by SaAnd SbSubstitution formula (2) and formula (3) calculate the intermediate initial value x for being protected to watermark0And ginseng
Number μ0;
Such as: By formula (2), formula (3) be can be calculated:
Step 4: by SaAnd SbLogistic chaotic maps initial value x is generated by formula (6)~(9)1,y1And parameter μ1,μ2, so
Afterwards by initial value x1,y1And parameter μ1,μ2Pseudo random number is generated by formula (10) and formula (11), then it is pressed into formula (12) and formula (13)
Generate random coordinates, judge its whether with it is all Chong Die as the piecemeal of starting point using the random coordinates generated, if so, iterative
(10)~formula (13) continues to judge, otherwise, size is carried out by starting point of the random coordinates for the piecemeal of el × el, until generating 64
× 64 piecemeals, wherein the specific production method of each piecemeal is: the random coordinates point for remembering generation is (X, Y), the size of selection
Piecemeal for el × el is Bk=(bkr,s)el×el, then it is that starting point chooses piecemeal Bk=with random coordinates (X, Y) in step 6
(bkr,s)el×elSpecific method is bkr,s=P '(X+r)modM,(Y+s)modN, wherein P '(X+r)modM,(Y+s)modNFor the A ' containing watermarking images=
(P′i,j)M×NOn pixel.
Such as: el=8,
By taking first group of pseudo random number of generation as an example: yt+1=μ1xt(1-xt)=0.9707855550412292, xt+1=
μ2yt(1-yt)=0.02468364164788569, then random coordinates of first piecemeal starting point are as follows: Piecemeal rises
Point coordinate (648,516), bk0,0=P648,516,bk0,1=P648,517,…,bk7,7=P655,523;;
Step 5: successively to each piecemeal block of generation0,block1,…,blockm×n-1Carry out L grades of wavelet pyramids
Transformation forms 3L+1 subband, and meets
Such as: if L=3,3 grades of Haar wavelet pyramids are carried out to first piecemeal and convert to form 10 subbands, form 1
Low frequency sub-band LL3With 9 high-frequency sub-band HLL,LHLAnd HHL, L=1,2,3,
After 3 grades of Haar wavelet pyramid transformation:
Step 6: note blockk, k=0,1 ..., the HL of m × n-1LAnd LHLAll coefficients of frequency band areWherein num isWithThe number of frequency band coefficient, byCalculate fusion coefficients weightk;
Such as: block0HL3And LH3All coefficients of frequency band areBy
Calculate fusion coefficients weight0=52.5-37.5=15;
Step 7: by formula (20) in the watermark figure EW '=(ew ' for extracting encryptionii,jj)m×n, q is quantization step;
Such as:
Step 8: by x0And μ0Substitution formula (1) generates the pseudo-random sequence S that length is 64 × 64c=< c0,…,ck,…,
c4095>, ck∈ (0,1) abandons count=3000 result before iteration, wherein c0=0.4993912172500605, c1
=0.9764121237695018 ..., c4095=0.4993912172500605;
Step 9: formula (4) are pressed by ScIt is converted into matrix E=(eii,jj)m×n, by formula (21) to EW ' carry out exclusive or decryption, obtain
Watermark figure EW=(ew after decryptionii,jj)m×n, to decryption watermark EW=(ewii,jj)m×nW=(w is printed with raw waterii,jj)m×n(such as
Shown in Fig. 5) compare to obtain authentication result figure method particularly includes: if ewi,j=wi,j, then blockk, k=i × n+j is by recognizing
Card, otherwise, by block labeled as white in authentication result figure;
Such as:ew′0,0=
0,ew0,0=w0,0, block0Pass through certification.
Fig. 6 is the watermarking images extracted from Fig. 4, is 82 value images of 64 × 64 resolution ratio;
To verify proposed method to cover-image SHA-1 value, the sensibility of user key, Fig. 7~14 are respectively to repair
Change the watermark of one of them corresponding extraction, Figure 15~23 are respectively to modify one of them corresponding authentication image;From mentioning
The watermark and authentication image taken can be seen that the method for the present invention, the key and cover-image feature extreme sensitivity provided user.
To verify the robustness that proposed method operates some image attacks, Figure 24~28 are respectively to pass through JPEG
Compress, cut out, noise, covering, scribble aqueous impression, Figure 29~33 be respectively corresponding extraction watermark, Figure 34~38
Respectively corresponding authentication image;Watermark and authentication image from extraction can be seen that the method for the present invention, still have tradition
The feature of robust watermarking, therefore relative to traditional robust watermarking method, have higher safety and better practical application valence
Value.
Claims (10)
1. a kind of combination non-overlap piecemeal and the watermark embedding method for waiting frequency bands fusion, it is characterised in that the following steps are included:
Step 1: note cover-image A=(Pi,j)M×NCorresponding 160 SHA-1 values are Asha-1, by Asha-1Be converted to 40 16 into
Number processed constitutes 16 system sequences, is denoted as
Step 2: by user key xinit∈ (0,1) and μinit∈ [3.57,4] is as the initial value x and parameter μ in formula (1) and drives
Dynamic formula (1) generates the pseudo-random sequence that length is 40To eliminate transient state effect
It answers, count >=0 time result before iteration is abandoned, S is generated from count+1 timesb, count gives by user;
X=μ x (1-x) (1);
Step 3: by SaAnd SbSubstitution formula (2) and formula (3) calculate the intermediate initial value x for watermark protection0And parameter μ0;
Step 4: 2 value watermark figures of note are W=(wii,jj)m×n, by x0And μ0Substitution formula (1) generates the pseudorandom sequence that length is m × n
Arrange Sc=< c0,…,ck,…,cm×n-1>, ck∈ (0,1), is similarly elimination transient effect, by count >=0 time result before iteration
It abandons, S is generated from count+1c, count gives by user;
Step 5: by pseudo-random sequence Sc2 values are carried out, are scanned as 2 value matrix E=(eii,jj)m×n, W is encrypted and is added
Watermark figure W '=(w ' after closeii,jj)m×n;
Step 6: S is utilizedaAnd SbThe random coordinates of [0, M-1] × [0, N-1] range are generated, and by the coordinate in cover-image A
=(Pi,j)M×NM × n size of upper interception for el × el not overlap partition, meet el be even number and
Wherein M × N is cover-image A=(Pi,j)M×NSize;
Step 7: note block0,block1,…,blockm×n-1For each piecemeal of generation, successively each piecemeal is carried out L grades small
Wave pyramid transform forms 3L+1 subband, and meets
Step 8: note blockk, k=0,1 ..., the HL of m × n-1LAnd LHLAll coefficients of frequency band areWherein num isWithThe number of frequency band coefficient, byFusion coefficients weight is calculated by formula (15)k;
Step 9: fusion coefficients variable quantity is calculated using formula (16) and formula (17), is modified by formula (18)It is embedded in watermark, wherein q
For quantization step:
Δk=weight 'k-weightk (17)
Step 10: each piecemeal is subjected to L grades of wavelet pyramid inverse transformations, reconstructs the A ' containing watermarking images=(P 'i,j)M×N。
2. a kind of combination non-overlap piecemeal as described in claim 1 and the watermark embedding method for waiting frequency bands fusion, feature exist
In: step 1 by Asha-140 16 system numbers are converted to, the specific method for constituting 16 system sequences is from Asha-1Highest from left to right
Position starts successively by Asha-1Be converted to 40 16 system numbers.
3. a kind of combination non-overlap piecemeal as described in claim 1 and the watermark embedding method for waiting frequency bands fusion, feature exist
In:
Step 5 is by pseudo-random sequence Sc2 values are carried out, are scanned as 2 value matrix E=(eii,jj)m×nSpecific method be by formula (4)
By pseudo-random sequence ScIt is converted into 2 value matrix E=(eii,jj)m×n:
Step 5 encrypts W to obtain encrypted watermark figure W '=(w 'ii,jj)m×nMethod particularly includes:
Step 6 utilizes SaAnd SbThe random coordinates of [0, M-1] × [0, N-1] range are generated in cover-image A=(Pi,j)M×NIt is upper to cut
The specific method for taking the not overlap partition that m × n size is el × el is by SaAnd SbIt is mixed that Logistic is generated by formula (6)~(9)
Ignorant mapping initial value x1,y1And parameter μ1,μ2, then by initial value x1,y1And parameter μ1,μ2It is generated by formula (10) and formula (11) pseudo-
Random number, then by its by formula (12) and formula (13) generation random coordinates, judge its whether the random coordinates with all to have generated
It is overlapped for the piecemeal of starting point, if so, iterative (10)~formula (13) continues to judge, otherwise, is taken using the random coordinates as starting point
Size is the piecemeal of el × el, until generating m × n piecemeal:
yt+1=μ1xt(1-xt) (10)
xt+1=μ2yt(1-yt) (11)
The random coordinates point that note generates is (X, Y), and the piecemeal that the size of selection is el × el is Bk=(bkr,s)el×el, then step 6
In with random coordinates (X, Y) be starting point choose piecemeal Bk=(bkr,s)el×elSpecific method is bkr,s=P(X+r)modM,(Y+s)modN,
Middle P(X+r)modM,(Y+s)modNFor the pixel on cover-image A.
4. a kind of combination non-overlap piecemeal as described in claim 1 and the watermark embedding method for waiting frequency bands fusion, feature exist
In:
The specific method that step 7 successively carries out the transformation of L grades of wavelet pyramids to each piecemeal is to carry out L grades by Haar wavelet transformation
Wavelet pyramid transformation, shown in the specific method of Haar wavelet transformation such as formula (14):
In formula (14), f0,f1,f2,f3It is 4 elements that 2 × 2 fritter of matrix is adjacent up and down before converting, f0′,f1′,f2′,
f3' it is transformed coefficient, transformed all coefficients are put together by locating frequency band, that is, are constituted to original matrix element
Level-one pyramid decomposition, wherein f0′,f1′,f2′,f3' respectively correspond as LL, HL, LH, HH subband.
5. a kind of combination non-overlap piecemeal as described in claim 1 and the watermark embedding method for waiting frequency bands fusion, feature exist
In:
Each piecemeal is carried out L grades of wavelet pyramid inverse transformations by step 10, reconstructs the A ' containing watermarking images=(P 'i,j)M×NIt is specific
Method is the inverse transformation converted using formula (19) Haar small echo pyramid:
In formula (19), f0,f1,f2,f3It is 4 elements that 2 × 2 fritter of matrix is adjacent up and down after converting, f0′,f1′,f2′,
f3' it is LL, HL, LH before converting, HH subband corresponding position element.
6. the watermark extracting method of a kind of combination non-overlap piecemeal corresponding with claim 1 and equal frequency bands fusion, feature
Be the following steps are included:
Step 1: the A ' containing watermarking images=(P ' of certification to be extracted is inputtedi,j)M×NWith cover-image corresponding 160 before insertion watermark
Position SHA-1 value is Asha-1, by Asha-140 16 system numbers are converted to, 16 system sequences are constituted
Step 2: by user key xinit∈ (0,1) and μinit∈ [3.57,4] is as the initial value x and parameter μ in formula (1) and drives
Dynamic formula (1) generates the pseudo-random sequence that length is 40To eliminate transient state effect
It answers, count >=0 time result before iteration is abandoned, S is generated from count+1 timesb, count gives by user;
X=μ x (1-x) (1);
Step 3: by SaAnd SbSubstitution formula (2) and formula (3) calculate the intermediate initial value x for being protected to watermark0And parameter μ0;
Step 4: S is utilizedaAnd SbThe random coordinates of [0, M-1] × [0, N-1] range are generated in the A ' containing watermarking images=(P 'i,j)M×N
M × n size of upper interception for el × el not overlap partition, meet el be even number and
Step 5: successively to each piecemeal block of generation0,block1,…,blockm×n-1Carry out L grades of wavelet pyramid transformation
3L+1 subband is formed, and is met
Step 6: note blockk, k=0,1 ..., the HL of m × n-1LAnd LHLAll coefficients of frequency band areWherein num isWithThe number of frequency band coefficient, byFusion coefficients weight is calculated by formula (15)k;
Step 7: watermark figure EW '=(ew ' of encryption is extractedii,jj)m×n;
Step 8: for by x0And μ0Substitution formula (1) generates the pseudo-random sequence S that length is m × nc=< c0,…,ck,…,cm×n-1
>, ck∈ (0,1), is similarly elimination transient effect, and count >=0 time result before iteration is abandoned, and S is generated from count+1c,
Count is inputted by user;
X=μ x (1-x) (1);
Step 9: by pseudo-random sequence Sc2 values are carried out, then scanning is 2 value matrix E=(eii,jj)m×n, EW ' is decrypted
Watermark figure EW=(ew after being decryptedii,jj)m×n, and by decryption watermark EW=(ewii,jj)m×nW=is printed with raw water
(wii,jj)m×nComparison, obtain authentication result figure.
7. a kind of combination non-overlap piecemeal as claimed in claim 6 and the watermark extracting method for waiting frequency bands fusion, feature exist
In: step 1 by Asha-140 16 system numbers are converted to, 16 system sequences are constitutedSpecific method be from Asha-1Highest order from left to right start successively
By Asha-1Be converted to 40 16 system numbers.
8. a kind of combination non-overlap piecemeal as claimed in claim 6 and the watermark extracting method for waiting frequency bands fusion, feature exist
S is utilized in step 4aAnd SbThe random coordinates of [0, M-1] × [0, N-1] range are generated in the A ' containing watermarking images=(P 'i,j)M×NOn
The not overlap partition that m × n size of interception is el × el method particularly includes:
By SaAnd SbLogistic chaotic maps initial value x is generated by formula (6)~(9)1,y1And parameter μ1,μ2, then by initial value
x1,y1And parameter μ1,μ2Pseudo random number is generated by formula (10) and formula (11), then it is generated into random sit by formula (12) and formula (13)
Mark, judge its whether with it is all Chong Die as the piecemeal of starting point using the random coordinates generated, if so, iterative (10)~formula
(13) continue to judge, otherwise, size is carried out by starting point of the random coordinates for the piecemeal of el × el, until generating m × n points
Block:
yt+1=μ1xt(1-xt) (10)
xt+1=μ2yt(1-yt) (11)
The random coordinates point that note generates is (X, Y), and the piecemeal that the size of selection is el × el is Bk=(bkr,s)el×el, then step 4
In with random coordinates (X, Y) be starting point choose piecemeal Bk=(bkr,s)el×elSpecific method is bkr,s=P '(X+r)modM,(Y+s)modN,
Wherein P '(X+r)modM,(Y+s)modNFor the A ' containing watermarking images=(P 'i,j)M×NOn pixel.
9. a kind of combination non-overlap piecemeal as claimed in claim 6 and the watermark extracting method for waiting frequency bands fusion, feature exist
In: step 5 carries out L grades of wavelet pyramids and converts the specific method to form 3L+1 subband as by l grades of Haar wavelet transformation progress
Wavelet pyramid transformation, shown in the specific method of Haar wavelet transformation such as formula (14):
Step 7 extracts encrypted watermark figure EW '=(ew 'ii,jj)m×nSpecific method be by formula (20) extracting encryption
Watermark figure EW '=(ew 'ii,jj)m×n, q is quantization step;
10. a kind of combination non-overlap piecemeal as claimed in claim 6 and the watermark extracting method for waiting frequency bands fusion, feature exist
In: step 9 by pseudo-random sequence ScBinaryzation, then scanning is 2 value matrix E=(eii,jj)m×nSpecific method be by formula (4)
By ScIt is converted into matrix E=(eii,jj)m×n;
The specific method that EW ' is decrypted is by formula (21) to EW ' carry out exclusive or decryption by step 9;
Step 9 is to decryption watermark EW=(ewii,jj)m×nW=(w is printed with raw waterii,jj)m×nComparison obtain the tool of authentication result figure
Body method is if ewii,jj=wii,jj, then blockk, k=ii × n+jj is by certification, otherwise, by block in authentication result figure
Labeled as white.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610052872.1A CN105719226B (en) | 2016-01-26 | 2016-01-26 | A kind of combination non-overlap piecemeal and the watermark insertion and extracting method for waiting frequency bands to merge |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610052872.1A CN105719226B (en) | 2016-01-26 | 2016-01-26 | A kind of combination non-overlap piecemeal and the watermark insertion and extracting method for waiting frequency bands to merge |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105719226A CN105719226A (en) | 2016-06-29 |
CN105719226B true CN105719226B (en) | 2019-03-15 |
Family
ID=56155030
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610052872.1A Active CN105719226B (en) | 2016-01-26 | 2016-01-26 | A kind of combination non-overlap piecemeal and the watermark insertion and extracting method for waiting frequency bands to merge |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105719226B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108537716B (en) * | 2018-01-24 | 2022-05-03 | 重庆邮电大学 | Color image encryption embedding method based on discrete domain |
CN110879894B (en) * | 2019-11-11 | 2023-04-07 | 陕西师范大学 | Image encryption and decryption method based on lazy scale transformation and random layered fusion |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1377184A (en) * | 2002-04-30 | 2002-10-30 | 中山大学 | Method for embedding digital watermark |
CN102892048A (en) * | 2012-09-18 | 2013-01-23 | 天津大学 | Video watermark anti-counterfeiting method capable of resisting geometric attacks |
CN104732474A (en) * | 2015-03-30 | 2015-06-24 | 陕西师范大学 | Digital watermark embedding and extracting method based on multi-level wavelet coefficient weighting and quantification |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
KR100576802B1 (en) * | 2003-10-29 | 2006-05-10 | 한국전자통신연구원 | The Method for embedding and extracting blindly watermarks by using wavelet transform and HVS |
CN101309410A (en) * | 2008-07-08 | 2008-11-19 | 武汉大学 | Video coding and decoding architecture |
-
2016
- 2016-01-26 CN CN201610052872.1A patent/CN105719226B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1377184A (en) * | 2002-04-30 | 2002-10-30 | 中山大学 | Method for embedding digital watermark |
CN102892048A (en) * | 2012-09-18 | 2013-01-23 | 天津大学 | Video watermark anti-counterfeiting method capable of resisting geometric attacks |
CN104732474A (en) * | 2015-03-30 | 2015-06-24 | 陕西师范大学 | Digital watermark embedding and extracting method based on multi-level wavelet coefficient weighting and quantification |
Non-Patent Citations (1)
Title |
---|
对基于量化水印的两频带滤波攻击的估计;王津申等;《电子科技大学学报》;20080130;第37卷(第1期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN105719226A (en) | 2016-06-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Kashyap et al. | Image watermarking using 3-level discrete wavelet transform (DWT) | |
Gunjal et al. | Discrete wavelet transform based strongly robust watermarking scheme for information hiding in digital images | |
CN104732474B (en) | The digital watermark embedding and extracting method quantified based on multilevel wavelet coefficient weighted sum | |
Khalifa et al. | A robust non-blind algorithm for watermarking color images using multi-resolution wavelet decomposition | |
CN103279917A (en) | Transforming domain geometric-attack-resistant gray level image digital watermark technology | |
CN111882477A (en) | Self-adaptive zero-watermarking method combining visual password and enhanced singular value decomposition | |
CN103020496A (en) | Digital watermark encryption realization method | |
Joseph et al. | Image security enhancement using DCT & DWT watermarking technique | |
Shoaib et al. | Authenticating using secret key in digital video watermarking using 3-level DWT | |
CN105719226B (en) | A kind of combination non-overlap piecemeal and the watermark insertion and extracting method for waiting frequency bands to merge | |
CN108985082B (en) | Medical image encryption method based on digital watermark | |
Al-Dabbas et al. | Digital Watermarking, Methodology, Techniques, and Attacks: A Review | |
CN111861847A (en) | GF-2 image double watermarking method applying DWT and SIFT | |
Jin | A wavelet-based method of zero-watermark utilizing visual cryptography | |
Kekre et al. | Robust Watermarking Technique using Hybrid Wavelet Transform Generated from Kekre Transform and Discrete Cosine Transform | |
Sadreazami et al. | Highly robust image watermarking in contourlet domain using singular value decomposition | |
Tsai et al. | A wavelet-based semi-fragile watermarking with recovery mechanism | |
Abu-Errub et al. | Optimized DWT-based image watermarking | |
Surekha et al. | Digital image ownership verification based on spatial correlation of colors | |
JP4257444B2 (en) | Digital watermark insertion / extraction apparatus and method | |
Hien et al. | RDWT domain watermarking based on independent component analysis extraction | |
Patvardhan et al. | A robust wavelet packet based blind digital image watermarking using HVS characteristics | |
Mehta et al. | Secure audio watermarking based on Haar wavelet and discrete cosine transform | |
Shen et al. | Blind watermarking algorithm based on henon chaos system and lifting scheme wavelet | |
Sun et al. | High-definition digital color image watermark algorithm based on QR code and DWT |
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 |