CN117173001A - Robust reversible watermark embedding and extracting method based on attack simulation - Google Patents
Robust reversible watermark embedding and extracting method based on attack simulation Download PDFInfo
- Publication number
- CN117173001A CN117173001A CN202311169752.6A CN202311169752A CN117173001A CN 117173001 A CN117173001 A CN 117173001A CN 202311169752 A CN202311169752 A CN 202311169752A CN 117173001 A CN117173001 A CN 117173001A
- Authority
- CN
- China
- Prior art keywords
- watermark
- image
- robust
- embedding
- harmonic transformation
- 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.)
- Pending
Links
- 230000002441 reversible effect Effects 0.000 title claims abstract description 68
- 238000000034 method Methods 0.000 title claims abstract description 66
- 238000004088 simulation Methods 0.000 title claims abstract description 47
- 230000009466 transformation Effects 0.000 claims abstract description 128
- 238000010606 normalization Methods 0.000 claims abstract description 32
- 238000013139 quantization Methods 0.000 claims abstract description 25
- 230000003044 adaptive effect Effects 0.000 claims description 12
- 238000000605 extraction Methods 0.000 claims description 11
- 238000012360 testing method Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 9
- 230000001965 increasing effect Effects 0.000 claims description 4
- 238000006467 substitution reaction Methods 0.000 claims description 3
- 230000006835 compression Effects 0.000 description 11
- 238000007906 compression Methods 0.000 description 11
- 235000002566 Capsicum Nutrition 0.000 description 4
- 241000758706 Piperaceae Species 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 3
- 239000003086 colorant Substances 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
Landscapes
- Editing Of Facsimile Originals (AREA)
- Image Processing (AREA)
Abstract
The invention relates to the technical field of digital watermarking, and discloses a robust reversible watermarking embedding and extracting method based on attack simulation, which comprises the following specific steps: acquiring an original image; performing self-adaptive normalization operation based on attack simulation based on an original image to obtain normalized polar harmonic transformation moment; performing the watermark embedding with the lowest embedding strength to obtain a normalized polar harmonic transformation moment with the watermark and quantization distortion of the normalized polar harmonic transformation moment; performing self-adaptive normalized inverse operation through normalized polar harmonic transformation moment to obtain an error image; obtaining candidate watermark images through the error images; carrying out attack simulation on the candidate watermark image; calculating rounding distortion; obtaining an intermediate image containing the watermark and the auxiliary information; and generating a hash value; robust reversible watermark embedding is accomplished by hash values. The invention solves the problems of insufficient watermark robustness and embedding capacity in the prior art, and has the characteristic of self-adaptive watermark embedding.
Description
Technical Field
The invention relates to the technical field of digital watermarks, in particular to a robust reversible watermark embedding and extracting method based on attack simulation.
Background
With the widespread use and spread of digital images, it is important to protect the copyright and integrity of digital images.
The digital watermarking technology plays roles of copyright protection and integrity authentication by embedding digital information into digital media. In recent years, there has been great progress in image robust reversible watermarking research capable of resisting conventional signal processing attacks such as noise, compression, filtering and the like, but there are still disadvantages of poor robustness and insufficient embedding capacity for geometric attacks such as rotation and scaling. The robust reversible watermark has the characteristics that when the carrier image is not attacked, the embedded watermark information can be correctly extracted and the carrier image can be completely recovered, and when the carrier image is attacked to a certain extent, the watermark information can still be correctly extracted without damage. The robust reversible watermark integrates the robust watermark and the reversible watermark technology, and is suitable for scenes needing integrity authentication and copyright protection, such as digital works of art and high-fidelity images. The technology not only can effectively extract watermark information after the image is attacked, but also can accurately recover the original image under the condition of not being attacked. However, the robust reversible watermark method capable of simultaneously resisting common signal processing and geometric deformation attacks still has a shortage in terms of both robustness and embedding capacity, and there is no robust reversible watermark embedding method and extraction method that utilize target attacks as a priori knowledge to improve robustness and embedding capacity.
The prior art has a reversible robust watermarking method based on Zernike moment to resist geometric attack by calculating the Zernike moment of an image and embedding the watermark on the Zernike moment. Judging whether the image is attacked or not by comparing the hash value of the watermark image with the hash value extracted from the watermark image, and extracting watermark information and recovering the original image by using the Zernike moment of the image when judging that the image is not attacked; when the image is judged to be attacked, the Zernike moment of the attacked image with watermark information is calculated, and the Zernike moment of the attacked image with watermark information is utilized to extract watermark information.
However, the problem of insufficient watermark robustness and embedding capacity still exists in the prior art, and how to invent a robust reversible watermark embedding and extracting method and system with sufficient watermark robustness and embedding capacity is a technical problem to be solved in the technical field.
Disclosure of Invention
The invention provides a robust reversible watermark embedding and extracting method and system based on attack simulation, which have the characteristic of self-adaptive watermark embedding, and aims to solve the problems of insufficient watermark robustness and embedding capacity in the prior art.
In order to achieve the above purpose of the present invention, the following technical scheme is adopted:
a robust reversible watermark embedding method based on attack simulation comprises the following specific steps:
s1: acquiring an original image I;
s2: calculating an n-order M-order harmonic transformation moment M of an original image I n,l ;
S3: for said polar harmonic transformation moment M n,l Performing self-adaptive normalization operation based on attack simulation to obtain normalized polar harmonic transformation moment
S4: for normalized polar harmonic transformation momentWatermark embedding with lowest embedding strength to obtain normalized polar harmonic transformation moment with watermark>And quantization distortion d thereof quantified ;
S5: for normalized polar harmonic transformation moment with watermarkPerforming self-adaptive normalized inverse operation to obtain polar harmonic transformation moment with watermark>
S6: obtaining polar harmonic transformation moments with watermarksAnd polar harmonic transformation moment M n,l To obtain an error imageSuperimposed error image +.>And the original image I, a candidate watermark image is obtained>
S7: for candidate watermark imagePerforming attack simulation and extracting watermark information in the attacked image>If watermark bit is in the extracted watermark information>Extracting errors and watermark bits +.>If the embedding strength level of the watermark bit is smaller than or equal to the set threshold value, increasing the embedding strength level of the watermark bit, and returning to the step S4; otherwise, candidate watermark image +.>As a robust watermark image->
S8: for robust watermark imagePerforming a dewatering operationObtaining a water-jet printing image I watermarking-removal ;
S9: calculation of the original image and the watermark image I watermarking-removal Rounding distortion d between rounding ;
S10: to quantize distortion d quantified Rounding distortion d rounding Robust watermark imageThe least significant bit of the first several pixels of (a) is used as auxiliary information; embedding auxiliary information into a robust watermark image using a reversible watermark method>In which watermark and auxiliary information w are contained 2 Intermediate image +.>
S11: intermediate imageThe least significant bit of the first several pixels of (a) is replaced by a 0 value and a hash value H is generated 1 The method comprises the steps of carrying out a first treatment on the surface of the Substitution of intermediate images by hash values +.>The least significant bit of the first several pixels of (a) to obtain a robust reversible watermark pattern, and completing the embedding of the robust reversible watermark>
Preferably, in the step S2, an n-order M-order harmonic transformation moment M of the original image I is calculated n,l The method specifically comprises the following steps:
setting an order M, and taking the center of an original image I with the size of K multiplied by K as a circle center to make an inscribed circle; taking the inscribed circle as a unit circle, solving polar harmonic transformation moment of pixels in the unit circle; constructing polar harmonic transform basis V based on the inscribed circle n,l (x,y)The method comprises the steps of carrying out a first treatment on the surface of the Through V n,l (x, y) calculating the polar harmonic transformation moment to obtain an n-order l-order polar harmonic transformation moment M n,l :
Wherein Deltax is s And Deltay t Representing the step sizes of the x-axis and the y-axis of the image unit circle, f (x s ,y t ) Representing pixels within a unit circle, (·) * Representing the complex number of conjugates, V n,l (x s ,y t ) Is a polar harmonic transformation basis in a circle.
Further, in the step S3, the polar harmonic transformation moment M n,l Performing self-adaptive normalization operation based on attack simulation to obtain normalized polar harmonic transformation momentThe method comprises the following specific steps:
s301: 1000 images are randomly selected from the BOWS2 database, and are subjected to attack simulation test, and the test result is fitted into a binary cubic polynomial by a double square polynomial fitting method:
wherein T is ni,li Representing the fitting result and simultaneously being self-adaptive normalized weight; n is n i And l i Polar harmonic transformation moments M for embedding the ith watermark bit, respectively ni,li The number of steps and the number of times; p is p 1 ~p 10 The fitting operation is carried out;
s302: selecting polar harmonic transform moment M for watermark embedding n,l Performing self-adaptive normalization to obtain normalized polar harmonic transformation moment
Wherein M is 00 Is the polar harmonic transformation moment of 0 order 0, T ni,li Is an adaptive normalized weight.
Further, in the step S4, the polar harmonic transformation moment is normalizedWatermark embedding with lowest embedding strength to obtain normalized polar harmonic transformation moment with watermark>And quantization distortion d thereof quantified The method comprises the following specific steps of;
s401: grouping normalized polar harmonic transformation moments; each group comprises P normalized polar harmonic transformation moments for embedding a watermark bit; the normalized polar harmonic transformation moment of the embedded ith watermark bit is expressed as:
the specific embedding formula for embedding the watermark is as follows:
wherein,is normalized polar harmonic transformation moment->Transpose of->For normalized polar harmonic transformation moment with watermark, < >>Is a multi-stage quantizer with embedded intensity level of z-stage, u is a random vector with length of P; the multi-stage quantizer is specifically:
wherein [ (S)]Is a rounding operation, (. Cndot.) for the purpose of handling the round-off operations t Is a transpose operation, delta is the quantization step size, Q w (. Cndot.) is a single-stage quantizer; q (Q) w The (-) is specifically as follows:
where w is the embedded watermark bit, β i (w) is constrained to be beta i (1)=β i (0) A jitter value of +Δ/2;
s402: obtaining quantization distortion d corresponding to the ith watermark bit quantified-i :
Wherein,is a multi-level quantizer with embedded intensity level z.
Further, in the step S5, the normalized polar harmonic transformation moment with watermark is obtainedPerforming self-adaptive normalized inverse operation to obtain polar harmonic transformation moment with watermark>The method comprises the following steps:
the inverse operation of the adaptive normalization operation is adopted and is carried withNormalized polar harmonic transformation moment of watermarkMultiplying by M 00 Self-adaptive normalized weight T ni,li Divided by 10 3 Obtaining polar harmonic transformation moment with watermark:
further, in the step S6, polar harmonic transformation moments with watermarks are obtainedAnd polar harmonic transformation moment M n,l To obtain an error image +.>The method comprises the following steps: obtaining polar harmonic transformation moment with watermark>Polar harmonic transformation moment M with original image ni,li Is a difference in (2); performing polar harmonic transformation and inverse transformation on the difference value and rounding operation to obtain an error image I error :
Wherein V is ni,li For polar harmonic transform basis for embedding the ith watermark bit, V ni,-li Is V (V) ni,li Is used to determine the complex number of the conjugate,is->Complex conjugate of M ni,-li) Is M ni,li Is a complex conjugate of (a) and (b).
Further, in the step S8, the robust watermark image is subjected to a watermark removal operation to obtain a watermark image, which specifically includes the steps of:
s801: computing robust watermark imagesPolar harmonic transformation moment->Performing adaptive normalization to obtain ++>And extracting watermark information->
Wherein,and->Respectively robust watermark image +.>And a watermark image I watermarking-removal The normalized polar harmonic transformation moment for embedding the ith watermark bit;
s802: the inverse operation of the self-adaptive normalization operation is adopted to obtain
By passing throughCalculation of the watermark image I watermarking-removal :
Where P x L represents the number of polar harmonic transformed moments for which watermark embedding is totally modified.
The robust reversible watermark extraction method based on attack simulation comprises the following specific steps:
s21: extracting least significant bits of first S pixels of the robust reversible watermark image to obtain a hash value H 1 ;
S22: replacing least significant bits of first S pixels of the robust reversible watermark image with 0 value, and generating hash value H of the image at the moment 2 ;
S23: if the hash value H 1 Equal to hash value H 2 Judging that the robust reversible watermark image is not attacked, and executing step S05; if the hash value H 1 Not equal to hash value H 2 When the robust reversible watermark image is attacked, executing a step S06;
s24, calculating quantization distortion of the robust reversible watermark image which is not attacked, extracting watermark information in the robust watermark image according to the quantization distortion, and recovering the robust reversible watermark image which is not attacked into an original image;
s25: robust reversible watermark image after computing is attackedPolar harmonic transformation moment->For->Performing self-adaptive normalization operation, and extracting waterAnd printing information.
Preferably, the step S24 specifically includes the steps of:
s051: extraction by reversible watermarkingIs quantized distortion d of (2) quantified The method comprises the steps of carrying out a first treatment on the surface of the Combining robust watermark imagesIs the rounding distortion d of (2) rounding And the least significant bits of the first S pixels, recovering the robust watermark image +.>
S052: computing robust watermark imagesPolar harmonic transformation moment->Performing self-adaptive normalization operation to obtain +.>And calculate the ith watermark information w i :
w∈{0,1}
S053: according to quantization distortion d quantified Extracting a robust watermark imageWatermark information in the image is obtained to obtain a watermark image I watermarking-removal :
S054: computing robust watermark imagesPolar harmonic transformation moment->Will->Performing adaptive normalization to obtain ++>Extracting its quantization distortion:
wherein,and->Respectively robust watermark image +.>And a watermark image I watermarking-removal The normalized polar harmonic transformation moment for embedding the ith watermark bit;
s055: by passing throughCalculation of the watermark image I watermarking-removal :
S056: using de-watermarked image I watermarking-removal And rounding distortion d rounding Restoring the original image I:
I=I watermarking-removal +d rounding 。
further, in the step S25, watermark information is extracted, specifically:
the beneficial effects of the invention are as follows:
the invention discloses a robust reversible watermark embedding method based on attack simulation. The robust reversible watermarking method based on attack simulation measures the stability of different orders and subharmonic transformation moments by carrying out attack simulation test on the image, and carries out self-adaptive normalization operation on the stability, thereby improving the robustness of the watermark under the condition of the same embedded distortion, realizing the extraction of the watermark and the restoration of the image when the watermark is not attacked, and effectively extracting the watermark when the watermark is attacked. The method provided by the invention can solve the problems of insufficient watermark robustness and embedding capacity in the prior art, and has the characteristic of self-adaptive watermark embedding.
Drawings
Fig. 1 is a schematic flow chart of a robust reversible watermark embedding method based on attack simulation.
Fig. 2 is a schematic diagram of an embedded watermark using a robust reversible watermark embedding method based on attack simulation of the present invention.
Fig. 3 is a schematic flow chart of a robust reversible watermark extraction method based on attack simulation.
Detailed Description
The invention is described in detail below with reference to the drawings and the detailed description.
Example 1
As shown in FIG. 1, the robust reversible watermark embedding method based on attack simulation comprises the following specific steps:
s1: acquiring an original image I;
s2: calculating an n-order M-order harmonic transformation moment M of an original image I n,l ;
S3: for said polar harmonic transformation moment M n,l Performing self-adaptive normalization operation based on attack simulation to obtain normalized polar harmonic transformation moment
S4: for normalized polar harmonic transformation momentWatermark embedding with lowest embedding strength to obtain normalized polar harmonic transformation moment with watermark>And quantization distortion d thereof quantified ;
S5: for normalized polar harmonic transformation moment with watermarkPerforming self-adaptive normalized inverse operation to obtain polar harmonic transformation moment with watermark>
S6: obtaining polar harmonic transformation moments with watermarksAnd polar harmonic transformation moment M n,l To obtain an error imageSuperimposed error image +.>And the original image I, a candidate watermark image is obtained>
S7: for candidate watermark imagePerforming attack simulation and extracting watermark information in the attacked image>If watermark bit is in the extracted watermark information>Extracting errors and watermark bits +.>If the embedding strength level of the watermark bit is smaller than or equal to the set threshold value, increasing the embedding strength level of the watermark bit, and returning to the step S4; otherwise, candidate watermark image +.>As a robust watermark image->
S8: for robust watermark imageCarrying out a dewatering operation to obtain a dewatering image I watermarking-removal ;
S9: calculation of the original image and the watermark image I watermarking-removal Rounding distortion d between rounding ;
S10: to quantize distortion d quantified Rounding distortion d rounding Robust watermark imageThe least significant bit of the first several pixels of (a) is used as auxiliary information; embedding auxiliary information into a robust watermark image using a reversible watermark method>In which watermark and auxiliary information w are contained 2 Intermediate image +.>
S11: intermediate imageThe least significant bit of the first several pixels of (a) is replaced by a 0 value and a hash value H is generated 1 The method comprises the steps of carrying out a first treatment on the surface of the Substitution of intermediate images by hash values +.>The least significant bit of the first several pixels of (a) to obtain a robust reversible watermark pattern, and completing the embedding of the robust reversible watermark>
Example 2
More specifically, in one embodiment, in the step S2, an n-order M-th harmonic transformation moment M of the original image I is calculated n,l The method comprises the following specific steps of:
s201: in this embodiment, n=38, using M n,l Represents the PHT moment of order n, where n and l satisfy:
0<n≤N
0<l≤L
base V of PHT moment n,l (x, y) is a set of perfect orthogonal bases on a unit circle, n is the order of the transform, l is the number of transforms, V n,l (x, y) is:
V n,l (x,y)=R n (r)e imθ
wherein,θ=tan -1 (y/x),R n (r) is:
S202: setting an order M, and taking the center of an original image I with the size of K multiplied by K as a circle center to make an inscribed circle; taking the inscribed circle as a unit circle, solving polar harmonic transformation moment of pixels in the unit circle; constructing polar harmonic transform basis V based on the inscribed circle n,l (x, y); through V n,l (x, y) calculating the polar harmonic transformation moment to obtain an n-order l-order polar harmonic transformation moment M n,l :
Wherein Deltax is s And Deltay t Representing the step sizes of the x-axis and the y-axis of the image unit circle, f (x s ,y t ) Representing pixels within a unit circle, (·) * Representing the complex number of conjugates, V n,l (x s ,y t ) Is a polar harmonic transformation base in a circle;
in this embodiment:
and k=512.
In a specific embodiment, in said step S3, said polar harmonic transformation moment M n,l Performing self-adaptive normalization operation based on attack simulation to obtain normalized polar harmonic transformation momentThe method comprises the following specific steps:
s301: randomly selecting 1000 images from the BOWS2 database, and performing attack simulation test on the images; selecting PHT moment Z nm Satisfies (0)<n≤N)&&PHT moment Z of (l+.4j) nm Performing self-adaptive normalization operation on the obtained product to obtain
Wherein M is 00 PHT moment, T, of 0 th order ni,li Randomly selecting 1000 images from a BOWS2 database, and respectively applying Gaussian noise attack with variance of 0.03, wherein the quality factor is 10, and the compression ratio is 100; calculating the average value of the PHT moments of each order and the times of the 1000 original images and the average value of the PHT moments of each order and the times of the attacked images; fitting the test result into a binary cubic polynomial by a double square polynomial fitting method:
wherein T is ni,li Representing the fitting result and simultaneously being self-adaptive normalized weight; n is n i And l i Polar harmonic transformation moments M for embedding the ith watermark bit, respectively ni,li The number of steps and the number of times; p is p 1 ~p 10 The fitting operation is carried out;
in this embodiment, p 1 To p 10 The values of (2) are p respectively 1 =0.13,p 2 =-3.59*10 -3 ,p 3 =6.59*10 -6 ,p 4 =1.14*10 -4 ,p 5 =-5.11*10 -7 ,p 6 =-1.43*10 -5 ,p 7 =-1.39*10 -6 ,p 8 =2.27*10 -8 ,p 9 =4.09*10 -7 ,p 10 =-1.29*10 -8 ;
S302: selecting polar harmonic transform moment M for watermark embedding n,l Performing self-adaptive normalization to obtain normalized polar harmonic transformation moment
Wherein M is 00 Is the polar harmonic transformation moment of 0 order 0, T ni,li Is an adaptive normalized weight.
In a specific embodiment, in the step S4, the polar harmonic transformation moment is normalizedWatermark embedding with lowest embedding strength to obtain normalized polar harmonic transformation moment with watermark>And quantization distortion d thereof quantified The method comprises the following specific steps of;
s401: embedding watermark information: normalized PHT moment using spread spectrum transform dither modulation based on multi-stage quantizerAbsolute value of +.>Watermark embedding is carried out to obtain normalized PHT moment with watermarkFirstly grouping normalized PHT moments, wherein each group of P normalized PHT moments is used for embedding a watermark bit; wherein the normalized PHT moment for embedding the ith watermark bit is
The normalized polar harmonic transformation moment of the embedded ith watermark bit is expressed as:
the specific embedding formula for embedding the watermark is as follows:
wherein,is normalized polar harmonic transformation moment->Transpose of->For normalized polar harmonic transformation moment with watermark, < >>Is a multi-stage quantizer with embedded intensity level of z-stage, u is a random vector with length of P; the multi-stage quantizer is specifically:
wherein [ (S)]Is a rounding operation, (. Cndot.) for the purpose of handling the round-off operations T Is a transpose operation, delta is the quantization step size, Q w (. Cndot.) is a single-stage quantizer; q (Q) w The (-) is specifically as follows:
where w is the embedded watermark bit, β i (w) is constrained to be beta i (1)=β i (0) A jitter value of +Δ/2;
s402: obtaining quantization distortion d corresponding to the ith watermark bit quantified-i :
Wherein,is a multi-level quantizer with embedded intensity level z.
In a specific embodiment, in the step S5, the normalized polar harmonic transformation moment with watermark is calculatedPerforming self-adaptive normalized inverse operation to obtain polar harmonic transformation moment with watermark>The method comprises the following steps:
normalized PHT moment with watermark by inverse operation of adaptive normalization operationPHT moment M multiplied by 0 th order 0 00 Multiplying by self-adaptive normalized weight T ni,li Finally divided by 10 3 Obtaining polar harmonic transformation moment with watermark:
in a specific embodiment, in the step S6, polar harmonic transformation moments with watermarks are obtainedAnd polar harmonic transformation moment M n,l To obtain an error image +.>The method comprises the following steps: obtaining polar harmonic transformation moment with watermark>Polar harmonic transformation moment M with original image ni,li Is a difference in (2); performing polar harmonic transformation and inverse transformation on the difference value and rounding operation to obtain an error image I error :
Wherein V is ni,li For polar harmonic transform basis for embedding the ith watermark bit, V ni,-li Is V (V) ni,li Is used to determine the complex number of the conjugate,is->Complex conjugate of M ni,-li) Is M ni,li Is a complex conjugate of (a) and (b).
In this embodiment, L represents the length of watermark information, and in this embodiment, p=2, l=512.
In this embodiment, the error image I is described in S6 error Superimposed with the original image I to obtain candidate watermark imageThe method comprises the following specific steps:
PHT inverse transformation is carried out to obtain candidate watermark imagesThe specific formula is as follows:
in a specific embodiment, in the step S8, the robust watermark image is obtainedCarrying out a dewatering operation to obtain a dewatering image I watermarking-removal The method comprises the following specific steps of:
s801: computing robust watermark imagesPolar harmonic variation of (2)Moment->Performing adaptive normalization to obtain ++>And extracting watermark information->
Wherein,and->Respectively robust watermark image +.>And a watermark image I watermarking-removal The normalized polar harmonic transformation moment for embedding the ith watermark bit;
s802: the inverse operation of the self-adaptive normalization operation is adopted to obtain
By passing throughCalculation of the watermark image I watermarking-removal :
Where P x L represents the number of polar harmonic transformed moments for which watermark embedding is totally modified.
In the present embodiment, in the step S9, the original image and the watermark image I are calculated watermarking-removal Rounding distortion d between rounding The method specifically comprises the following steps:
d rounding =I-I watermarking-removal 。
the robust reversible watermark embedding method based on attack simulation is shown in figure 2; compared with the prior art, the invention has higher robustness against conventional signal processing attacks and geometric deformation attacks and larger embedding capacity. The invention adopts the self-adaptive normalization operation based on the attack simulation, the stability of PHT moments with different orders and times is measured through the attack simulation test, and the corresponding self-adaptive normalization weight is obtained, so that the robustness of resisting the conventional signal processing attack and the geometric deformation attack can be improved by carrying out the self-adaptive normalization operation; the self-adaptive watermark embedding based on attack simulation used in the invention determines the optimal embedding strength of each watermark bit by carrying out attack simulation test on the candidate watermark image, and carries out watermark embedding by using a multi-stage quantizer, thereby not only enhancing the robustness under the same embedding distortion, but also improving the embedding capacity of the watermark.
Example 3
As shown in fig. 3, the robust reversible watermark extraction method based on attack simulation further comprises the following specific steps:
s21: extracting least significant bits of first S pixels of the robust reversible watermark image to obtain a hash value H 1 ;
S22: replacing least significant bits of first S pixels of the robust reversible watermark image with 0 value, and generating hash value H of the image at the moment 2 ;
S23: if the hash value H 1 Equal to hash value H 2 Judging that the robust reversible watermark image is not attacked, and executing step S05; if hashValue H 1 Not equal to hash value H 2 When the robust reversible watermark image is attacked, executing a step S06;
s24, calculating quantization distortion of the robust reversible watermark image which is not attacked, extracting watermark information in the robust watermark image according to the quantization distortion, and recovering the robust reversible watermark image which is not attacked into an original image;
s25: robust reversible watermark image after computing is attackedPolar harmonic transformation moment->For->And performing self-adaptive normalization operation and extracting watermark information.
In a specific embodiment, the step S24 specifically includes:
s051: extraction by reversible watermarkingIs quantized distortion d of (2) quantified The method comprises the steps of carrying out a first treatment on the surface of the Combining robust watermark imagesIs the rounding distortion d of (2) rounding And the least significant bits of the first S pixels, recovering the robust watermark image +.>
S052: for candidate watermark imageA gaussian noise attack with variance of 0.03 and a JPEG2000 attack with a quality factor of 10 and a compression ratio of 100 were applied, respectively. Computing candidate watermark images after attackIs>And calculate the ith watermark information w i :
w∈{0,1}
If the watermark information is extractedIf the embedding strength level is inconsistent with the original watermark information and is smaller than or equal to the set maximum watermark strength level Z, the embedding strength level is increased and a new candidate watermark image +.>Otherwise, candidate watermark image +.>Is considered as a robust watermark image +.>
S053: according to quantization distortion d quantified Extracting a robust watermark imageWatermark information in the image is obtained to obtain a watermark image I watermarking-removal :
S054: computing robust watermark imagesPolar harmonic transformation moment->Will->Performing adaptive normalization to obtain ++>Extracting its quantization distortion:
wherein,and->Respectively robust watermark image +.>And a watermark image I watermarking-removal The normalized polar harmonic transformation moment for embedding the ith watermark bit;
s055: by passing throughCalculation of the watermark image I watermarking-removal :
S056: using de-watermarked image I watermarking-removal And rounding distortion d rounding Restoring the original image I:
I=I watermarking-removal +d rounding 。
in a specific embodiment, in the step S25, watermark information is extracted, specifically:
the self-adaptive watermark embedding method of the invention determines the optimal embedding strength grade of each watermark bit by carrying out attack simulation test on the candidate watermark image, and adopts the multi-stage quantizer to endow different embedding strengths for different watermark bits, thereby not only enhancing the robustness under the same embedding distortion, but also realizing larger watermark embedding capacity, having the characteristic of unchanged rotation scaling based on the absolute value of polar harmonic transformation moment, being capable of effectively resisting rotation scaling attack, and being capable of effectively extracting watermark information and recovering images when not attacked.
Example 4
In the embodiment, gray images of a picture Lena, a picture Goldhill, a picture Peppers and a picture Barbara are used as experimental objects, and the robust reversible watermark embedding and extracting method based on attack simulation is used for carrying out win embedding and extracting on the images; the four groups of pictures have different characteristics, such as the picture Lena comprises a flat block, clear and fine lines, gradually changing light shadows, deep and shallow layers of colors and the like; the picture Goldhill has a large number of regular textures and irregular textures at the same time; the picture Peppers has more bright and dark areas, the colors inside the blocks are similar, and the color difference between the blocks is large; the picture barbera has a large number of regular textures and the like. Various pictures in daily life have the characteristics, so that the four groups of pictures are adopted as experimental objects, so that experimental results have popularization; the picture size selected in this embodiment is 512×512, and the watermark embedding is not greatly affected by the images with different sizes, so that the method can be popularized to various images.
The bit error rate results of the test pictures Lena under attack are shown in table 1; the watermark embedded in table 1 is 512bits, the error rate exceeds 20% and is indicated by "-" and the experimental result based on the picture Lena shows that the method of the embodiment resists gaussian noise attack with variance of 0.005 to 0.023, JPEG compression with quality factor of 10 to 100 and compression ratio of 10: JPEG2000 attack of 1 to 100:1, rotation attack of 0 to 360 degrees, scaling attack of 0.5 to 2.0;
TABLE 1
The bit error rate results of the picture Goldhill when attacked are shown in table 2; in table 2, the experimental results based on the picture Goldhill show that the method of this embodiment can resist gaussian noise attack with variance of 0.005 to 0.023, JPEG compression with quality factor of 10 to 100, compression ratio of 10: JPEG2000 attack of 1 to 100:1, rotation attack of 0 to 360 degrees, scaling attack of 0.5 to 2.0;
TABLE 2
The bit error rate results of the pictures Peppers under attack are shown in table 3; the experimental result based on the picture Peppers shows that the method of the embodiment can resist gaussian noise attack with variance of 0.005 to 0.023, JPEG compression with quality factor of 10 to 100 and compression ratio of 10: JPEG2000 attack of 1 to 100:1, rotation attack of 0 to 360 degrees, scaling attack of 0.5 to 2.0;
/>
the bit error rate results of the picture barbera under attack are shown in table 4; the experimental result based on the picture Barbara shows that the method of the embodiment can resist Gaussian noise attack with variance of 0.005 to 0.023, JPEG compression with quality factor of 10 to 100 and compression ratio of 10: JPEG2000 attack of 1 to 100:1, rotation attack of 0 to 360 degrees, scaling attack of 0.5 to 2.0;
/>
it is to be understood that the above examples of the present invention are provided by way of illustration only and not by way of limitation of the embodiments of the present invention. Any modification, equivalent replacement, improvement, etc. which come within the spirit and principles of the invention are desired to be protected by the following claims.
Claims (10)
1. A robust reversible watermark embedding method based on attack simulation is characterized in that: the method comprises the following specific steps:
s1: acquiring an original image I;
s2: calculating an n-order M-order harmonic transformation moment M of an original image I n,l ;
S3: for said polar harmonic transformation moment M n,l Performing self-adaptive normalization operation based on attack simulation to obtain normalized polar harmonic transformation moment
S4: for normalized polar harmonic transformation momentWatermark embedding with lowest embedding strength to obtain normalized polar harmonic transformation moment with watermark>And quantization distortion d thereof quantified ;
S5: for normalized polar harmonic transformation moment with watermarkPerforming self-adaptive normalized inverse operation to obtain polar harmonic transformation moment with watermark>
S6: obtaining polar harmonic transformation moments with watermarksAnd polar harmonic transformation moment M n,l To obtain an error imageSuperimposed error image +.>And the original image I, a candidate watermark image is obtained>
S7: for candidate watermark imagePerforming attack simulation and extracting watermark information in the attacked image>If watermark bit is in the extracted watermark information>Extracting errors and watermark bits +.>If the embedding strength level of the watermark bit is smaller than or equal to the set threshold value, increasing the embedding strength level of the watermark bit, and returning to the step S4; otherwise, candidate watermark imageAs a robust watermark image->
S8: for robust watermark imageCarrying out a dewatering operation to obtain a dewatering image I watermarking-removal ;
S9: calculation of the original image and the watermark image I watermarking-removal Rounding distortion d between rounding ;
S10: to quantize distortion d quantified Rounding distortion d rounding Robust watermark imageThe least significant bit of the first several pixels of (a) is used as auxiliary information; embedding auxiliary information into a robust watermark image using a reversible watermark method>In which watermark and auxiliary information w are contained 2 Intermediate image +.>
S11: intermediate imageThe least significant bit of the first several pixels of (a) is replaced by a 0 value and a hash value H is generated 1 The method comprises the steps of carrying out a first treatment on the surface of the Substitution of intermediate images by hash values +.>The least significant bit of the first pixels of the image is used for obtaining a robust reversible watermark image to finish the robust reversible watermarkPrint insert->
2. The robust reversible watermark embedding method based on attack simulation according to claim 1, wherein: in the step S2, an n-order M-order harmonic transformation moment M of the original image I is calculated n,l The method specifically comprises the following steps:
setting an order M, and taking the center of an original image I with the size of K multiplied by K as a circle center to make an inscribed circle; taking the inscribed circle as a unit circle, solving polar harmonic transformation moment of pixels in the unit circle; constructing polar harmonic transform basis V based on the inscribed circle n,l (x, y); through V n,l (x, y) calculating the polar harmonic transformation moment to obtain an n-order l-order polar harmonic transformation moment M n,l :
Wherein Deltax is s And Deltay t Representing the step sizes of the x-axis and the y-axis of the image unit circle, f (x s ,y t ) Representing pixels within a unit circle, (·) * Representing the complex number of conjugates, V n,l (x s ,y t ) Is a polar harmonic transformation basis in a circle.
3. The robust reversible watermark embedding method based on attack simulation according to claim 2, wherein: in the step S3, the polar harmonic transformation moment M n,l Performing self-adaptive normalization operation based on attack simulation to obtain normalized polar harmonic transformation momentThe method comprises the following specific steps:
s301: 1000 images are randomly selected from the BOWS2 database, and are subjected to attack simulation test, and the test result is fitted into a binary cubic polynomial by a double square polynomial fitting method:
wherein T is ni,li Representing the fitting result and simultaneously being self-adaptive normalized weight; n is n i And l i Polar harmonic transformation moments M for embedding the ith watermark bit, respectively ni,li The number of steps and the number of times; p is p 1 ~p 10 The fitting operation is carried out;
s302: selecting polar harmonic transform moment M for watermark embedding n,l Performing self-adaptive normalization to obtain normalized polar harmonic transformation moment
Wherein M is 00 Is the polar harmonic transformation moment of 0 order 0, T ni,li Is an adaptive normalized weight.
4. A robust reversible watermark embedding method based on attack simulation according to claim 3, characterized in that: in the step S4, the normalized polar harmonic transformation momentWatermark embedding with lowest embedding strength to obtain normalized polar harmonic transformation moment with watermark>And quantization distortion d thereof quantified The method comprises the following specific steps of;
s401: grouping normalized polar harmonic transformation moments; each group comprises P normalized polar harmonic transformation moments for embedding a watermark bit; the normalized polar harmonic transformation moment of the embedded ith watermark bit is expressed as:
the specific embedding formula for embedding the watermark is as follows:
wherein,is normalized polar harmonic transformation moment->Transpose of->For normalized polar harmonic transformation moment with watermark, < >>Is a multi-stage quantizer with embedded intensity level of z-stage, u is a random vector with length of P; the multi-stage quantizer is specifically:
wherein [ (S)]Is a rounding operation, (. Cndot.) for the purpose of handling the round-off operations T Is a transpose operation, delta is the quantization step size, Q w (. Cndot.) is a single-stage quantizer; q (Q) w The (-) is specifically as follows:
where w is the embedded watermarkBits, beta i (w) is constrained to be beta i (1)=β i (0) A jitter value of +Δ/2;
s402: obtaining quantization distortion d corresponding to the ith watermark bit quantified-i :
Wherein,is a multi-level quantizer with embedded intensity level z.
5. The robust reversible watermark embedding method based on attack simulation according to claim 4, wherein: in the step S5, the normalized polar harmonic transformation moment with the watermark is obtainedPerforming self-adaptive normalized inverse operation to obtain polar harmonic transformation moment with watermark>The method comprises the following steps:
normalized polar harmonic transformation moment with watermark by inverse operation of adaptive normalization operationMultiplying by M 00 Self-adaptive normalized weight T ni,li Divided by 10 3 Obtaining polar harmonic transformation moment with watermark:
6. the attack simulation-based simulation of claim 5The robust reversible watermark embedding method is characterized by comprising the following steps of: in the step S6, polar harmonic transformation moment with watermark is takenAnd polar harmonic transformation moment M n,l To obtain an error imageThe method comprises the following steps: obtaining polar harmonic transformation moment with watermark>Polar harmonic transformation moment M with original image ni,li Is a difference in (2); performing polar harmonic transformation and inverse transformation on the difference value and rounding operation to obtain an error image I error :
Wherein V is ni,li For polar harmonic transform basis for embedding the ith watermark bit, V ni,-li Is V (V) ni,li Is used to determine the complex number of the conjugate,is->Complex conjugate of M ni,-li) Is M ni,li Is a complex conjugate of (a) and (b).
7. The robust reversible watermark embedding method based on attack simulation according to claim 6, wherein: in the step S8, for the robust watermark imageCarrying out a dewatering operation to obtain a dewatering image I watermarking-removal The method comprises the following specific steps of:
s801: computing robust watermark imagesPolar harmonic transformation moment->Performing adaptive normalization to obtain ++>And extracting watermark information->
Wherein,and->Respectively robust watermark image +.>And a watermark image I watermarking-removal The normalized polar harmonic transformation moment for embedding the ith watermark bit;
s802: the inverse operation of the self-adaptive normalization operation is adopted to obtain
By passing throughCalculation of the watermark image I watermarking-removal :
Where P x L represents the number of polar harmonic transformed moments for which watermark embedding is totally modified.
8. A robust reversible watermark extraction method based on attack simulation is characterized in that: comprising the embedding method of any one of claims 1 to 7: the method also comprises the following specific steps:
s21: extracting least significant bits of first S pixels of the robust reversible watermark image to obtain a hash value H 1 ;
S22: replacing least significant bits of first S pixels of the robust reversible watermark image with 0 value, and generating hash value H of the image at the moment 2 ;
S23: if the hash value H 1 Equal to hash value H 2 Judging that the robust reversible watermark image is not attacked, and executing step S05; if the hash value H 1 Not equal to hash value H 2 When the robust reversible watermark image is attacked, executing a step S06;
s24, calculating quantization distortion of the robust reversible watermark image which is not attacked, extracting watermark information in the robust watermark image according to the quantization distortion, and recovering the robust reversible watermark image which is not attacked into an original image;
s25: robust reversible watermark image after computing is attackedPolar harmonic transformation moment->For->And performing self-adaptive normalization operation and extracting watermark information.
9. The robust and reversible watermark extraction method based on attack simulation according to claim 8, wherein: the step S24 specifically comprises the following steps:
s051: extraction by reversible watermarkingIs quantized distortion d of (2) quantified The method comprises the steps of carrying out a first treatment on the surface of the Combining robust watermark images->Is the rounding distortion d of (2) rounding And the least significant bits of the first S pixels, recovering the robust watermark image +.>
S052: computing robust watermark imagesPolar harmonic transformation moment->Performing self-adaptive normalization operation to obtain +.>And calculate the ith watermark information w i :
S053: according to quantization distortion d quantified Extracting a robust watermark imageWatermark information in the image is obtained to obtain a watermark image I watermarking-removal :
S054: computing robust watermark imagesPolar harmonic transformation moment->Will->Performing adaptive normalization to obtain ++>Extracting its quantization distortion:
wherein,and->Respectively robust watermark image +.>And a watermark image I watermarking-removal The normalized polar harmonic transformation moment for embedding the ith watermark bit;
s055: by passing throughCalculation of the watermark image I watermarking-removal :
S056: using de-watermarked image I watermarking-removal And rounding distortion d rounding Restoring the original image I:
I=I watermarking-removal +d rounding 。
10. the robust and reversible watermark extraction method based on attack simulation according to claim 8, wherein: in the step S25, watermark information is extracted, specifically:
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311169752.6A CN117173001A (en) | 2023-09-11 | 2023-09-11 | Robust reversible watermark embedding and extracting method based on attack simulation |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202311169752.6A CN117173001A (en) | 2023-09-11 | 2023-09-11 | Robust reversible watermark embedding and extracting method based on attack simulation |
Publications (1)
Publication Number | Publication Date |
---|---|
CN117173001A true CN117173001A (en) | 2023-12-05 |
Family
ID=88944737
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202311169752.6A Pending CN117173001A (en) | 2023-09-11 | 2023-09-11 | Robust reversible watermark embedding and extracting method based on attack simulation |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN117173001A (en) |
-
2023
- 2023-09-11 CN CN202311169752.6A patent/CN117173001A/en active Pending
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Keshavarzian et al. | ROI based robust and secure image watermarking using DWT and Arnold map | |
CN104616244B (en) | Image watermark insertion and extracting method based on BP neutral net compression domains | |
CN111242831B (en) | Zernike moment-based geometric attack resistant reversible robust watermarking method | |
CN110349073B (en) | Four-system color digital image blind watermarking method based on Schur decomposition | |
Fu et al. | Watermarking scheme based on support vector machine for colour images | |
Kang et al. | Color Image Steganalysis Based on Residuals of Channel Differences. | |
Chang et al. | Preserving robustness and removability for digital watermarks using subsampling and difference correlation | |
Loukhaoukha et al. | Hybrid watermarking algorithm based on SVD and lifting wavelet transform for ownership verification | |
CN101833745A (en) | Method for detecting embedding and extracting of multiple binary embedded watermarks of digital image | |
CN117173001A (en) | Robust reversible watermark embedding and extracting method based on attack simulation | |
CN109829846B (en) | Digital image blind watermarking method based on two-dimensional discrete cosine transform | |
CN116993567A (en) | Frequency domain blind watermarking method based on Hadamard transform and teaching optimization algorithm | |
CN110415155B (en) | Blind watermarking method for airspace color image fused with haar transformation | |
Huynh-The et al. | Robust image watermarking framework powered by convolutional encoder-decoder network | |
Lai et al. | Image watermarking scheme using genetic algorithm | |
Chang et al. | A robust DWT-based copyright verification scheme with Fuzzy ART | |
CN110415154B (en) | Haer transformation-based quaternary color digital image blind watermarking method | |
CN111242830B (en) | Image reversible robust data hiding method based on polar harmonic transformation | |
CN113962839A (en) | Multi-color mode quantum image blind watermarking method based on TMQIR | |
Peungpanich et al. | An improving method for image watermarking using image averaging and tuned pixels prediction | |
Ma et al. | Color image watermarking using local quaternion Fourier spectral analysis | |
CN117217972A (en) | Robust reversible watermark embedding and extracting method based on fractional PECT moment | |
Patel et al. | SEMI-FRAGILE BLIND WATERMARKING MECHANISM FOR COLOR IMAGE AUTHENTICATION AND TAMPERING. | |
Jin et al. | Semi-fragile watermark based on adaptive quantization for image content authentication | |
Hemida et al. | Block-DCT based alterable-coding restorable fragile watermarking scheme with superior localization |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |