CN109829846A

CN109829846A - A kind of digital picture blind watermark method based on two-dimension discrete cosine transform

Info

Publication number: CN109829846A
Application number: CN201910085504.0A
Authority: CN
Inventors: 苏庆堂; 袁子涵; 刘得成; 王环英; 王刚; 邓冠龙
Original assignee: Ludong University
Current assignee: Shenzhen Wanzhida Information Consulting Co ltd
Priority date: 2019-01-29
Filing date: 2019-01-29
Publication date: 2019-05-31
Anticipated expiration: 2039-01-29
Also published as: CN109829846B

Abstract

The invention discloses a digital image blind watermarking method based on two-dimensional discrete cosine transform by taking advantage of the strong robustness of the frequency domain digital watermarking algorithm. The invention selects some intermediate frequency coefficients of image blocks after two-dimensional discrete cosine transformation in the transform domain, and completes the embedding and blind extraction of digital watermarks by modifying the size relationship between the selected intermediate frequency coefficients. The invention embeds the digital image watermark into the host image, which not only has good watermark invisibility, but also has strong robustness, and is suitable for the occasion of digital image copyright protection.

Description

A kind of digital picture blind watermark method based on two-dimension discrete cosine transform

Technical field

The invention belongs to field of information security technology, are related to high invisibility, the strong robustness digital watermarking of digital picture Copyright protection.

Background technique

With the fast development of Internet and multimedia technology, internet is permeated in the every aspect of daily life, I Can more accurately and efficiently obtain information needed, but consequently also produce a series of serious problems, it is such as pirate, encroach right, usurp Change.Therefore, extremely urgent to the protection of intellectual property and copyright, digital watermark technology comes into being.Digital figure watermark Success be embedded in and extract can effective protection its copyright, to solve this problem.In recent years in the research of digital watermarking, tool There is digital watermarking algorithm while preferable watermark invisibility with higher robustness to be increasingly becoming digital watermark technology development Mainstream.Therefore, a kind of high invisibility how is designed, the digital watermarking algorithm of strong robustness is Contemporary Digital digital watermark One of difficult point.

Summary of the invention

The object of the present invention is to provide a kind of the digital picture blind watermark method based on two-dimension discrete cosine transform, feature It is that the realization of watermark is divided into two processes of watermark insertion and watermark extracting, watermark telescopiny is described as follows:

Step 1: the pretreatment of host image and watermarking images: being by a width sizeM×MGray scale host imageHIt is divided into size Form×mNon-overlap block of pixels；It is to a width sizeN×NGray-level Watermarking imageWIt carries out based on private keyKa _iArnold become It changes to improve the safety of watermark, and by watermarking imagesWIn each decimal system pixel value be converted to 8 binary sequences, according to The secondary length that connects into is 8N ²Watermark bit sequenceSW；

Step 2: the embedded block of selection host image: the pseudo-random sequence generated using randperm function is from host imageHIn Select block of pixelsA；

Step 3: according to formula (1), to the block of pixels of selectionAIt carries out two-dimension discrete cosine transform (DCT) and obtains transformation matrixdctA,And transformation matrix is chosen according to zigzagdctA4 pairs of DCT intermediate frequency coefficients (c _i1, c _i2), whereini=1,2,3,4, difference table ShowiTo DCT intermediate frequency coefficient；

(1)

Wherein, dct2 () is two-dimension discrete cosine transform function；

Step 4: according to sequencing from watermark sequenceSWIn successively choose 4 watermark bitsw _i；Utilize formula (2), (3), modification DCT middle frequency pair (c _i1, c _i2) between size relation to be embedded in watermark bitw _i,i=1,2,3,4；

(2)

(3)

Custom function signc () establishes rules then as follows really:

(4)

Wherein, sign () is to take sign function,avg=(abs(c _i1)+abs(c _i2))/2,TFor quantization step；

Step 5: with modified DCT middle frequency pair (c _i1 ^* _, c _i2 ^*) the original DCT middle frequency pair of replacement (c _i1, c _i2) obtain Transformation matrix containing watermarkdctA ^*, whereini=1,2,3,4, respectively indicateiTo DCT intermediate frequency coefficient；

Step 6: according to formula (5), to transformation matrixdctA ^*It carries out inverse two-dimension discrete cosine transform and obtains the pixel containing watermark Block, and by aqueous print block of pixelsUpdate host imageHIn；

(5)

Wherein, idct2 () is inverse two-dimension discrete cosine transform function；

Step 7: repeating above-mentioned second step to the 6th step, until all watermark informations are all embedded into completion, thus Obtain the host image containing watermark；

Its watermark extraction process is described as follows:

Step 1: the pretreatment of the host image containing watermark: watermark host image will be containedBeing divided into size ism×mNon-overlap Block of pixels；

Step 2: the extraction block of selection host image containing watermark: the pseudo-random sequence generated using randperm function, from aqueous Print host imageMiddle block of pixels of the selection containing watermark；

Step 3: to the aqueous print block of pixels of selectionIt carries out two-dimension discrete cosine transform and obtains transformation matrixdctA ^*, and press Transformation matrix is chosen according to zigzagdctA ^*In with 4 pairs of DCT intermediate frequency coefficients of telescopiny opposite position (c _i1 ^* _, c _i2 ^*), whereini= 1,2,3,4, respectively indicateiTo DCT intermediate frequency coefficient；

Step 4: according to DCT middle frequency pair (c _i1 ^* _, c _i2 ^*) between size relation, using formula (6) from aqueous print block of pixelsMiddle extraction watermark bit；

(6)

Wherein, abs () is ABS function,i=1,2,3,4；

Step 5: second step is repeated to the 4th step, the binary system watermark sequence containing watermarking images extracted, By watermark sequenceIn every 8 binary messages divide one group into and be converted into metric pixel value, obtain a decimal system Ordered series of numbers；

Step 6: carrying out the decimal system ordered series of numbers based on private keyKa _iInverse Arnold conversion, obtain the watermark finally extracted。

This method completes digital watermarking using the size relation between the DCT intermediate frequency coefficient in two-dimension discrete cosine transform Insertion and Blind extracting；This method had not only had preferable watermark invisibility, but also had stronger watermark robustness.

Detailed description of the invention

Fig. 1 (a), Fig. 1 (b) are two width original gradation host images.

Fig. 2 (a), Fig. 2 (b) are two width original gradation watermarking images.

Fig. 3 (a), Fig. 3 (b) are that watermark shown in Fig. 2 (a) is sequentially embedded to host image Fig. 1 (a), Fig. 1 (b) institute afterwards What is obtained contains watermarking images, and structural similarity SSIM value is successively 0.9283,0.9696, and Y-PSNR PSNR value is successively It is 42.8192dB, 41.3844dB.

Fig. 4 (a), Fig. 4 (b) are the watermark successively extracted from Fig. 3 (a), Fig. 3 (b), normalized-cross-correlation function NC value It is 1.0000,1.0000 respectively.

Fig. 5 (a), Fig. 5 (b), Fig. 5 (c), Fig. 5 (d), Fig. 5 (e), Fig. 5 (f) be will shown in Fig. 3 (a) containing watermarking images according to Secondary progress JPEG2000 compression (6:1), salt-pepper noise (0.2%), JPEG compression (60), low-pass filtering (100,4), scaling (75%), extracted watermark after the attack such as (12.5%) is sheared, normalized-cross-correlation function NC value is 0.9959 respectively, 0.9623、0.9799、0.9740、1.0000、0.9865。

Fig. 6 (a), Fig. 6 (b) are that watermark shown in Fig. 2 (b) is sequentially embedded to host image Fig. 1 (a), Fig. 1 (b) institute afterwards What is obtained contains watermarking images, and structural similarity SSIM value is successively 0.9303,0.9689, and Y-PSNR PSNR value is successively It is 42.9986dB, 41.0945dB.

Fig. 7 (a), Fig. 7 (b) are the watermark successively extracted from Fig. 6 (a), Fig. 6 (b), normalized-cross-correlation function NC value It is 1.0000,1.0000 respectively.

Fig. 8 (a), Fig. 8 (b), Fig. 8 (c), Fig. 8 (d), Fig. 8 (e), Fig. 8 (f) be will shown in Fig. 6 (a) containing watermarking images according to Secondary progress JPEG2000 compression (6:1), salt-pepper noise (0.2%), JPEG compression (60), low-pass filtering (100,4), scaling (75%), extracted watermark after the attack such as (12.5%) is sheared, normalized-cross-correlation function NC value is 0.9967 respectively, 0.9693、0.9810、0.9750、0.9999、0.9885。

Specific embodiment

Step 1: the pretreatment of host image and watermarking images: the gray scale host image for being 512 × 512 by a width sizeHIt is divided into The non-overlap block of pixels that size is 8 × 8；The Gray-level Watermarking image for being 32 × 32 to a width sizeWIt carries out based on private keyKa _i's Arnold conversion is to improve the safety of watermark, and by watermarking imagesWIn each decimal system pixel value be converted to 8 binary systems Sequence (such as: decimal number 214 can be converted into binary sequence ' 11010110 '), it is in turn connected into watermark bit sequenceSW, SWLength be 8 × 32² =8192；

Step 2: the embedded block of selection host image: the pseudo-random sequence generated using randperm function is from host imageHIn Select block of pixelsA,Herein, if the block of pixels chosenAFor；

(1)

Wherein, dct2 () is two-dimension discrete cosine transform function；At this point, to block of pixelsATwo-dimension discrete cosine transform is carried out to obtain Transformation matrixdctAFor

；

Transformation matrix is chosen according to zigzagdctA4 pairs of DCT intermediate frequency coefficients (c _i1, c _i2) be respectively (- 9.4018,11.4220), (15.9976,-5.8750),(-9.0920,13.1825),(4.7516,3.8495)；

(2)

(3)

Custom function signc () establishes rules then as follows really:

(4)

Wherein, sign () is to take sign function,avg=(abs(c _i1)+abs(c _i2))/2,TFor quantization step；At this point, from watermark SequenceSW4 watermark bits of middle selectionw _iIt is ' 0 ', ' 0 ', ' 0 ', ' 0 '；Modified DCT middle frequency pair (c _i1 ^* _, c _i2 ^*) be (- 27.2119,3.2119),(15.9976,-5.8750),(-27.9372,3.9372),(21.1005,-2.8995)；

Step 5: with modified DCT middle frequency pair (c _i1 ^* _, c _i2 ^*) the original DCT middle frequency pair of replacement (c _i1, c _i2) obtain Transformation matrix containing watermarkdctA ^*, whereini=1,2,3,4, respectively indicateiTo DCT intermediate frequency coefficient；At this point, the change containing watermark Change matrixdctA ^*For

；

(5)

Wherein, idct2 () is inverse two-dimension discrete cosine transform function；At this point, the block of pixels containing watermarkFor

；

Its watermark extraction process is described as follows:

Step 1: the pretreatment of the host image containing watermark: watermark host image will be containedIt is divided into the non-overlap that size is 8 × 8 Block of pixels；

Step 2: the block of pixels of selection host image containing watermark: the pseudo-random sequence generated using randperm function, from aqueous Print host imageMiddle block of pixels of the selection containing watermark；At this point, setting the block of pixels containing watermarkFor

；

Step 3: carrying out two-dimension discrete cosine transform to the aqueous print block of pixels of selection obtains transformation matrixdctA ^*, and press Transformation matrix is chosen according to zigzagdctA ^*In with 4 pairs of DCT intermediate frequency coefficients of telescopiny opposite position (c _i1 ^* _, c _i2 ^*), whereini= 1,2,3,4, respectively indicateiTo DCT intermediate frequency coefficient；At this point, the transformation matrix obtained after two-dimension discrete cosine transformdctA ^* For；4 chosen To DCT intermediate frequency coefficient (c _i1 ^* _, c _i2 ^*) be respectively (- 27.1567,3.2463), (15.8511, -5.5000), (- 27.7883, 4.2891),(21.0825,-2.2370)；

(6)

Wherein, abs () is ABS function, i=1,2,3,4；At this point, the watermark bit extractedRespectively ' 0 ', ' 0 ', ‘0',‘0'；

This method is simple and fast, and strong robustness, watermark invisibility are good, and the copyright suitable for digital picture as watermark is protected Shield.

Validation verification of the present invention

In order to prove effectiveness of the invention, selecting the two width sizes as shown in Fig. 1 (a), Fig. 1 (b) is 24 of 512 × 512 Standard picture is as host image, and 24 gray scales for being respectively 32 × 32 with the two width sizes as shown in Fig. 2 (a), Fig. 2 (b) Image is verified as digital watermarking.

Fig. 3 (a), Fig. 3 (b) are that watermark shown in Fig. 2 (a) is sequentially embedded to host image Fig. 1 (a), Fig. 1 (b) institute afterwards What is obtained contains watermarking images, and structural similarity SSIM value is successively 0.9283,0.9696, and Y-PSNR PSNR value is successively It is 42.8192dB, 41.3844dB；Fig. 4 (a), Fig. 4 (b) are the watermark successively extracted from Fig. 3 (a), Fig. 3 (b), normalization Cross-correlation coefficient NC value is 1.0000,1.0000 respectively；Fig. 5 (a), Fig. 5 (b), Fig. 5 (c), Fig. 5 (d), Fig. 5 (e) are by Fig. 3 (a) JPEG2000 compression (6:1), salt-pepper noise (0.2%), JPEG compression (60), low pass are successively carried out containing watermarking images shown in Filter extracted watermark after (100,4), scaling (75%), shearing (12.5%) etc. are attacked, normalized-cross-correlation function NC value It is 0.9959,0.9623,0.9799,0.9740,1.0000,0.9865 respectively.

Fig. 6 (a), Fig. 6 (b) are that watermark shown in Fig. 2 (b) is sequentially embedded to host image Fig. 1 (a), Fig. 1 (b) institute afterwards What is obtained contains watermarking images, and structural similarity SSIM value is successively 0.9303,0.9689, and Y-PSNR PSNR value is successively It is 42.9986dB, 41.0945dB；Fig. 7 (a), Fig. 7 (b) are the watermark successively extracted from Fig. 6 (a), Fig. 6 (b), normalization Cross-correlation coefficient NC value is 1.0000,1.0000 respectively；Fig. 8 (a), Fig. 8 (b), Fig. 8 (c), Fig. 8 (d), Fig. 8 (e), Fig. 8 (f) are JPEG2000 compression (6:1), salt-pepper noise (0.2%), JPEG compression are successively carried out by watermarking images are contained shown in Fig. 6 (a) (60), extracted watermark after the attack such as low-pass filtering (100,4), scaling (75%), shearing (12.5%), normalized crosscorrelation Coefficient NC value is 0.9967,0.9693,0.9810,0.9750,0.9999,0.9885 respectively.

In conclusion embedded digital figure watermark has preferable invisibility, meet watermarking algorithm can not The requirement of opinion property；Meanwhile the digital figure watermark extracted from various under fire images is with preferable identifiable and higher NC value illustrates that this method has stronger robustness, meets the needs of digital image copyright protection.

Claims

1. a kind of digital image blind watermarking method based on two-dimensional discrete cosine transform, it is characterized in that the realization of watermark is divided into two processes of watermark embedding and watermark extraction, and its watermark embedding process is described as follows:

The first step: preprocessing of host image and watermark image: divide a grayscale host image H of size M × M into non - overlapping pixel blocks of size m × m ; The watermark image W is subjected to _Arnold transformation based on the private key Kai to improve the security of the watermark, and each decimal pixel value in the watermark image W is converted into an 8-bit binary sequence, which is sequentially concatenated into watermark bits with a length of 8 N ² sequence SW ;

The second step: select the embedded block of the host image: use the pseudo-random sequence generated by the randperm function to select the pixel block A from the host image H ;

Step 3: According to formula (1), perform two-dimensional discrete cosine transform (DCT) on the selected pixel block A to obtain the transformation matrix dctA, and select 4 pairs of DCT intermediate frequency coefficients ( c _i1, c _i2 ) of the transformation matrix dctA according to the zigzag shape ), where i = 1, 2, 3, 4, respectively representing the i -th pair of DCT intermediate frequency coefficients;

(1)

Among them, dct2(.) is a two-dimensional discrete cosine transform function;

The fourth step: select 4 watermark bits w _i from the watermark sequence SW in sequence; use formulas (2) and (3) to modify the magnitude relationship between the DCT intermediate frequency coefficient pairs ( c _{i1 ,} c _i2 ) to embed Watermark bit w _i , i =1,2,3,4;

(2)

(3)

The determination rules of the custom function signc(.) are as follows:

(4)

Among them, sign(.) is the sign function, avg =( abs ( c _i1 )+ abs ( c _i2 ))/2, T is the quantization step size;

The fifth step: replace the original DCT intermediate frequency coefficient pair ( c _{i1 ,} c _i2 ) with the modified DCT intermediate frequency coefficient pair ( c _i1 ^* _, c _i2 ^* ) to obtain the watermark transformation matrix dctA ^* , where i =1, 2, 3, and 4, respectively represent the i -th pair of DCT intermediate frequency coefficients;

Step 6: According to formula (5), perform an inverse two-dimensional discrete cosine transform on the transformation matrix dctA ^* to obtain a pixel block with a watermark , and will watermark the pixel block Update to the host image H ;

(5)

Among them, idct2(.) is the inverse two-dimensional discrete cosine transform function;

Step 7: Repeat the above steps 2 to 6 until all the watermark information is embedded, thereby obtaining a watermarked host image ;

The watermark extraction process is described as follows:

Step 1: Preprocessing of the watermarked host image: Convert the watermarked host image into non-overlapping pixel blocks of size m × m ;

Step 2: Select the extraction block of the watermarked host image: use the pseudo-random sequence generated by the randperm function, from the watermarked host image Select the watermarked pixel block in ;

The third step: for the selected watermark pixel block Perform a two-dimensional discrete cosine transform to obtain a transformation matrix dctA ^* , and select 4 pairs of DCT intermediate frequency coefficients ( c _i1 ^* _, c _i2 ^* ) corresponding to the position of the embedding process in the transformation matrix dctA ^* according to the zigzag shape, where i =1,2 , 3, 4, respectively represent the i -th pair of DCT intermediate frequency coefficients;

Step 4: According to the size relationship between the DCT intermediate frequency coefficient pairs ( c _i1 ^* _, c _i2 ^* ), use formula (6) to extract the watermark from the pixel block Extract the watermark bit ;

(6)

Among them, abs(.) is the absolute value function, i =1,2,3,4;

Step 5: Repeat steps 2 to 4 to obtain the binary watermark sequence of the extracted watermark image , the watermark sequence Each 8-bit binary information is grouped into a group and converted into a decimal pixel value to obtain a decimal number sequence;

Step 6: _Perform the inverse Arnold transform based on the private key Kai on the decimal sequence to obtain the final extracted watermark .