CN104484850A

CN104484850A - Robust image watermark detecting method based on fuzzy classification

Info

Publication number: CN104484850A
Application number: CN201410793176.7A
Authority: CN
Inventors: 王向阳; 刘宇男; 牛盼盼
Original assignee: Liaoning Normal University
Current assignee: Liaoning Normal University
Priority date: 2014-12-20
Filing date: 2014-12-20
Publication date: 2015-04-01

Abstract

The invention discloses a robust image watermark detecting method based on fuzzy classification. The robust image watermark detecting method comprises the following steps: in a watermark-containing image correcting process, extracting an eight-circumcircle lower PHT matrix (low 7-order), obtained by circumcircle lower PHT decomposition, of a to-be-detected image, taking the eight-circumcircle lower PHT matrix as a training characteristic to combine with a fuzzy support vector machine theory, training a sample to obtain a training module for performing geometric correction. The fuzzy vector machine FSVM has very good learning capacity and the matrix obtained by PHT decomposition has high precision, so that watermark information under different attacks can be correctly extracted, and the watermark robustness is improved. Moreover, the method has the characteristics of simple calculation, no need of an initial carrier during watermark extraction, and the like, so that the copyright protection practicability of digital image works is strengthened.

Description

Based on the robust image watermark detection method of fuzzy classification

Technical field

The invention belongs to Information hiding and digital watermark technology field in multi-media information security, especially one not only has preferably not sentience, and all has the robust image watermark detection method based on fuzzy classification of good robustness to normal signal process (medium filtering, edge sharpening, superimposed noise and JPEG compression etc.) and desynchronization attack (rotation, translation, convergent-divergent, shearing, upset etc.).

Background technology

Digital watermarking (Digital Watermarking) is as effective means of supplementing out economy of conventional encryption methods; it is a kind of new technology can protecting copyright and certification source and integrality under open network environment; cause people to pay much attention in recent years, and become a focus of international academic community research.So-called digital figure watermark; the mark (watermark) of certain sense will be had exactly; the method utilizing data to embed is hidden in digital picture product; in order to prove the entitlement of creator to its works; and as qualification, the illegal foundation of encroaching right of prosecution; simultaneously by ensureing the complete reliability of numerical information to the determination and analysis of watermark, thus become intellectual property protection and the false proof effective means of digital multimedia.

So-called desynchronization attack, not refers to that this kind of attack can remove watermark information from containing watermarking images, and refers to that it can destroy synchronous (namely the changing watermark embedment position) of digital watermarking component, thus cause detecting device to can not find effective watermark.In recent years, propose a series of digital image watermark detection method successively, but regrettably, existing digital image watermark detection method is mainly primarily focused in the research of antagonism normal signal process (as lossy compression method, low-pass filtering, noise etc.), and such as rotations, convergent-divergent, translation, ranks are removed, shearing, etc. the opposing effect of geometric attack bad.

Summary of the invention

The present invention is directed to the upper technical matters existing for prior art, there is provided one not only to have preferably not sentience, and to normal signal process (medium filtering, edge sharpening, superimposed noise and JPEG compression etc.) and desynchronization attack (rotation, translation, convergent-divergent, shearing, upset etc.), all there is the embedding of the Color digital watermarking based on machine learning and the detection method of good robustness.

Technical solution of the present invention is: a kind of robust image watermark detection method based on fuzzy classification, it is characterized in that carrying out as follows:

Step 1: embed watermark in original image, obtains containing watermarking images F;

Step 2: to carrying out X-translation, Y-translation containing watermarking images F, Rotation and Zoom operates, with the image after operating for training image constructs training sample set , , under carrying out circumscribed circle to each training image in sample set, PHT decomposes, and PHT square (low 7 rank) under calculating each training image 8 circumscribed circles, obtain 8 proper vectors reflecting corresponding training image feature, use relevant information coefficient , state each training image, obtain training pattern;

Step 3: treat detected image under carrying out circumscribed circle, PHT decomposes, and calculates image to be detected 8 circumscribed circles under PHT square (low 7 rank), obtain 8 proper vectors reflecting characteristics of image to be detected;

Step 4: utilize training pattern to treat detected image proper vector train, obtain a geometric transformation parameter , utilize this geometric transformation parameter treat detected image carry out inverse transformation, obtain the image after correcting ;

Step 5: the image after correction middle extraction watermark.

Described step 1 is as follows:

Step 11: get and be of a size of gray level image be carrier image , most-significant byte plane picture is extracted for carrier image;

Step 12: carry out non-lower sampling shearlet conversion to extracted most-significant byte plane picture, obtains high-frequency sub-band coefficient and low frequency sub-band coefficient;

Step 13: utilize HVS to choose quantization step, utilizes the low frequency sub-band that quantization method will extract in watermark embedment to step 12;

Step 14: convert against shearlet according to non-lower sampling, merges the low-frequency image of embed watermark and high frequency imaging, obtains the high bit-planes image of embed watermark.

Described step 12 is as follows:

Step 121: to the most-significant byte plane picture obtained, carries out one-level non-lower sampling shearlet and converts, obtain a low frequency sub-band with four direction high-frequency sub-band;

Step 122: by the low frequency sub-band obtained coefficient is divided into block, after carrying out piecemeal process, the size of each fritter is P*Q, obtains the low frequency sub-band coefficient after piecemeal .

Described step 13 is as follows:

Step 131: the low frequency coefficient that non-lower sampling shearlet to be modified is converted, be quantified as odd number or even number with embed watermark according to quantization step and its corresponding bit watermark information (1 or 0), watermark embedment formula is as follows:

Described step 2 is as follows:

Step 21: under carrying out circumscribed circle to each training image in sample set, PHT decomposes, and its process of decomposing is:

Exponent number is n, multiplicity is l, and pCET be defined as:

Wherein, represent conjugate complex number, represent the original image function under polar coordinate system, basis function be broken down into radial polynomial and angle polynomial expression, as follows

Radial polynomial directly provides with complex exponential form, as follows

And meet orthogonality condition

And have

Wherein, for normalized factor, kronecker symbol, represent conjugation.Therefore, the image function under polar coordinate system pHT can be expressed as:

Step 22: under calculating 8 circumscribed circles of each training image, PHT square (low 7 rank) is respectively: P (0,1), P (0,2), P (1,0), P (1,1), P (1,2), P (2,0), P (2,1), P (2,2), these 8 proper vectors reflecting corresponding training image feature are obtained;

Step 23: utilize in step 22 proper vector extracting image to carry out yojan, obtain training sample set, and then to train FSVM model be training pattern.

Described step 5 is as follows:

Step 51: calculate image to be detected high bit-planes figure obtain ;

Step 52: to the high bit-planes image obtained , carry out non-lower sampling Shearlet conversion, obtain a low frequency sub-band with several high-frequency sub-band;

Step 53: utilize quantization method to extract watermark, utilize HV to choose quantization step, first extract according to the step-by-step of watermark extracting formula, draw individual watermark, watermark extracting formula is as follows:

Wherein: , for downward bracket function, for what extract individual watermark;

Step 54: according to , according to the value of individual watermarking images correspondence position and be averaged and obtain watermarking images

；

Step 55: take intermediate grey values as threshold value, by following formula Gray-level Watermarking image be converted into binary bitmap ,

be the digital watermarking image detected.

The present invention is containing in watermarking images testing process, extract PHT square (low 7 rank) make it as training characteristics, in conjunction with fuzzy support vector machine theory, to sample training under 8 circumscribed circles of original image after PHT under circumscribed circle decomposes, draw training pattern, thus carry out geometry correction.Because fuzzy vector machine FSVM has good learning ability and PHT decomposes the accuracy that the square of gained has height; so watermark information all can correctly be extracted under difference is attacked; thus improve the robustness of watermark; the present invention has simultaneously and calculates the feature such as simple, enhances its practicality for digital picture Works copyright protection.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the embodiment of the present invention.

Embodiment

As shown in Figure 1, method of the present invention is carried out in accordance with the following steps:

Concrete steps are as follows:

Step 13: utilize HVS to choose quantization step, utilizes the low frequency sub-band that quantization method will extract in watermark embedment to step 12:

Wherein: original low-frequency coefficients, quantize amended low frequency coefficient, , round is the bracket function that rounds off, and mod is modular arithmetic function, for quantization step.

Concrete steps are as follows:

Exponent number is n, multiplicity is l, and pCET be defined as:

Radial polynomial directly provides with complex exponential form, as follows

And meet orthogonality condition

And have

Concrete steps are as follows:

Step 41: the mathematical model setting up geometry correction, sets up the mapping relations between image slices point coordinate (row, column number) and object space (or reference diagram) corresponding point coordinate;

Step 42: solution asks the unknown parameter in mapping relations, then corrects each pixel coordinate of image according to mapping relations;

Step 43: the gray-scale value (gray scale interpolation) determining each pixel.

Step 5: the image after correction middle extraction watermark:

Step 5 is specific as follows:

Step 51: calculate image to be detected high bit-planes figure obtain ;

；

be the digital watermarking image detected.

Claims

1., based on a robust image watermark detection method for fuzzy classification, it is characterized in that entering as follows

OK:

Step 2: to carrying out X-translation, Y-translation containing watermarking images F, Rotation and Zoom operates, with the image after operating for training image constructs training sample set , , under carrying out circumscribed circle to each training image in sample set, PHT decomposes, and PHT square under calculating each training image 8 circumscribed circles, obtain 8 proper vectors reflecting corresponding training image feature, use relevant information coefficient , state each training image, obtain training pattern;

Step 3: treat detected image under carrying out circumscribed circle, PHT decomposes, and calculates image to be detected 8 circumscribed circles under PHT square, obtain 8 proper vectors reflecting characteristics of image to be detected;

Step 5: the image after correction middle extraction watermark.

2. the robust image watermark detection method based on fuzzy classification according to claim 1, is characterized in that described step 1 is as follows:

Step 13: utilize HVS to choose quantization step, utilizes the low frequency sub-band that quantization method will extract in watermark embedment to step 12,

3. the robust image watermark detection method based on fuzzy classification according to claim 2, is characterized in that described step 12 is as follows:

4. the robust image watermark detection method based on fuzzy classification according to claim 2, is characterized in that described step 13 is as follows:

Step 131: the low frequency coefficient that non-lower sampling shearlet to be modified is converted, be quantified as odd number or even number with embed watermark according to quantization step and its corresponding bit watermark information, watermark embedment formula is as follows:

5. the robust image watermark detection method based on fuzzy classification according to claim 3, is characterized in that described step 2 is as follows:

Exponent number is n, multiplicity is l, and pCET be defined as:

Radial polynomial directly provides with complex exponential form, as follows

And meet orthogonality condition

And have

Wherein, for normalized factor, kronecker symbol, represent conjugation;

Therefore, the image function under polar coordinate system pHT can be expressed as:

Step 22: under calculating 8 circumscribed circles of each training image, PHT square is respectively: P (0,1), P (0,2), P (1,0), P (1,1), P (1,2), P (2,0), P (2,1), P (2,2), these 8 proper vectors reflecting corresponding training image feature are obtained;

6. the robust image watermark detection method based on fuzzy classification according to claim 5, its feature

Be that described step 5 is as follows:

Step 51: calculate image to be detected high bit-planes figure obtain ;

；

be the digital watermarking image detected.