CN1598877A - Positioning vulnerable water print generating and recognizing method capable of distigushing image and watermark distortion - Google Patents
Positioning vulnerable water print generating and recognizing method capable of distigushing image and watermark distortion Download PDFInfo
- Publication number
- CN1598877A CN1598877A CN 200410040433 CN200410040433A CN1598877A CN 1598877 A CN1598877 A CN 1598877A CN 200410040433 CN200410040433 CN 200410040433 CN 200410040433 A CN200410040433 A CN 200410040433A CN 1598877 A CN1598877 A CN 1598877A
- Authority
- CN
- China
- Prior art keywords
- image
- watermark
- msub
- msubsup
- low
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title 1
- 238000013139 quantization Methods 0.000 claims description 20
- 230000000739 chaotic effect Effects 0.000 claims description 18
- 238000000354 decomposition reaction Methods 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 5
- 230000005540 biological transmission Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 3
- 238000004422 calculation algorithm Methods 0.000 abstract description 15
- 230000000007 visual effect Effects 0.000 abstract description 4
- 238000001514 detection method Methods 0.000 description 7
- 230000035945 sensitivity Effects 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 239000011521 glass Substances 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 229910002056 binary alloy Inorganic materials 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011084 recovery Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 238000004891 communication Methods 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000004807 localization Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000008929 regeneration Effects 0.000 description 1
- 238000011069 regeneration method Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Landscapes
- Editing Of Facsimile Originals (AREA)
- Image Processing (AREA)
Abstract
The invention discloses orientation flimsy watermark generating and certificating method which can distinguish image and juggled watermark. Original image's wavelet low frequency coefficient of 4 bits is scalar quantized and used as watermark. The watermark is disturbed using random number to encrypt to improve its security. The encrypted watermark is embedded into LSB bit of original image. Original image's general picture is displayed through original watermark image's low frequency compress image recovered from watermark when certificating. Combining juggled position of image information located by difference value images with difference between image information and juggled watermark, how attacker juggled the image can be directly judged. At the premise of guaranteeing image's authenticity, switch efficiency of digital image is improved, which is helpful for digital images' application and spread. Adopting the invention, the authentication result is direct, visual effect is good, key's space is bigger and watermark algorithm is safer.
Description
Technical Field
The invention relates to a positioning type fragile watermark embedding and authentication method, which is used for authenticating a digital image, namely proving the integrity and the authenticity of the digital image.
Background
In recent years, with the development of computer and multimedia technologies, copyright protection and content authentication of digital products are increasingly emphasized, and digital watermarks are generated under such a background and are receiving wide attention from academia. The research on robust watermarking for copyright protection has been relatively deep, but for content integrity certification of digital products, i.e. the reliability of transmitting multimedia data, robust watermarking techniques have not been able to fully accommodate this requirement, which fragile watermarking techniques are able to achieve. A fragile watermark should generally satisfy the following points: the method comprises the following steps of (1) invisibility, sensitivity to tampering of images added with watermarks, capability of tampering and positioning, and no need of original images for extracting watermarks; in the existing positioning type fragile watermark, the following two methods are better:
(1) hash function method based on safety
Hash functions are widely used in traditional cryptography. The method utilizes the extreme sensitivity of the Hash function to the initial value to realize the requirement of the fragile watermark to the tampering sensitivity, and is a hot spot of watermark research in recent years. In the reference "Secret and Public Key Image watermark Schemes for Image authentication and owner Verification" (P W Wong et al, IEEE trans. Image Processing, vol.10 No.10 o segment 2001. pp: 1593-.
(2) Chaos-based fragile watermarking technology
Chaotic techniques have wide application in both cryptography and secure communications. Patent application number 03116160, X proposes a "positioning type chaos fragile digital watermark embedding and extracting method" based on the extreme sensitivity of the chaotic system to the initial value. The pixel gray value after the LSB position of the original image is zero is subjected to chaos iteration for a plurality of times to generate a watermark, then the watermark is embedded into the LSB position of the original image, the integrity of the image is checked through a generated tampering judgment matrix during authentication, and the algorithm can accurately position the tampering of the image added with the watermark.
The two watermarking algorithms have the common characteristics in image authentication: the watermark obtained by a specific algorithm by using seven high-order bits of information on the tested image with eight bits is compared with the watermark stored in the Least Significant Bit (LSB) to judge whether the image is falsified. It is disadvantageous in that it can only indicate the location of tampering with the watermark image (watermarked image) and cannot distinguish between tampering with the seven-bit-high image information and tampering with the lowest-bit watermark. Since the tampering of the image information can destroy the use value of the original image, the fragile watermarking algorithm can detect and accurately position the tampering so as to ensure the reliability and effectiveness of authentication; the falsification of the watermark does not affect the use value of the image, the image with the falsified watermark only needs to pass authentication, otherwise, the image which can be directly used originally needs to be retransmitted because the image does not pass authentication indiscriminately, so that the exchange efficiency of the digital image is reduced, the real image cannot be effectively utilized, and the popularization and the application of the digital image are hindered. For example: if the digital image used as evidence of litigation is authenticated and identified by adopting the method, the requirement is too strict and the digital image is unfairly correct; if the method is adopted for authentication and identification of the news image, the news information which can reflect the truth of the fact is suspected, and the timely spreading of the news is hindered.
Disclosure of Invention
The invention aims to provide a positioning type fragile watermark generation and authentication method capable of distinguishing image and watermark tampering, which has accurate positioning capability on tampering, can distinguish whether image information or watermark tampering is carried out, improves the exchange efficiency of digital images, and is beneficial to application and popularization of the digital images; the authentication result is visual, the visual effect is good, the key space is larger, and the watermarking algorithm is safer.
The invention solves the technical problem, and adopts the technical scheme that: a positioning type fragile watermark generation and authentication method capable of distinguishing image and watermark tampering comprises the following steps:
(1) and watermark generation: performing two-dimensional one-level wavelet decomposition on the original image I to obtain: { LL, LH, HL, HH }, and low-frequency wavelet coefficient LL therein is scalar-quantized with 4 bits to generate a low-frequency compressed image ILLThe corresponding binary numbers are arranged in a spatial sequence to generate a binary image ILb(ii) a Using random sequences xL＝{x1，x2，...，xi，...，xLIs represented as { x } in an ordered sequencep1，xp2，xp3，...，xpi，...，xpLUsing the address sequence p1，p2，...，pL-1，pLTo the binary image ILbCarrying out block scrambling, and taking the scrambled binary image as a watermark W to be embedded;
(2) watermark embedding and key generation: embedding the watermark W generated in step (1) into LSB bits (Least Significant Bit, minimum Bit) of the original image I:，Iw＝ I+W，Iwto obtain a watermark image embedded with a watermark W; for watermark image IwPerforming two-dimensional first-level wavelet decomposition to obtain: { LL, LH, HL, HH }, 4-bit scalar quantization of the low-frequency coefficient LL to generate a low-frequency compressed image ILL WCalculating to obtain a difference key <math> <mrow> <msub> <mi>ΔI</mi> <mi>LL</mi> </msub> <mo>=</mo> <msub> <mi>I</mi> <mi>LL</mi> </msub> <mo>-</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mi>w</mi> </msubsup> </mrow> </math> In the formula ILLA low-frequency compressed image which is an original image I; (3) and watermark extraction: for watermark image IwMeasured image I received after transmission*And taking out LSB bit to obtain watermark W of detected image*And then restoring the low-frequency compressed image I stored in the watermark according to the key by the reverse calculation process of the step (1)LL', and using a difference key delta ILLAnd formulas thereof <math> <mrow> <msub> <mi>ΔI</mi> <mi>LL</mi> </msub> <mo>=</mo> <msub> <mi>I</mi> <mi>LL</mi> </msub> <mo>-</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mi>w</mi> </msubsup> </mrow> </math> Calculating a low frequency compressed image <math> <mrow> <msubsup> <mi>I</mi> <mi>LL</mi> <msup> <mi>w</mi> <mo>′</mo> </msup> </msubsup> <mo>=</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mo>′</mo> </msubsup> <mo>-</mo> <msub> <mi>ΔI</mi> <mi>LL</mi> </msub> <mo>;</mo> </mrow> </math>
(4) And authentication: for the detected image I*Performing two-dimensional first-level wavelet decomposition to obtain: { LL, LH, HL, HH }, performing 4-bit scalar quantization on the low-frequency wavelet coefficient LL therein to generate a detected image I*Low frequency compressed image I ofLL *(ii) a Then, the low-frequency compressed image I recovered in the step (3) is processedLL w′And the image I to be measured*Low frequency compressed image I ofLL *Subtracting to obtain a difference image <math> <mrow> <mi>Γ</mi> <mo>=</mo> <mo>|</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <msup> <mi>w</mi> <mo>′</mo> </msup> </msubsup> <mo>-</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mo>*</mo> </msubsup> <mo>|</mo> </mrow> </math> (ii) a If there are several non-zero points distributed randomly in the difference image, the image I is determined*The watermark is tampered, otherwise, the detected image I is identified*The watermark of (2) has not been tampered with; if a plurality of non-zero points in the difference image are concentrated in a certain region and the area delta S of the region is larger than the threshold template T, determining the detected image I*The image information of the corresponding area is falsified, otherwise, the detected image I is determined*Has not been tampered with.
The method for 4-bit scalar quantization of the low-frequency wavelet coefficient LL is 4-bit mean scalar quantization or 4-bit non-mean scalar quantization.
Random sequence x as described aboveL＝{x1，x2，...，xi，...，xLThe method can be a chaos sequence, and can also be a multi-value random sequence generated by other random models such as multiplication and congruence, so that the method is easier to realize and higher in reliability.
Compared with the prior art, the invention has the beneficial effects that:
1. since the tampering of the transmission image by the attacker is generally a local centralized change, only the tampering needs to be detected, the tampering position is located, and the tampering position is not authenticated. The invention detects the tampered area of the template larger than the set threshold value and does not pass the authentication, thereby ensuring the reliability and the authenticity of the digital image; meanwhile, the tampering of the watermark can be distinguished, the tampering which does not affect the authenticity of the image is authenticated, the application and exchange efficiency of the digital image is improved on the basis of ensuring the authenticity of the image, and the application and popularization of the digital image are facilitated. Such as: the method is used for authenticating and distinguishing the digital image, the authentication result of the method is consistent with the reality and truth, and the method can be used as litigation evidence and news materials, so that judicial activities and news activities are more efficient and fair.
2. The low-frequency compressed image recovered from the watermark during authentication reflects the general appearance of the original image; the authentication result is reflected by the distribution of non-zero points on the difference image, and is very visual; and the low-frequency compressed image recovered from the watermark can be combined to intuitively judge how the attacker tampers the image.
3. When the watermark is generated, the watermark is generated by directly utilizing eight-bit information of the original image instead of the high seven-bit information of the original image, so that the key space is enlarged, and the watermark algorithm is safer.
Detailed Description
The present invention will be further described with reference to the following examples.
FIG. 1 is a block diagram of a watermark generation algorithm of the present invention
FIG. 2 is a block diagram of the algorithm for embedding watermark and generating secret key
FIG. 3 is a block diagram of watermark extraction and authentication in the present algorithm
FIG. 4 is a diagram of a detection and authentication legend for detecting a large tampering of an image under test, and distinguishing the tampering of the image and a watermark by using the method of the present invention
In fig. 4: i is the original image, IwAs a watermark image, Ia *For tested images in which both the image information and the watermark have been tampered with, Ib *For the inspected image with tampered image information, Ic *The watermark is a tampered detected image; i isLL-a、ILL-bAnd ILL-cAre respectively slave and Ia *、Ib *And Ic *Recovering low-frequency compressed image from corresponding watermark, gamma b and gamma c are respectively equal to Ia *、Ib *And Ic *The corresponding difference image.
FIG. 5 is a diagram of a detection and authentication method for detecting tampering and locating a tampered location of an image of a subject image with minimal tampering
In fig. 5: i iswAs a watermark image, I*To tamper with the image, ILLFor the low-frequency compressed image recovered from the watermark, Γ is a difference image for detecting tampering, and T is an enlarged view of the tampered area before tampering. T is*An enlarged view of the tampered area after tampering.
Examples
A positioning type fragile watermark generation and authentication method capable of distinguishing image and watermark tampering comprises the following steps:
firstly, watermark generation:
1. fig. 1 shows the early stage process of watermark generation as follows: performing two-dimensional one-level wavelet decomposition on the original image I to obtain: { LL, LH, HL, HH }, and low-frequency wavelet coefficient LL therein is scalar-quantized with 4 bits to generate a low-frequency compressed image ILLThe corresponding binary systems are arranged in a spatial sequence to generate a binary image ILb。
In the embodiment, a DB1 wavelet base is selected to perform two-dimensional one-level wavelet decomposition on an image with the size of m × n, and 4-bit mean scalar quantization is performed on a low-frequency coefficient LL in the image to generate a low-frequency compressed image ILL. The quantization process is described by the formula:
ILL＝Q(LL) (1)
the rule of correspondence Q is:
wherein i is 1, 2, 1., m/2, j is 1, 2, 1., n/2,
called quantization step size, max and min are the maximum and minimum values, respectively, of the elements in LL.
Compressing low frequency of picture ILLIs converted into a binary image I of 4-bit binary system arranged in a spatial orderLbThe formula is described as follows:
wherein,
2. fig. 1 shows the watermark generation later process as follows: using random sequences xL＝{x1，x2，...，xi，...，xLIs represented as { x } in an ordered sequencep1，xp2，xp3，...，xpi...xpLUsing the address sequence p1，p2，...，pL-1，pLTo the binary image ILbCarrying out block scrambling, and taking the scrambled binary image as a watermark W to be embedded;
based on the characteristics of numerous chaotic mappings, extreme sensitivity of the chaotic system to initial values, good randomness and easy regeneration, the chaotic sequence is used as an encryption sequence, so that the chaotic system has better safety. In order to increase the period of the chaotic sequence, make the sensitivity to the initial value higher and overcome the influence of the finite word length, the embodiment selects the formula (4) chaotic iterative system to generate the random sequence:
wherein, ynIn this embodiment, a Logistic chaotic map is selected for any chaotic map. It and the initial value of the chaotic system form a chaotic key together.
Using the chaotic system to generate a length ofChaotic sequence x of LL＝x1，x2，...，xi，...，xLAnd (6) carrying out address sequencing on the chaotic sequence by using a stable sequencing method to generate an address sequence:
A(i)＝pi i＝1，2，3，....，L (5)
using the address sequence A to the binary image ILbAnd carrying out block scrambling to generate the watermark W to be embedded. The principle of partitioning the image is as follows: the block sizes are the same and the total number of blocks of the image is equal to L.
Let mk, nk denote binary image I respectivelyLbThe number of blocks per row and per column, then L ═ mk × nk, the size of each block is: (m/mk) × (n/nk).
The scrambling method is as follows:
ILb(iLb，jLb)→W(iw，jw) (6)
wherein: temp. is usedij＝(iLb-1)*nk+jLb
jw＝A(tempij)-(iw-1)*nk
Watermark embedding and key generation:
fig. 2 shows the watermark embedding and key generation process of this example as follows: embedding the watermark W generated in the step (a) into LSB bits (Least Significant bits) of the original image I:Iw＝ I+W，Iwto obtain a watermark image embedded with a watermark W; for watermark image IwPerforming two-dimensional first-level wavelet decomposition to obtain: { LL, LH, HL, HH }, for low frequency coefficient LLPerforming 4-bit scalar quantization to generate a low-frequency compressed 5-picture ILL WCalculating to obtain a difference key <math> <mrow> <msub> <mi>ΔI</mi> <mi>LL</mi> </msub> <mo>=</mo> <msub> <mi>I</mi> <mi>LL</mi> </msub> <mo>-</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mi>w</mi> </msubsup> </mrow> </math> In the formula ILLIs a low frequency compressed image of the original image I.
Thirdly, watermark extraction:
fig. 3 shows that the watermark extraction process of this example is: for received tested image I embedded with watermark*And taking out LSB bit to obtain watermark W of detected image*And then recovering the low-frequency compressed image I stored in the watermark according to the chaos key and the inverse calculation process of step (one)LL', and using a difference key delta ILLAnd formulas thereof <math> <mrow> <msub> <mi>ΔI</mi> <mi>LL</mi> </msub> <mo>=</mo> <msub> <mi>I</mi> <mi>LL</mi> </msub> <mo>-</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mi>w</mi> </msubsup> </mrow> </math> Calculating a low frequency compressed image <math> <mrow> <msubsup> <mi>I</mi> <mi>LL</mi> <msup> <mi>w</mi> <mo>′</mo> </msup> </msubsup> <mo>=</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mo>′</mo> </msubsup> <mo>-</mo> <msub> <mi>ΔI</mi> <mi>LL</mi> </msub> <mo>;</mo> </mrow> </math>
Recovery of ILLThe specific process of' is as follows: chaotic sequence x generated by chaotic keyL＝{x1，x2，...，xi，...，xLSorting the chaos sequence to generate an address sequence A, recovering the scrambled watermark by using the address sequence A, and converting the corresponding space block after recovery into a corresponding decimal number, namely a low-frequency compressed image I recovered from the watermarkLL′。
If the watermark is not tampered with, ILL' equal to originalLow frequency compressed image of the starting image I, I calculated from the above formulaLL w′Equal to the watermark image IwLow frequency compressed image, e.g. I in FIG. 4LL-bShown; if the watermark is distorted, after the distortion of the local watermark is scrambled and recovered, recovering the local watermarkLL' there will appear randomly distributed noise points, I calculated by the above formulaLL w′There will also be randomly distributed noise points, but the basic content of the watermark image can still be seen, as shown in fig. 4ILL-aAnd ILL-cAs shown.
Fourthly, authentication:
fig. 3 shows the authentication process of the present invention as follows: for the detected image I*Performing two-dimensional first-level wavelet decomposition to obtain: { LL, LH, HL, HH }, performing 4-bit mean scalar quantization on the low-frequency wavelet coefficient LL therein to generate a detected image I*Low frequency compressed image I ofLL *(ii) a Low frequency compressed image I to be recovered from watermarkLL w′And the image I to be measured*Low frequency compressed image I ofLL *Subtracting to obtain a difference image <math> <mrow> <mi>Γ</mi> <mo>=</mo> <mo>|</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <msup> <mi>w</mi> <mo>′</mo> </msup> </msubsup> <mo>-</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mo>*</mo> </msubsup> <mo>|</mo> <mo>;</mo> </mrow> </math>
If a plurality of non-zero points which are randomly distributed exist in the difference image, the image I to be detected is determined*The watermark is tampered, otherwise, the detected image I is identified*The watermark of (2) has not been tampered with; if a plurality of non-zero points in the difference image are concentrated in a certain region and the area delta S of the region is larger than the threshold template T, determining the detected image I*The image information of the corresponding area is tampered, otherwise, the detected image I is identified*Has not been tampered with. Such as: Γ a and Γ c in fig. 4 reflect that the detected image has watermark and is tampered; Γ a and Γ b in fig. 4 and Γ in fig. 5 reflect that the image information of the measured image is falsified and accurately reflects the image informationThe position and shape of the tamper portion.
The threshold template T is selected in consideration of two factors: the size of the image scrambling block and the size of the tampered area in the image under test. In consideration of practical tampering, in the present embodiment, the size of the scrambled block is 2 × 2 (i.e., the image block is composed of 2 × 2 pixels), and the size of the selected threshold template T is 3 × 3.
Random sequence x employed in the present inventionL＝{x1，x2，...，xi，...，xLThe sequence can be a chaotic sequence, or a multi-value random sequence generated by other random models such as multiplication congruence and the like; the method for 4-bit scalar quantization of the low-frequency wavelet coefficient LL may use a 4-bit non-mean scalar quantization method in addition to the 4-bit mean scalar quantization of the present example. The two-dimensional one-level wavelet decomposition of the original image I can use the DB1 wavelet basis of this example, and can also use any other existing wavelet basis.
The effects of the invention can be verified by the following performance analysis:
invisibility analysis:
the fragile watermark requires that the added watermark is imperceptible, watermark information is embedded into LSB bits of an image, P is set as the probability of changing the lowest bit of the original image, and the independence of each bit of the LSB bits can indicate that P is 0.5. Theoretically, the algorithm of the invention obtains a high peak signal-to-noise ratio and meets the requirement of fragile watermarks on imperceptibility.
Tamper localization capability analysis:
the invention locates the tampered position of the image and judges how the image is tampered by the low-frequency compressed image and the difference image recovered from the watermark. The low-frequency compressed image is obtained by quantizing the wavelet low-frequency coefficient, and the number of rows and columns of the low-frequency compressed image is half of that of the original image. The tamper detection capability of the algorithm is illustrated by taking the DB1 wavelet basis employed in the present embodiment as an example.
The two-dimensional wavelet transform of the DB1 wavelet basis can be described mathematically approximately by the following formula:
ILL(i，j)＝sum(I(2*i-1∶2*i，2*j-1∶2*j))/2 (7)
wherein, i is 1, 2, 1, m/2, j is 1, 2, n/2
The value range of each pixel of the eight-bit image is [0, 255], so the value range of LL is [0, 510], and the quantization step size when 4-bit mean scalar quantization is performed is q < (510/16) > 32. If the change to one pixel value or the sum of 2 x 2 pixel values exceeds 64, the difference image is not equal, and the position of the image information falsification can be located.
Differential tamper analysis:
the authentication algorithm locates the tampered position of the image through the difference image gamma and distinguishes whether the image information or the watermark is tampered.
If the watermark has not been tampered with, recovering a low frequency compressed image I from the watermarkLL' Low frequency compressed image equal to original image I, represented by formula <math> <mrow> <msubsup> <mi>I</mi> <mi>LL</mi> <msup> <mi>w</mi> <mo>′</mo> </msup> </msubsup> <mo>=</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mo>′</mo> </msubsup> <mo>-</mo> <msub> <mi>ΔI</mi> <mi>LL</mi> </msub> </mrow> </math> Calculated ILL w′A low frequency compressed image equal to the watermark image; if the watermark is distorted, after the distortion of the local watermark is scrambled and recovered, recovering the low-frequency compressed image I from the watermarkLL' randomly distributed noise points, ILL w′There also appear randomly distributed noise points, but ILL w′The underlying content of the watermark image can still be approximated. I isLL *Is a detected image I*Of low frequency compressed image of formula <math> <mrow> <mi>Γ</mi> <mo>=</mo> <mo>|</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <msup> <mi>w</mi> <mo>′</mo> </msup> </msubsup> <mo>-</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mo>*</mo> </msubsup> <mo>|</mo> </mrow> </math> A difference image Γ is calculated. When the image content is tampered, the corresponding position in the difference image gamma is not zero, and the shape of the non-zero area is similar to that of the tampered area; when the watermark is tampered with, tamper points appear in Γ that resemble randomly distributed small blocks that are not zero.
The computer simulation analysis of the method of the invention:
matlab is used to simulate the algorithm of the present invention, and FIG. 4 and FIG. 5 show the detection and authentication results of two different tampering situations.
Fig. 4 shows the detection and authentication result of the detected image with greater tampering, and the method of the present invention distinguishes the image from the watermark tampering. 208, 328, 8 'vase' grey scale original image I and watermarked image IwThe peak signal-to-noise ratio of (c) is 51.2108 dB. Watermark image I by using photoshop image editing softwarewAdding a wine glass, and marking the tampered image as Ia *(ii) a By using a watermark image IwLSB bit replacement ofa *The LSB bit of (A) obtains a detected image I only tampering with image informationb *( High 7 + I ofwLSB bit of); by using a watermark image IwHigh seven-bit information replacement Ia *The high seven bits of the image to be detected I only tampering with the watermarkc *( High 7 + I ofa *LSB bit of); i isLL-a、ILL-bAnd ILL-cAre respectively slave and Ia *Ib *And Ic *Corresponding toRecovering low-frequency compressed image in watermark, gamma a, gamma b and gamma c are respectively equal to Ia *、Ib *And Ic *The corresponding difference image.
ILL-aAnd ILL-cIn the image I, the similar randomly distributed 'noise' indicates the corresponding detected image Ia *And Ic *The watermark of (2) is tampered with; i isLL-bThere is no "noise" like random distribution, which indicates the corresponding detected image Ib *The watermark in (1) has not been tampered with; i isLL-aAnd ILL-bWithout a wine glass, and the image I to be measureda *And Ib *The method is characterized in that a wine glass is arranged in the image, and the tampering can be visually judged by adding the wine glass in the image. I isLL-cAnd the image I to be measuredc *Compared with the detected image I which can not be seen with obvious change, the detected image I can be basically judgedc *Only the watermark part is tampered, and the use value of the image is not influenced. The difference images Γ a and Γ b have regions where non-zero points are concentrated, and the image I to be measured is determineda *And Ib *The image information of the region is falsified, and the shape of the falsified image information (in this example, "cup") can be determined from the shape of the non-zero-point concentrated region; the difference images gamma and gamma have random-like distributed noise, which indicates that the image I to be measured is Ia *And Ic *The watermark is tampered; wherein, the image I is judged to be the detected image I only with 'noise' and no non-zero point concentration area in the Γ cc *Only the watermark is tampered with. Therefore, the invention can accurately position the tampered position and can very intuitively judge how to tamper the detected image.
Fig. 5 shows the detection and authentication results of the image information of the detected image with slight tampering, and the tampering is detected and the tampered position is located by the method of the present invention. 512 by 8 "Jet" grayscale image I and watermark image IwHas a peak signal-to-noise ratio of 50.2801. The first letter in the U.S. air FORCE on the airplane body in the Jet image is edited by using photoshop image editing softwareU, removing and by using the watermark image IwReplacing LSB bits of the image to obtain a detected image I only tampering with image information*For more clearly observing the tampered content, an enlarged image T before the region is tampered and an enlarged image T after the region is tampered are given*. Detection of tampering by the algorithm proposed by the invention, ILLThe presence of "u." is clearly visible from the low frequency compressed image recovered from the watermark. The position of the image falsified is accurately located in the difference image Γ. It can be seen that the present invention has good tamper sensitivity.
Claims (3)
1. A positioning type fragile watermark generation and authentication method capable of distinguishing image and watermark tampering comprises the following steps:
(1) and watermark generation: performing two-dimensional one-level wavelet decomposition on the original image I to obtain: { LL, LH, HL, HH }, and low-frequency wavelet coefficient LL therein is scalar-quantized with 4 bits to generate a low-frequency compressed image ILLThe corresponding binary numbers are arranged in a spatial sequence to generate a binary image ILb(ii) a Using random sequences xL＝{x1，x2，...，xi，...，xL}, ordered sequences thereofDenoted as { xp1，xp2，xp3，...，xpi，...，xpLUsing the address sequence p1，p2，...，pL-1，pLTo the binary image ILbCarrying out block scrambling, and taking the scrambled binary image as a watermark W to be embedded;
(2) watermark embedding and key generation: embedding the watermark W generated in step (1) into LSB bits of the original image I:
IW＝ I+W，IWis a watermark image embedded with a watermark W; for watermark image IWPerforming two-dimensional first-level wavelet decomposition to obtain: { LL, LH, HL, HH }, 4-bit scalar quantization of the low-frequency coefficient LL to generate a low-frequency compressed image ILL WCalculating to obtain a difference key <math> <mrow> <msub> <mi>ΔI</mi> <mi>LL</mi> </msub> <mo>=</mo> <msub> <mi>I</mi> <mi>LL</mi> </msub> <mo>-</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mi>w</mi> </msubsup> <mo>,</mo> </mrow> </math> In the formula ILLA low-frequency compressed image which is an original image I;
(3) and watermark extraction: for watermark image IWMeasured image I received after transmission*And taking out LSB bit to obtain watermark W of detected image*And then restoring the low-frequency compressed image I stored in the watermark W according to the key by the reverse calculation process of the step (1)LL', and using a difference key delta ILLAnd formulas thereof <math> <mrow> <msub> <mi>ΔI</mi> <mi>LL</mi> </msub> <mo>=</mo> <msub> <mi>I</mi> <mi>LL</mi> </msub> <mo>-</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mi>w</mi> </msubsup> <mo>,</mo> </mrow> </math> Computing low frequency compressed images <math> <mrow> <msubsup> <mi>I</mi> <mi>LL</mi> <msup> <mi>w</mi> <mo>′</mo> </msup> </msubsup> <mo>=</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mo>′</mo> </msubsup> <mo>-</mo> <msub> <mi>ΔI</mi> <mi>LL</mi> </msub> <mo>;</mo> </mrow> </math>
(4) And authentication: for the detected image I*Performing two-dimensional first-level wavelet decomposition to obtain: { LL, LH, HL, HH }, performing 4-bit scalar quantization on the low-frequency wavelet coefficient LL therein to generate a detected image I*Low frequency compressed image I ofLL *(ii) a Then, the low-frequency compressed image I recovered in the step (3) is processedLL W′And the image I to be measured*Low frequency compressed image I ofLL *Subtracting to obtain a difference image <math> <mrow> <mi>Γ</mi> <mo>=</mo> <mo>|</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <msup> <mi>w</mi> <mo>′</mo> </msup> </msubsup> <mo>-</mo> <msubsup> <mi>I</mi> <mi>LL</mi> <mo>*</mo> </msubsup> <mo>|</mo> <mo>;</mo> </mrow> </math> If there are several non-zero points distributed randomly in the difference image, the image I is determined*The watermark is tampered, otherwise, the detected image I is identified*The watermark of (2) has not been tampered with; if a plurality of non-zero points in the difference image are concentrated in a certain region and the area delta S of the region is larger than the threshold template T, determining the detected image I*The image information of the corresponding area is falsified, otherwise, the detected image I is determined*Has not been tampered with.
2. The method of claim 1, wherein the method comprises: the method for 4-bit scalar quantization of the low-frequency wavelet coefficient LL is 4-bit mean scalar quantization or 4-bit non-mean scalar quantization.
3. The method of claim 1, wherein the method comprises: the random sequence xL＝{x1，x2，...，xi，...，xLAnd the frequency is generated by using a chaotic system.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2004100404336A CN1322433C (en) | 2004-08-12 | 2004-08-12 | Positioning vulnerable water print generating and recognizing method capable of distigushing image and watermark distortion |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2004100404336A CN1322433C (en) | 2004-08-12 | 2004-08-12 | Positioning vulnerable water print generating and recognizing method capable of distigushing image and watermark distortion |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1598877A true CN1598877A (en) | 2005-03-23 |
| CN1322433C CN1322433C (en) | 2007-06-20 |
Family
ID=34664665
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2004100404336A Expired - Fee Related CN1322433C (en) | 2004-08-12 | 2004-08-12 | Positioning vulnerable water print generating and recognizing method capable of distigushing image and watermark distortion |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN1322433C (en) |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100340111C (en) * | 2005-03-28 | 2007-09-26 | 南方医科大学 | Medical image fragile watermark method based on wavelet transform |
| CN100525365C (en) * | 2005-09-28 | 2009-08-05 | 佳能株式会社 | Image processing apparatus and control method therefor |
| CN101021938B (en) * | 2006-02-16 | 2010-09-01 | 刘粉林 | Safety image identifying method based on breakable watermark and value merge technique |
| CN102214352A (en) * | 2010-04-09 | 2011-10-12 | 致伸科技股份有限公司 | Method and device for embedding information in input image and extracting information from input image |
| CN103123719A (en) * | 2013-03-11 | 2013-05-29 | 九江学院 | Zero watermark method for image authentication and tamper localization |
| CN103279918A (en) * | 2013-06-20 | 2013-09-04 | 海南大学 | Volume data watermark realizing method based on three-dimension DCT and chaotic scrambling |
| CN103325084A (en) * | 2013-06-27 | 2013-09-25 | 宁波大学 | Three-dimensional image fragile watermark method based on joint just noticeable distortion |
| CN103780913A (en) * | 2014-01-24 | 2014-05-07 | 西安空间无线电技术研究所 | Data compression method based on error hiding |
| CN106408495A (en) * | 2015-11-06 | 2017-02-15 | 河南师范大学 | High PSNR fragile watermarking method based on chaos theory |
| CN113014742A (en) * | 2021-02-09 | 2021-06-22 | 安徽师范大学 | Image tampering detection positioning and self-recovery method based on fragile watermark |
| CN113572913A (en) * | 2021-01-18 | 2021-10-29 | 腾讯科技（深圳）有限公司 | Image encryption method, device, medium and electronic equipment |
| CN114493976A (en) * | 2022-04-15 | 2022-05-13 | 湖北微模式科技发展有限公司 | JPEG image tampering detection method based on fragile watermark |
| CN117057971A (en) * | 2023-10-07 | 2023-11-14 | 湖北微模式科技发展有限公司 | JPEG image semi-fragile watermarking algorithm and device based on brightness shrinkage calibration |
Family Cites Families (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2002244685A (en) * | 2001-02-22 | 2002-08-30 | Kowa Co | Embedding and detection of digital watermark |
| CA2350029A1 (en) * | 2001-06-08 | 2002-12-08 | Cloakware Corporation | Sustainable digital watermarking via tamper-resistant software |
| CN1184795C (en) * | 2002-12-30 | 2005-01-12 | 上海交通大学 | Fragile digital water print embedding and extracting method based on address code |
| CN1189030C (en) * | 2003-04-03 | 2005-02-09 | 上海交通大学 | Positioned Chaos fragile digital implanting and extracting method |
| CN1246779C (en) * | 2003-06-19 | 2006-03-22 | 上海交通大学 | Embedding and extracting method for wavelet domain chaotic semi-frageile digit water print |
-
2004
- 2004-08-12 CN CNB2004100404336A patent/CN1322433C/en not_active Expired - Fee Related
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN100340111C (en) * | 2005-03-28 | 2007-09-26 | 南方医科大学 | Medical image fragile watermark method based on wavelet transform |
| CN100525365C (en) * | 2005-09-28 | 2009-08-05 | 佳能株式会社 | Image processing apparatus and control method therefor |
| CN101021938B (en) * | 2006-02-16 | 2010-09-01 | 刘粉林 | Safety image identifying method based on breakable watermark and value merge technique |
| CN102214352A (en) * | 2010-04-09 | 2011-10-12 | 致伸科技股份有限公司 | Method and device for embedding information in input image and extracting information from input image |
| CN103123719A (en) * | 2013-03-11 | 2013-05-29 | 九江学院 | Zero watermark method for image authentication and tamper localization |
| CN103123719B (en) * | 2013-03-11 | 2015-07-29 | 九江学院 | The Zero watermarking method of a kind of image authentication and tampering location |
| CN103279918A (en) * | 2013-06-20 | 2013-09-04 | 海南大学 | Volume data watermark realizing method based on three-dimension DCT and chaotic scrambling |
| CN103325084B (en) * | 2013-06-27 | 2016-03-16 | 宁波大学 | A kind of stereo-picture vulnerable watermark method just can perceiveing distortion based on associating |
| CN103325084A (en) * | 2013-06-27 | 2013-09-25 | 宁波大学 | Three-dimensional image fragile watermark method based on joint just noticeable distortion |
| CN103780913B (en) * | 2014-01-24 | 2017-01-04 | 西安空间无线电技术研究所 | A kind of data compression method based on error concealment |
| CN103780913A (en) * | 2014-01-24 | 2014-05-07 | 西安空间无线电技术研究所 | Data compression method based on error hiding |
| CN106408495A (en) * | 2015-11-06 | 2017-02-15 | 河南师范大学 | High PSNR fragile watermarking method based on chaos theory |
| CN106408495B (en) * | 2015-11-06 | 2019-10-18 | 河南师范大学 | A kind of high PSNR vulnerable watermark method based on chaology |
| CN113572913A (en) * | 2021-01-18 | 2021-10-29 | 腾讯科技（深圳）有限公司 | Image encryption method, device, medium and electronic equipment |
| CN113572913B (en) * | 2021-01-18 | 2022-08-05 | 腾讯科技（深圳）有限公司 | Image encryption method, device, medium and electronic equipment |
| CN113014742A (en) * | 2021-02-09 | 2021-06-22 | 安徽师范大学 | Image tampering detection positioning and self-recovery method based on fragile watermark |
| CN113014742B (en) * | 2021-02-09 | 2022-04-12 | 安徽师范大学 | Image tampering detection positioning and self-recovery method based on fragile watermark |
| CN114493976A (en) * | 2022-04-15 | 2022-05-13 | 湖北微模式科技发展有限公司 | JPEG image tampering detection method based on fragile watermark |
| CN114493976B (en) * | 2022-04-15 | 2022-07-01 | 湖北微模式科技发展有限公司 | JPEG image tampering detection method based on fragile watermark |
| CN117057971A (en) * | 2023-10-07 | 2023-11-14 | 湖北微模式科技发展有限公司 | JPEG image semi-fragile watermarking algorithm and device based on brightness shrinkage calibration |
| CN117057971B (en) * | 2023-10-07 | 2023-12-29 | 湖北微模式科技发展有限公司 | JPEG image semi-fragile watermarking algorithm and device based on brightness shrinkage calibration |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1322433C (en) | 2007-06-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tai et al. | Image self-recovery with watermark self-embedding | |
| Chen et al. | A fuzzy c-means clustering-based fragile watermarking scheme for image authentication | |
| Bachrach et al. | 11 Survey of Image Steganography and Steganalysis | |
| Chang et al. | A secure fragile watermarking scheme based on chaos-and-hamming code | |
| Chang et al. | Four-scanning attack on hierarchical digital watermarking method for image tamper detection and recovery | |
| CN1835018A (en) | Block fragile watermark generation and authentication method based on chaos hash function | |
| CN100346353C (en) | Generating and vertification method for electronic seal based on vulnerable water mark | |
| Surekha et al. | Sensitive digital image watermarking for copyright protection | |
| CN1322433C (en) | Positioning vulnerable water print generating and recognizing method capable of distigushing image and watermark distortion | |
| He et al. | A wavelet-based fragile watermarking scheme for secure image authentication | |
| CN102208097A (en) | Network image copyright real-time distinguishing method | |
| CN101038661A (en) | Blind watermark embedding and abstracting method based on regression support vector machine | |
| CN1517855A (en) | Image digital watermark method | |
| CN103279917A (en) | Transforming domain geometric-attack-resistant gray level image digital watermark technology | |
| Xiao et al. | An improved hierarchical fragile watermarking scheme using chaotic sequence sorting and subblock post-processing | |
| CN111882477A (en) | Self-adaptive zero-watermarking method combining visual password and enhanced singular value decomposition | |
| Liu et al. | Lossless image hierarchical recovery based on POB number system | |
| CN1189030C (en) | Positioned Chaos fragile digital implanting and extracting method | |
| CN1294531C (en) | Generating and authenticating method for positioned authenticating water-mark for recoverying distorted image contents | |
| Kang et al. | An SVD-based fragile watermarking scheme with grouped blocks | |
| CN1885339A (en) | Electronic stamp certification method based on image features | |
| CN1184795C (en) | Fragile digital water print embedding and extracting method based on address code | |
| Surekha et al. | Visual secret sharing based digital image watermarking | |
| He et al. | Self-recovery fragile watermarking using block-neighborhood tampering characterization | |
| CN1246779C (en) | Embedding and extracting method for wavelet domain chaotic semi-frageile digit water print |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20070620 Termination date: 20100812 |