CN110648271A

CN110648271A - Method for embedding digital watermark in halftone image by using special dots

Info

Publication number: CN110648271A
Application number: CN201910926398.4A
Authority: CN
Inventors: 周博文
Original assignee: Individual
Current assignee: Individual
Priority date: 2019-09-27
Filing date: 2019-09-27
Publication date: 2020-01-03

Abstract

The present invention proposes a method of embedding a digital watermark in a halftone image, and reading the watermark from a photographed image. The process of embedding a watermark in a halftone image is: carrying out operations such as asymmetric encryption, LDPC encoding and the like on watermark information to generate a bit matrix; and carrying out image halftone by using the dots with special shapes, and encoding the matrix information into the image dots to obtain the halftone image with the watermark information. The process of reading out the watermark from the captured picture is: the picture is cut into 256-256 pixel blocks, bit information coded in all the blocks is gathered to generate a bit matrix, and the bit matrix is subjected to folding, averaging and threshold operation, cyclic translation, exclusive or with masks, LDPC decoding and decryption to obtain original watermark information. Compared with the existing image watermarking method, the method can embed more information in the image, has good confidentiality and robustness, and does not change the color of the image.

Description

Method for embedding digital watermark in halftone image by using special dots

Technical Field

The invention belongs to the field of image digital watermarking, and particularly relates to an image watermarking method based on image halftoning.

Background

Image digital watermarks are invisible information embedded in an image that can be identified using specific devices and software. Digital watermarks are widely introduced in the fields of image copyright protection, product anti-counterfeiting and the like. Most of the existing commonly used digital watermarking algorithms process frequency domain information or color statistical information of an image and encode the information into the frequency domain or histogram information of the image.

The digital watermarking method widely used at present has some defects: the amount of coded information is small, the robustness is low, and the color of the image is changed. The invention uses the dots with special shapes to carry out image halftone, watermark information generates a bit matrix through operations such as encryption, LDPC coding and the like, and then the information of the bit matrix is coded into the image dots. The information quantity embedded in one image is far larger than that of the common digital watermarking technology at present, the robustness of the watermark is greatly improved due to the LDPC coding technology, the watermark information can be still correctly read even if the image is deformed or partially stained, and in addition, the printed appearance of the halftone image generated by the method is consistent with that of the traditional halftone image.

Disclosure of Invention

In order to overcome the defects of the existing image digital watermarking technology, the invention provides a method for carrying out image halftone by using special dots, coding watermark information into the image dots and extracting digital watermarks from images by using specially designed software.

Compared with the existing image watermarking method, the method provided by the invention has the following four advantages: the embedded watermark information quantity is large, 728-3150-bit watermark information can be embedded into one picture by the method, and compared with the method, the Digimar image watermark method can only embed 32-bit information into one picture at most; the confidentiality is good, and the watermark information can be ensured not to be forged through asymmetric encryption; the robustness is higher, the LDPC coding is used for the watermark information, and even if the picture is stained or deformed, the watermark information can still be correctly extracted from the picture; the method provided by the invention has the advantages that the image watermarking algorithm widely used at present mostly changes the frequency domain information or the color statistical information of the image to cause the change of the image appearance, particularly the change is particularly obvious for the light-colored image with smooth tone, the method provided by the invention uses a special network to perform halftone processing on the image, and the color and the brightness of the image are not different from the traditional network printing effect after the image is printed on the surface of a medium.

The general process of embedding and retrieving a watermark in and from an image as proposed by the present invention is shown in fig. 1: and (3) after the binary string of the watermark information is subjected to operations of asymmetric encryption, LDPC coding, XOR with random mask, head addition of a fixed bit string and the like, obtaining a bit matrix E, and modulating the binary information in the matrix E into dots of a halftone image to obtain the image with the watermark.

The image with the watermark is printed on various product media, after the image is shot by using camera equipment, the binary information coded in the network points is obtained by using a specially designed algorithm through image segmentation, Fourier transform and the like, and then the original watermark information is obtained through operations of LDPC decoding, decryption and the like.

Drawings

FIG. 1: and embedding watermark information into the halftone image and reading the watermark information from the shot picture.

FIG. 2: an overall flow chart for embedding watermark information into a halftone image.

FIG. 3: a flow chart for obtaining noisy bit information from a captured image.

FIG. 4: and decoding and recovering the original watermark information from the bit information with noise.

Detailed Description

The method for embedding the digital watermark in the image by using the special dots comprises the following steps.

1) For bit string A of original watermark information, using asymmetric encryption method to form bit string A_CThe secret key used is K_EAs shown in fig. 2.

2) Using LDPC coding, the bit string A_CThe code is a bit string B, and LDPC code has high error correction performance and can improve the robustness of the watermark, as shown in FIG. 2.

3) In order to avoid a regular pattern in the string B, the string B is obtained by carrying out XOR operation on the string B and a preset random mask string M with higher entropy_MAs shown in fig. 2.

4) In B_MAdding a preset positioning string C in the front to generate all information D to be coded into the mesh points, namely D = [ C B =_M]As shown in fig. 2.

5) All bits of the string D are n × n bits in total, and are arranged into an n × n matrix E, and the value of n may be any even number between 32 and 64 according to different application environments, as shown in fig. 2. Note that in fig. 2, an 8 × 8 matrix is used as a schematic for clarity and readability.

6) From the information in matrix E, a special shape of the dots S is used to generate a threshold matrix H, where the dots S have two mutually perpendicular possible directions, and the dot direction in the threshold matrix H encodes the bit information in E, as shown in fig. 2. And (4) carrying out halftone processing on the image by using the threshold matrix H to obtain a halftone image Q with watermark information.

By the above steps, the process of embedding the watermark information a in the image P is completed. The generated halftone image Q with watermark information can be printed on different media by using a suitable printing device to generate documents, certificates, product packages and the like, and can also be issued through the internet and displayed on a screen. The specific identification software is installed on the equipment with the photographing function, the watermark information in the image can be automatically identified, and the authenticity of the document, the certificate and the product can be identified according to the existence and the correctness of the watermark information. The image Q may also be made an advertisement and the user may read the information in the picture using specific software installed on the cell phone or be directed to a designated website.

In step 1 above, the encryption key K is used_EThe original information A is encrypted, so that the aim of confidentiality can be fulfilled, and different users can be distinguished. The invention uses asymmetric encryption method to encrypt information, and different users use different encryption and decryption key pairs. For example, the secret key pair for user a is<K_Ea，K_Da >Using K_EaEncrypting watermark information, and using K in identifying watermark_DaDecrypt the information and user b uses the key pair<K_Eb,K_Db>. They use different keys, so they can only make their own dedicated watermarked picture, but cannot forge the watermark of the other party.

After the image Q with the watermark information is printed on a medium, an image I can be shot by a camera device, and the algorithm for reading the watermark information from the image comprises the following steps.

1) I is divided into image blocks I of size 256 pixels by 256 pixels₁,I₂…. As shown in fig. 3.

2) For each figureImage block I_iFourier transform is carried out to obtain an image block I_iFrequency domain image F_i. If the image block I_iComprising a part of image Q, since the dots of image Q are arranged in regular rows and columns, F_iThere will be two signals s significantly above the background at a particular angular and frequency position₀And s₁，s₀And s₁Is nearly vertical and the frequencies are approximately equal as shown in fig. 3. Signal s₀And s₁It is determined whether the image block contains key features of the watermark information. If there is no signal s satisfying the above-mentioned characteristics in the frequency domain signal of a certain image block₀And s₁Then the image block may be discarded without further processing.

3) For signals containing s₀And s₁Image block of, assuming its signal s₀Has an angle of a₀Frequency of f₀Phase of phi₀Signal s₁Has an angle of a₁Frequency of f₁Phase of phi₁According to a₀ ，f₀，φ₀And a₁ ，f₁，φ₁The image block is divided into many small parallelogram blocks as shown in fig. 3. Each small parallelogram block comprises a mesh point, and the bit information coded by the mesh point can be judged to be 0 or 1 according to the direction of the mesh point. If the dot is not clear enough due to dirt or noise, the direction of the dot cannot be judged, and the bit information is set to be null. Thus, for each tile I_iA corresponding bit block J can be obtained_iAs shown in fig. 3.

4) Bit information J of all image blocks_iAfter the aggregation, a large bit matrix L is formed by splicing, as shown in fig. 4. Since part of the image positions are not read to the watermark signal, part of the positions of L are empty.

5) Dividing the matrix L into small n × n matrices L₀，L₁… …, the matrix N is calculated using the following equation:

(equation 1).

6) Matrix N is matrix E_NCyclic translation of (2), wherein E_NRefers to the matrix E contaminated by noise. Searching a positioning string C in N, wherein the coordinate (a, b) of the first bit of the string C in N is a matrix E_NCircularly translating N, leftwards by a units and upwards by b units to obtain a matrix E_NAs shown in fig. 4.

7) Will E_NSpread into a string D_N，D_NThe string D is polluted by noise, and after the former positioning information is deleted, the information B is obtained_MN，B_MNIs a string B contaminated by noise_MAs shown in fig. 4.

8) Will bunch B_MNXOR'd with mask string M to get string B_N，B_NIs string B contaminated by noise.

9) Since string B uses LDPC coding, as long as B_NDoes not exceed the tolerance threshold of LDPC, can be obtained from B by using LDPC decoding algorithm_NRecovery information A_C。

10) Using a decryption key K_DFor information A_cAnd decrypting to obtain the bit string A of the original watermark information.

Claims

1. A method for embedding a digital watermark in a halftone image using special dots, the method comprising: the digital watermark information is modulated in halftone dots of the halftone image, the halftone image containing the watermark information can be printed on various products, pictures are shot to obtain the pictures, and the watermark information is extracted from the pictures by using a specially designed algorithm.

2. The method of claim 1, wherein the halftone image is halftoned with a special shape of halftone dots S, the halftone dots S having two mutually perpendicular possible directions and being used for encoding binary information, and the watermark information being encrypted and encoded and modulated into the halftone dots in the halftone image.

3. Method according to claims 1 and 2, characterized in that, to ensure that the watermark cannot be forged, the bit string a of the watermark is first subjected to an asymmetric encryption, generating a ciphertext bit string a_CThe encryption key used is K_EIn order to make the watermark more robust, ciphertext A_CGenerating a bit string B through LDPC coding, and carrying out XOR operation on the bit string B and a preset random mask bit string M with higher entropy to obtain the bit string B in order to avoid the possibility that the bit string B contains a regular pattern_M。

4. Method according to claim 3, characterized in that the bit string B_MAdding a preset fixed positioning bit string C in front of the image to generate a bit string D, arranging the bit string D into an n multiplied by n bit matrix E, using a halftone dot S to perform halftone processing on the image, modulating binary information of the matrix E into the direction of the halftone dot S, and generating a halftone image Q with watermark information.

5. The method according to claims 1 to 4, wherein the watermark image Q generated by the method of the above claims can be printed on different media, the image I of the product can be shot by using a device such as a mobile phone, and the device is provided with specially designed software for extracting watermark information from the shot image.

6. The method of claim 5, wherein the captured image is divided into equal-sized blocks I_iFor each block I_iFourier transform is carried out to obtain a frequency domain image F_iAccording to F_iThe signal characteristics of the image block can be judged whether the image block contains the mesh points S or not, and the arrangement mode direction and frequency of the mesh points S, and the image block packet is obtained from the angles of all the mesh points of the image blockContaining bit information J_i。

7. Method according to claim 6, characterized in that from all image blocks I_{i get}Bit block J of_iAre collected and spliced into large bit blocks L, which are divided into a number of n x n bit matrices L_iAnd from these L_iAnd generating a bit matrix N.

8. The method of claim 7, wherein searching the matrix N for the head position of the positioning string C, C is followed by matrix E_NSo that the matrix E can be restored_N，E_NIs a matrix E contaminated by noise_NSpread out to obtain a string D_NRun of string D_NIs a string D contaminated by noise, deleting D_NAfter the positioning string C of the head, a string B is obtained_MN，B_MNIs a string B contaminated by noise_M；B_MNPerforming exclusive or operation on the mask string M to obtain a string B_NRun of Chinese characters B_NIs string B contaminated by noise.

9. The method of claim 8, wherein the decoding is performed from B using LDPC decoding_NObtain a string A_CSince LDPC has high error correction capability, even B_NIs contaminated by noise, and bit string A can still be correctly recovered as long as the noise level does not exceed the tolerance value of LDPC_C。

10. Method according to claim 9, characterized in that the use and encryption key K is used_ECorresponding decryption key K_DTo A, a_CAnd decrypting to obtain the original watermark bit string A.