CN111583085A - Color image digital watermarking method based on transform domain - Google Patents

Abstract

The invention discloses a color image digital watermarking method based on a transform domain, which comprises the following steps: preprocessing a spline image; a watermark embedding process; and (5) watermark extraction. The other method comprises the following steps: watermark Arnold scrambling; a watermark embedding process; and (5) watermark extraction. The invention can effectively resist noise and object modification attack on the premise of ensuring that the visual quality of the carrier data is not reduced and meeting the transparency of the watermark.

Description

Color image digital watermarking method based on transform domain

Technical Field

The invention relates to the field of digital watermarking, in particular to a color image digital watermarking method based on a transform domain.

Background

Vector graphics, also known as object-oriented images or drawing images, use lines and curves to describe graphics, and are represented by geometric primitives such as points, lines, arcs, polygons, etc. based on mathematical equations. The vector diagram can be printed and output with the highest resolution, and the vector diagram is used for drawing in many industrial designs, billboard patterns and signs. Unlike the common raster image, the vector image has the following advantages: (1) regardless of the resolution, any reduction or enlargement of the vector image does not cause a change in the image definition. (2) The vector map can smooth vector lines and color edges, keeping the lines at the same scale. (3) The vector diagram stores line and color information, and is only related to complexity, so that the vector file storage space is small. (4) The vector data has high precision. Therefore, the digital vector diagram is the most important data of a Geographic Information System (GIS), and has been widely applied to the fields of navigation, cadastral management, digital cities, intelligent transportation and the like.

The slope vector transformed by Slant is in a discrete staggered tooth waveform with uniform step descending. The Slant transform is computationally simple and aims to match the basis vectors to regions of constant luminance slope, and the Slant transform has been applied to many image processing such as transform coding and image restoration. In terms of coding, the Slant transform is considered as a sub-optimal orthogonal transform for energy compression. The sign of most if coefficients in the Slant transform remains unchanged before and after JPEG compression and Gaussian noising.

Wavelet transform has been very important in image processing in recent years, and is often used in image compression, feature detection, and texture analysis. The wavelet transformation can carry out multi-scale research on the local characteristics of the signals in a time domain and a frequency domain at the same time, and can examine the frequency domain characteristics of the local time domain and the time domain characteristics of the local frequency domain process. Lifting wavelets based on the conventional wavelet transform were proposed by Swelden in 1996. Lifting wavelets have found wide application in the field of signal processing: in the video field, an object with any shape can be adaptively coded by using a lifting wavelet method, the coding efficiency is obviously improved, and the subjective evaluation effect is better; in terms of still image processing, the lifting wavelet has better compression performance at low bit rate, and can provide effective distortion and distortion-free compression in the same coding structure. The lifting method for constructing wavelet transform comprises three steps: and (4) subdivision, prediction and update.

For images, the main features of the human visual system are generally embodied in three aspects of image brightness characteristics, frequency domain characteristics and image type characteristics. The brightness characteristic refers to the sensitivity of human eyes to brightness changes, and generally, the sensitivity of human eyes to noise in high-brightness areas is small, so that the higher the brightness of the background of an image is, the more additional information can be embedded. For the frequency domain characteristics, the image is transformed from the spatial domain to the spectral domain, and the human eye is less sensitive to high frequency content. From the image type characteristics, the human visual system is more sensitive to smooth regions than to regions with dense textures, so the denser the image texture, the more information it represents can be embedded.

Disclosure of Invention

The invention provides a color image digital watermarking method based on a transform domain, which can effectively resist noise and object modification attack on the premise of ensuring that the visual quality of carrier data is not reduced and the transparency of the watermark is met, and is described in detail as follows:

a transform domain-based color image digital watermarking method, the method comprising:

obtaining transformation matrix of embedded watermark and executing inverse Slant transformation to obtain LL of embedded watermark information^*Sub-band, LL^*Combining the sub-bands with LH, HL and HH components to form a color image containing watermark information;

analyzing the color image along the reverse process of color image generation to obtain the attributes of left _ color and right _ color elements, and writing the attributes into a vector diagram XML file;

performing discrete wavelet transformation on a color image to be detected and an initial color image to respectively obtain low-frequency sub-bands and performing Slant transformation to obtain a transformation matrix;

sorting the two groups of transformation coefficients in the shape of Chinese character 'ji', extracting watermark information, and carrying out inverse DCT transformation to obtain a watermark image.

The obtaining of the transformation matrix of the embedded watermark specifically includes:

loading an XML file of the vector image, generating a corresponding DOM tree, and respectively obtaining attribute values of B, R and G of a left _ color element and a right _ color element; forming a two-dimensional array by taking values and quantizing the attribute values B, R and G;

performing DCT transformation on the watermark image once to generate a transformation matrix, wherein different frequency parts are distributed at different positions of the transformation matrix after transformation; performing discrete wavelet transform on the color image C (x, y) to obtain a low-frequency sub-band LL;

performing Slant transformation on the low-frequency sub-band LL to obtain a transformation matrix LL _ ST (u, v); for the obtained transform coefficients LL _ ST and W_dctSorting W according to zigzag, and adding W according to addition principle_dctIs embedded in the low frequency position of the corresponding LL _ ST (u, v), and a transformation matrix for embedding the watermark is generated.

Further, the addition principle specifically includes:

wherein m is_iRepresenting a watermark image W_dctCoefficient of (a), x_iRepresenting the coefficients of LL _ ST (u, v) after the LL sub-band Slant transform,

indicating after watermark embedding LL _ ST^*The coefficient of (u, v), oc is an embedding intensity factor.

The character sorting of the two groups of transformation coefficients and the extraction of the watermark information are specifically as follows:

a transform domain-based color image digital watermarking method, the method comprising:

merging the three-color components after scrambling to obtain a watermark image after Arnold scrambling, performing Arnold encryption and three-color-base separation operations, and performing Slant transformation on the three-color-base components respectively to obtain a transformation coefficient;

performing two-stage lifting wavelet transform on the original image to obtain LH₂Carrying out three-color base separation on the sub-bands and respectively carrying out Slant transformation to obtain respective transformation matrixes;

combining with a human visual system to adopt different embedding intensities for different color components to carry out watermark embedding operation;

performing inverse Slant transformation on the three-color basis embedded with the watermark, combining the three-color basis components, and performing lifting wavelet inverse transformation to generate an image containing watermark information;

respectively carrying out two-stage lifting wavelet transform on the original image to be detected and the obtained LH₂Separating the three color components, and respectively carrying out Slant transformation to obtain transformation matrixes;

extracting the watermark, performing one-time inverse Slant transformation on the extracted watermark information, combining the three-color basis components, and extracting the watermark image through Arnold decryption operation.

Wherein, the combination of the human visual system to different color components with different embedding intensity, the watermark embedding operation is specifically:

S_I^* _R(G,B)(i,j)＝S_I_R(G,B)(i,j)+α_R(G,B)×S_W_R(G,B)(i,j)

wherein, α_R(G,B)Is the embedding strength, S _ W_R(G,B)(I, j) is the transform coefficient, S _ I_R(G,B)(I, j) are respective transformation matrices, S _ I^* _R(G,B)And (i, j) is the transformation matrix after embedding the watermark.

Further, the extracting the watermark specifically includes:

wherein, S _ W_R(G,B) ^*And (i, j) extracted watermark information.

The technical scheme provided by the invention has the beneficial effects that:

1. the invention can effectively ensure the invisibility of the watermark and has better robustness to common noise attack, JPEG compression, scaling and shearing attack;

2. the invention can utilize the visual masking phenomenon to meet the characteristic that the watermark is perceived to be invisible; watermark information is widely distributed in the frequency spectrum domain of the whole image, which is beneficial to ensuring the concealment; strong capability of resisting attack and the like.

Drawings

FIG. 1 is a flow chart of a spline image digital watermarking method based on DCT, DWT and Slant transformation;

fig. 2 is a flowchart of a double-color image digital watermarking method based on LWT and Slant transformation.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention are described in further detail below.

Example 1

Spline image digital watermarking method based on DCT, DWT and Slant transformation

Wherein, DCT: discrete Cosine transform (Discrete Cosine transform form); DWT: discrete Wavelet Transform (Continuous Wavelet Transform); slant: and (5) oblique transformation.

A1: preprocessing a spline image;

b1: a watermark embedding process;

c1: and (5) watermark extraction.

In one embodiment, step a1 preprocesses the spline image, which includes the following steps:

"left color" represents the color value on the left side of the spline curve, "right color" represents the color value on the right side of the spline curve, and attribute information such as B, R, G, etc. represents the color value of the cubic bezier spline curve.

And acquiring the attribute values of B, R and G of the left _ color element and the right _ color element, and forming a two-dimensional array f (x, y) by the attribute values of B, R and G through value taking and quantization. Because the GRB color values are all integer type, (x, y) are discrete coordinates, and f is a discrete amplitude, it can be regarded as a pixel value of a certain gray image, which is called color image C in the present method. This can be considered as the pixel value of a certain gray scale image, which is referred to as color image C in the present invention.

In one embodiment, the step B1 performs watermark embedding on the basis of the step a1, and includes the following specific steps:

watermark image W (i)J) performing a DCT transformation to generate a transformation matrix W_dctThe different frequency portions are transformed and distributed at different positions of the transformation matrix. The color image C (x, y) is subjected to discrete wavelet transform to obtain a low frequency subband LL. Subsequently, the low frequency subband LL is subjected to a Slant transform to obtain a transform matrix LL _ ST (u, v). For the obtained transformation matrix LL _ ST (u, v) and transformation matrix W_dctSorting W according to zigzag, and adding W according to addition principle_dctIs embedded in a low-frequency position corresponding to LL _ ST (u, v), and generates a watermark-embedded transformation matrix LL _ ST^*(u, v). For LL _ ST^*(u, v) performing an inverse Slant transform to obtain a LL with embedded watermark information^*Sub-bands. LL (LL)^*The subbands are combined with components LH, HL, HH (where LH, HL, and HH all represent detail information such as image edge texture, as is well known to those skilled in the art) to form a color image C containing watermark information^*(x, y). Resolving the color image C along the inverse of the color image generation^*(x, y) to obtain the attributes of another left _ color and another right _ color element, written into the vector graphics XML file. Accordingly, a vector image with watermark information is successfully obtained.

In one embodiment, the step C1 performs watermark extraction based on the steps a1 and B1, and includes the following specific steps:

setting a color image C to be detected^*' (x, y), original color image C (x, y), and the extracted watermark image is W^*(i, j). Color image C^*' (x, y) and the initial color image C (x, y) are subjected to a discrete wavelet transform to obtain the low frequency sub-band LL, respectively^*' and LL. Then the low frequency sub-band LL^*' and LL are respectively subjected to Slant transformation to obtain a transformation matrix LL _ ST^*' (u, v) and LL _ ST (u, v). Then, two groups of transformation coefficients are arranged in a "" shape, and watermark information is extracted (the transformation coefficient is a transformation matrix LL _ ST)^*' (u, v) and LL _ ST (u, v). Finally, the extracted watermark information is subjected to inverse DCT transformation, and a watermark image W^*(i, j) was successfully obtained.

The algorithm is as follows: pseudo code description of DDS watermark extraction algorithm

Inputting: color image C to be detected^*' (x, y), initial color image C (x, y), embedding intensity α

And (3) outputting: color image W containing watermark information^*(i,j)

1.[LL^*′,LH^*′,HL^*′,HH^*′]＝dwt2(C,^*'haar')% wavelet decomposition

2.[LL,LH,HL,HH]＝dwt2(C,′haar′)

3. Generating a Slant identity matrix S _ n and an inverse identity matrix S _ nT

4.LL_ST^*’＝S_n×LL^*' × S _ nT% Slant transform

5.LL_ST＝S_n×LL×S_nT

6. Character arrangement of low frequency coefficient embedding _ locations ″

7.for location in embedding_locations:

8.W_dct(location)＝(LL_ST^*’(location)-LL_ST(location))/α

9.end for

10.W^*＝idct(W_dct)

11. Outputting a color image W containing watermark information^*

Two-color image digital watermarking method based on LWT and Slant transformation

Wherein, LWT: lifting Wavelet Transform (Lifting Wavelet Transform)

A2: watermark Arnold scrambling;

b2: a watermark embedding process;

c2: and (5) watermark extraction.

In one embodiment, the step a2 performs watermark Arnold scrambling, which includes the following specific steps:

image scrambling is an effective means for confidential storage and secure transmission of digital images, and is often used as an encryption method to preprocess watermark information to improve security. The scrambling algorithm for digital images can be divided into frequency domain scrambling and spatial domain scrambling, where the spatial domain includes a location space and a color space. In order for the original image to be accurately restored, it is necessary to ensure a one-to-one correspondence between the host image and the transformed image. Common deviceThe scrambling method comprises Arnold transformation, Hillbert curve scrambling, Torus self-isomorphic mapping transformation, magic square transformation, Gray code scrambling method and the like. The Arnold transform is used herein. The Arnold transformation, also known as Cat Mapping, encrypts by scrambling the position of each pixel in the image. For an N-order digital image, the Arnold transform is performed using the following formula:

wherein x is_n，y_nRepresenting the pixel value, x, of the image before Arnold transformation_n+1，y_n+1Representing pixel values after Arnold transformation, a and b are parameters, N represents transformation times, mod is modulo operation, and N is image width.

The digital image is recovered by an Arnold inverse transform as follows:

determining a scrambling key^[1]And randomly taking an integer K in the period range, and setting the integer K as the Arnold scrambling frequency. And then separating the three-color basis of the watermark image, and then respectively carrying out K times of Arnold transformation operations on the three-color basis matrix to obtain the three-color components after scrambling. Finally, the three color components after scrambling are combined to obtain a watermark image W after Arnold scrambling^*。

In one embodiment, the step B2 performs watermark embedding on the basis of the step a1, and includes the following specific steps:

let the original image be I, the watermark image be W, and the image containing the watermark be I. Firstly, Arnold encryption preprocessing is carried out on a watermark image W, and then three-color-base separation operation is carried out on the encrypted image to obtain three-color-base components W_R，W_G，W_BRespectively carrying out Slant transformation on the three-color basis components to obtain a transformation coefficient S _ W_R(G,B)(i, j). The original image I is then subjected to a two-level lifting wavelet transform, on LH2 subband (where LH2 represents LH2A frequency domain sub-band) to obtain corresponding components IR, IG and IB, and respectively performing Slant transformation to obtain respective transformation matrix S _ I_R(G,B)(i, j) and then performing a watermark embedding operation. Finally, performing inverse Slant transformation on the three-color basis embedded with the watermark, combining the three-color basis components, and generating an image I containing watermark information through lifting wavelet inverse transformation^*(x,y)。

In one embodiment, the step C2 performs watermark extraction based on the steps a2 and B2, and includes the following specific steps:

firstly, an image I to be detected^*' and the original image I are respectively subjected to two-level lifting wavelet transform. Secondly, separating three color bases of the obtained LH2 components, and respectively carrying out Slant transformation to obtain a transformation matrix S _ I^*′_R(G,B)(I, j) and S _ I_R(G,B)And (i, j) (namely, after Slant transformation is carried out on the three components, two matrixes are obtained). The watermark is then extracted. Finally, the extracted watermark information S _ W is extracted_R(G,B) ^*(i, j) performing an inverse Slant transformation, combining the three color base components, and finally watermarking the image W through Arnold decryption operation^*Is successfully extracted.

Example 2

The invention provides two digital watermarking algorithms, which are concrete network architecture diagrams of the product recommendation method of the invention, and comprise the following steps:

spline image digital watermarking algorithm based on DCT (discrete cosine transform), DWT (discrete wavelet transform) and Slant transform

Step S0101: loading an XML file of the vector image, generating a corresponding DOM tree (DOM is an abbreviation of Document object model), and respectively obtaining attribute values of B, R and G of a left _ color element and a right _ color element;

step S0102: through value taking and quantification, B, R and G attribute values form a two-dimensional array f (x, y);

since the GRB color values are all integer type, (x, y) are discrete coordinates and f is a discrete magnitude, it can be considered as a pixel value of a certain gray scale image, referred to herein as color image C.

Step S0201: performing a DCT transform on the watermark image W (i, j)Generating a transformation matrix W_dctDifferent frequency parts are distributed at different positions of the transformation matrix after being transformed;

step S0202: performing discrete wavelet transform on the color image C (x, y) to obtain a low-frequency sub-band LL;

step S0203: performing Slant transformation on the low-frequency sub-band LL to obtain a transformation matrix LL _ ST (u, v);

step S0204: for the obtained transform coefficients LL _ ST and W_dctSorting according to zigzag, and adding W according to the addition principle as shown in formula (3)_dctIs embedded in a low-frequency position corresponding to LL _ ST (u, v), and generates a watermark-embedded transformation matrix LL _ ST^*(u,v)。

Wherein m is_iRepresenting a watermark image W_dctCoefficient of (a), x_iRepresenting the coefficients of LL _ ST (u, v) after the LL sub-band Slant transform,

indicating after watermark embedding LL _ ST^*The coefficient of (u, v), oc is an embedding intensity factor used to balance visual quality and robustness of the algorithm.

Step S0205: for LL _ ST^*(u, v) performing an inverse Slant transform to obtain a LL with embedded watermark information^*Sub-bands. LL (LL)^*The sub-bands are combined with the LH, HL, HH components to form a color image C containing watermark information^*(x,y)。

Step S0206: along the inverse of color image generation, the color image is parsed to obtain the attributes of the left _ color and right _ color elements, which are written into the vector graphics XML file. Accordingly, a vector image with watermark information is successfully obtained.

Setting a color image C to be detected^*' (x, y), original color image C (x, y), and the extracted watermark image is W^*(i,j)。

Step S0301: color image C^*' (x, y) and C (x, y) are subjected to discrete wavelet transform to respectively obtainObtaining the low-frequency sub-band LL^*' and LL.

Step S0302: low frequency part LL^*' and LL are respectively subjected to Slant transformation to obtain a transformation matrix LL _ ST^*' (u, v) and LL _ ST (u, v).

Step S0303: sorting the two groups of transformation coefficients in the shape of Chinese character 'ji', and extracting watermark information according to formula (4)

Wherein the content of the first and second substances,

indicating an image LL _ ST to be detected^*' (u, v) coefficient, x_iLL _ ST (u, v) coefficient, m 'representing initial color image'_iRepresenting the extracted watermark information.

Step S0304: the extracted watermark information is subjected to inverse DCT transformation, and a watermark image W^*(i, j) was successfully obtained.

Two, double color image digital watermarking algorithm based on LWT and Slant transformation

Step S0101: and determining a scrambling key, randomly taking an integer K in the period range, and setting the integer K as the Arnold scrambling frequency.

Step S0102: and separating the three chromophores of the watermark image. And respectively separating the R, G and B components of the watermark image I to obtain a three-color basis matrix.

Step S0103: and respectively carrying out K times of Arnold transformation operations on the three-color basis matrix, wherein K is a scrambling key, and obtaining the three-color component after scrambling.

Step S0104: and merging the three scrambled color components to obtain the Arnold scrambled watermark image W.

Step S0201: performing Arnold encryption preprocessing on the watermark image W, and then performing three-color-based separation operation on the encrypted image to obtain a component W_R，W_G，W_BRespectively carrying out Slant transformation on the three-color basis components to obtain a transformation coefficient S _ W_R(G,B)(i,j)。

Step S0202: carrying out two on the original image ILevel lifting wavelet transform, pair LH₂The sub-band is subjected to three-color-base separation to obtain a corresponding component I_R，I_G，I_BAnd respectively carrying out Slant transformation to obtain respective transformation matrix S _ I_R(G,B)(i,j)。

Step S0203: and combining the human visual system to adopt different embedding intensities for different color components, and performing watermark embedding operation according to the formula (5).

S_I^* _R(G,B)(i,j)＝S_I_R(G,B)(i,j)+α_R(G,B)×S_W_R(G,B)(i,j) (5)

Wherein, α_R(G,B)Is the embedding strength, which is used to balance the visual quality and the robustness of the algorithm, in this experiment, let α_R＝0.06，α_G＝0.03，α_B＝0.12。

Step S0204: performing inverse Slant transformation on the three-color basis embedded with the watermark, combining the three-color basis components, and generating an image I containing watermark information through lifting wavelet inverse transformation^*(x,y)。

Step S0301: to detect the image I^*' and the original image I are respectively subjected to two-level lifting wavelet transform.

Step S0302: for the obtained LH₂Separating the three color components, and performing Slant transformation to obtain transformation matrix S _ I^*′_R(G,B)(I, j) and S _ I_R(G,B)(i,j)。

Step S0303: extracting watermark according to formula (6)

Wherein, S _ W_R(G,B) ^*(i, j) embedding strength α for the extracted watermark information_R(G,B)The ratio of R, G, B to 2, 1 to 4.

Step S0304: extracting watermark information S _ W_R(G,B) ^*(i, j) performing an inverse Slant transformation, combining the three color base components, and finally watermarking the image W through Arnold decryption operation^*Is successfully extracted.

Example 3

The following examples are presented to demonstrate the feasibility of the embodiments of examples 1 and 2, and are described in detail below:

with the development of internet and communication technology, the informatization of media resources is more and more popularized, people can acquire resources conveniently, and meanwhile, the problem of digital information security becomes a focus of attention, and a digital watermarking technology with copyright protection, authenticity and integrity authentication functions is rapidly developed under the background. On the basis of the digital watermarking theory, the method emphatically researches the digital watermarking algorithm of the transform domain, and mainly works as follows:

(1) through research and analysis of a color spline image of an XML file, a digital watermarking scheme (DDS algorithm) based on the combination of DCT and DWT and Slant transformation is provided. According to the scheme, color information in an XML file is captured by preprocessing a spline image to form a color matrix, and DCT low-frequency information of a watermark image is embedded into Slant low-frequency coefficients by performing Slant transformation on LL subband components of discrete wavelet decomposition. Through experimental simulation, the method can effectively resist noise and object modification attack on the premise of ensuring that the visual quality of the carrier data is not reduced and the transparency of the watermark is met. By analyzing the embedding capacity, the watermark scheme is proved to be capable of ensuring robustness and safety and simultaneously fully balancing the conflict requirements of fidelity and capacity.

(2) Aiming at the characteristic that a spline graph can be stored into a grid graph in a PNG format, the method provides a bicolor digital watermarking algorithm (LS algorithm) based on LWT and Slant transformation. The watermark image Arnold is encrypted by the image scrambling idea, and the embedding intensity of the trichromat is adjusted by combining different perception degrees of HVS to different colors. Simulation experiments show that the method can effectively ensure the invisibility of the watermark and has better robustness to common noise attacks, JPEG compression, scaling and shearing attacks.

A spline image digital watermarking algorithm based on DCT, DWT and Slant transformation is designed aiming at the copyright protection problem of a spline image. In the preprocessing stage, color values are obtained from a spline XML file to form a two-dimensional matrix, so that XML file objects which are difficult to operate are converted into gray images which are easy to operate, and the selectable watermarking methods are more diversified. And considering that the spatial domain watermark embedding capacity is limited and the robustness of resisting attack is poor, watermark information is selected to be embedded in the frequency domain. The embedding and extraction process of the watermark is explained in detail later. In order to verify the performance of the algorithm, different spline color images are tested, and the result shows that the watermarking algorithm meets the requirement of perception invisibility under the condition of no attack. Secondly, on the premise of embedding 628 low-frequency coefficients, the image added with the watermark is attacked, and experimental results show that in the face of noise attack and object modification attack with different strengths,

the watermark image can still be successfully extracted, and the provided watermark algorithm is proved to be good in robustness. Through evaluation and analysis of the embedded capacity, experiments show that the proposed scheme fully balances the conflicting requirements of fidelity and capacity while ensuring robustness and safety. Compared with NC values of other algorithms, the watermarking algorithm provided by the chapter is proved to be good in robustness. In conclusion, the proposed digital watermarking scheme has positive significance for solving the copyright protection problem of the color sample strip.

A significant color image is adopted as a digital watermark in a double-color image digital watermarking algorithm (LS algorithm) based on the combination of lifting wavelet transformation and oblique transformation, and Arnold scrambling preprocessing is carried out on the digital watermark before embedding operation, so that the invisibility and the safety of watermark information are improved. When the watermark is embedded, the original carrier image is firstly subjected to lifting wavelet transformation, then the separated three color bases are respectively subjected to Slant frequency domain transformation, and the embedding intensity is adjusted according to different perception degrees of HVS to different color components, so that the PSNR value of the processed image is higher. In order to verify the invisibility and robustness of the algorithm in the current chapter more intuitively, the experimental result is compared with a color image scrambling digital watermarking algorithm based on DCT transform in Wangtai and Lianwu. The results of the comparison of the two algorithms are shown in table 1.

TABLE 1 comparison of experimental results of the present method with the literature DCT Algorithm

As can be seen from table 1, PSNR of the method is higher than that of the DCT algorithm in the literature, which indicates that the similarity of the carrier image is higher in the method compared with the original image after the watermark is added, thereby proving scientific and effective idea of adjusting the embedding intensity for different color components according to HVS. No attack is added and noises with different intensities are gradually added, and when JPEG compression attacks are performed, the NC value of the method is higher than that of the DCT algorithm of the literature, which shows that the robustness of the method is stronger.

Reference to the literature

[1] All Ruiyang, Gunn Yongjun, Jupith, digital watermarking algorithm based on discrete cosine transform [ J ]. Zheng State university Committee (science edition), 2005(03):63-65.

Those skilled in the art will appreciate that the drawings are only schematic illustrations of preferred embodiments, and the above-described embodiments of the present invention are merely provided for description and do not represent the merits of the embodiments.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (7)

1. A color image digital watermarking method based on transform domain, the method comprising:

obtaining transformation matrix of embedded watermark and executing inverse Slant transformation to obtain LL of embedded watermark information^*Sub-band, LL^*Combining the sub-bands with LH, HL and HH components to form a color image containing watermark information;

analyzing the color image along the reverse process of color image generation to obtain the attributes of left _ color and right _ color elements, and writing the attributes into a vector diagram XML file;

performing discrete wavelet transformation on a color image to be detected and an initial color image to respectively obtain low-frequency sub-bands and performing Slant transformation to obtain a transformation matrix;

sorting the two groups of transformation coefficients in the shape of Chinese character 'ji', extracting watermark information, and carrying out inverse DCT transformation to obtain a watermark image.

2. The method according to claim 1, wherein the obtaining of the transform matrix for embedding the watermark specifically comprises:

loading an XML file of the vector image, generating a corresponding DOM tree, and respectively obtaining attribute values of B, R and G of a left _ color element and a right _ color element; forming a two-dimensional array by taking values and quantizing the attribute values B, R and G;

performing DCT transformation on the watermark image once to generate a transformation matrix, wherein different frequency parts are distributed at different positions of the transformation matrix after transformation; performing discrete wavelet transform on the color image C (x, y) to obtain a low-frequency sub-band LL;

performing Slant transformation on the low-frequency sub-band LL to obtain a transformation matrix LL _ ST (u, v); for the obtained transform coefficients LL _ ST and W_dctSorting W according to zigzag, and adding W according to addition principle_dctIs embedded in the low frequency position of the corresponding LL _ ST (u, v), and a transformation matrix for embedding the watermark is generated.

3. The transform-domain-based color image digital watermarking method according to claim 1, wherein the addition principle is specifically:

wherein m is_iRepresenting a watermark image W_dctCoefficient of (a), x_iRepresenting the coefficients of LL _ ST (u, v) after the LL sub-band Slant transform,

indicating after watermark embedding LL _ ST^*The coefficient of (u, v), oc is an embedding intensity factor.

4. The method according to claim 2, wherein said sorting the two groups of transform coefficients into "zigzag" order to extract watermark information specifically comprises:

5. a color image digital watermarking method based on transform domain, the method comprising:

merging the three-color components after scrambling to obtain a watermark image after Arnold scrambling, performing Arnold encryption and three-color-base separation operations, and performing Slant transformation on the three-color-base components respectively to obtain a transformation coefficient;

performing two-stage lifting wavelet transform on the original image to obtain LH₂Carrying out three-color base separation on the sub-bands and respectively carrying out Slant transformation to obtain respective transformation matrixes;

combining with a human visual system to adopt different embedding intensities for different color components to carry out watermark embedding operation;

performing inverse Slant transformation on the three-color basis embedded with the watermark, combining the three-color basis components, and performing lifting wavelet inverse transformation to generate an image containing watermark information;

respectively carrying out two-stage lifting wavelet transform on the original image to be detected and the obtained LH₂Separating the three color components, and respectively carrying out Slant transformation to obtain transformation matrixes;

extracting the watermark, performing one-time inverse Slant transformation on the extracted watermark information, combining the three-color basis components, and extracting the watermark image through Arnold decryption operation.

6. The method according to claim 5, wherein the embedding strength of different color components is different in combination with human visual system, and the watermark embedding operation is specifically:

S_I^* _R(G,B)(i,j)＝S_I_R(G,B)(i,j)+α_R(G,B)×S_W_R(G,B)(i,j)

wherein, α_R(G,B)Is the embedding strength, S _ W_R(G,B)(I, j) is the transform coefficient, S _ I_R(G,B)(I, j) are respective transformation matrices, S _ I^* _R(G,B)And (i, j) is the transformation matrix after embedding the watermark.

7. The method according to claim 5, wherein the extracting the watermark specifically comprises:

wherein, S _ W_R(G,B) ^*And (i, j) extracted watermark information.

Images