CN110827188B - Quick blind digital watermarking method for color image and extraction method - Google Patents

Abstract

The invention relates to a digital media information security technology, in particular to a quick blind digital watermarking method and an extraction method of a color digital image. The invention uses the digital watermark information to be embedded as binary 0/1 character string, then uses the original digital picture to be embedded as RGB color image, then uses the discrete wavelet transformation level to divide the three-direction components into several blocks, then converts them into RGB watermark picture, and calculates the capacity of watermark. The invention has the advantages that the color image digital watermark adopting the scheme has better invisibility, and meets the invisibility requirement of a watermark algorithm; meanwhile, the digital watermark adopting the scheme has better robustness under the conditions that the color carrier picture is subjected to format conversion, correction, JPEG compression and the like. In addition, the average running time of the algorithm is less than 100 milliseconds, and the actual requirement of quick digital watermarking of the color picture is met.

Description

Quick blind digital watermarking method for color image and extraction method

Technical Field

The invention relates to a digital media information security technology, in particular to a quick blind digital watermarking method and an extraction method of a color digital image.

Background

With the rapid development of network technology, more and more color digital images emerge and spread on the network, and the subsequent illegal actions such as piracy, infringement and the like are frequent, and the copyright protection problem is also more and more widely focused by students at home and abroad. For this reason, on the one hand, the identification requiring copyright protection tends to be attractive, practical, color image digital watermarks with high information content, and is not limited to pseudo-random sequences, binary images or gray images with smaller capacity; on the other hand, with the wide popularization of mobile terminal devices, the existing watermarking algorithm, especially the digital watermarking algorithm adopting the technologies of artificial intelligence, deep learning and the like, is difficult to meet the requirements of quick and efficient application due to the longer running time, and needs to further improve the running speed.

Currently, blind digital watermarking technology using gray images as digital watermarking carriers has been widely studied. However, with the rapid development of science and technology, color images occupy an increasingly important place in today's society. Compared with the gray level image, the color image contains richer information, and has incomparable superiority to the gray level image, whether the color image is used for visual perception of people or subsequent image understanding and analysis. Therefore, the blind digital watermarking technology research of the color image is of more practical significance.

Disclosure of Invention

Aiming at the defects of the prior art, the patent aims to solve the problems.

The invention aims to solve the technical problem of providing a rapid blind digital watermarking method for color images based on discrete wavelet transformation, which has the advantages of simple and rapid watermarking algorithm, high instantaneity, good watermark invisibility, capability of ensuring that the digital watermark has strong robustness, no need of any information of an original color image and an original digital watermark when the digital watermark is extracted at a detection end, and belongs to a blind watermark detection mode.

The invention aims to provide a rapid blind digital watermarking method of a color image based on discrete wavelet transformation, which is characterized by being realized through a specific watermarking embedding process and an extraction process, wherein the watermarking embedding process is described as follows:

the first step:

the digital watermark information to be embedded is a binary 0/1 character string.

And a second step of:

the original digital picture to be embedded with the digital watermark is an RGB color image, which is denoted as F-RGB, the red, green and blue components of F-RGB are denoted as R, G, B, and the size of F-RGB is i×j (unit pixel). F-RGB is converted from RGB color space to YUV color space, color pictures converted to YUV color space are denoted as F-YUV, and luminance component, chrominance component, and density chrominance component of F-YUV are denoted as Y, U, V.

And a third step of:

the level of the discrete wavelet transformation is marked as L, and an image with the size of M (the minimum value in I, J is taken by the unit pixel) is selected on Y to carry out the discrete wavelet transformation, Y is divided into 2 x L mutually non-overlapped sub-blocks, and the size of each sub-block is marked as Y0, wherein the size of each sub-block is M/2^L. The L layer is divided into 4 sub-images, LL sub-image is a low frequency component, LH is a horizontal direction component, HL is a vertical direction component, and HH is a diagonal direction component.

Fourth step:

the components in the directions LH, HL, HH are divided into squares with side lengths S, denoted LH, HL, HH, respectively. Traversing all square blocks with side lengths S from left to right and from top to bottom in lh, hl and hh, randomly selecting a position from a main diagonal intermediate node of each square block, marking a row and column as r and c, and reading wavelet transformation coefficients of the r and c positions in lh, hl and hh as h, v and d. The three coefficient magnitude relations are adjusted to satisfy h < = v < = d. The difference value of d and h is adjusted to be min_diff (the min_diff can be set to be 20.0), the larger the min_diff value is, the larger the invisible effect on the picture is, but the better the robustness is, and the corresponding value can be set by testing according to actual conditions. The vertical direction coefficient v is set to the diagonal coefficient d if the embedded bit character is 1, otherwise it is set to the watermark coefficient h. And rewriting the adjusted h, v and d to wavelet coefficients of the c and r point positions. This process is repeated continuously embedding all characters. The algorithm has good robustness and high running speed.

Fifth step:

the combined Y, U, V component information is combined into YUV and then converted into RGB watermark pictures.

Sixth step:

according to a watermark embedding algorithm, the maximum watermark capacity which can be embedded is recorded as SIZE, and a watermark capacity calculation formula is obtained as follows:

X＝(M–S)/pow(2，L)

SIZE＝X^2/S^2＝(M-S)^2/(4^L*S^2)；

the larger the watermark level L, the smaller the watermark capacity embedded, and the larger the step size S, the smaller the watermark amount embedded. In order to ensure the embedding amount of the watermark and the running speed Lmax to be 3, the value range of the L is [1, 2 and 3]. Since it is necessary to embed binary characters at a randomly selected position in the matrix block of S-2, the minimum value of S is 3. The maximum watermark embedding amount is as follows: (M-3)/(2/(4*9);

since the larger the watermark level L is, the better the watermark robustness is, the larger the S is, the better the watermark embedding security is, and the faster the running speed is.

The watermark extraction process is described as follows:

the first step to the third step: the same as in the first to third steps of the watermark embedding process.

Fourth step:

the components in the directions LH, HL, HH are divided into squares with side lengths S, denoted LH, HL, HH, respectively. Traversing all square blocks with side lengths S from left to right and from top to bottom in lh, hl and hh, randomly selecting a position from a main diagonal intermediate node of each square block, marking a row and column as r and c, and reading wavelet transformation coefficients of the r and c positions in lh, hl and hh as h, v and d. The three coefficient magnitude relations are adjusted to satisfy h < = v < = d. If d-v < v-h is satisfied, the embedded watermark information is character 1, otherwise, the embedded information is character 0. This process is repeated continuously to read all embedded watermark information.

The beneficial effects are that: the color image digital watermark adopting the scheme has better invisibility, and meets the invisibility requirement of a watermark algorithm; meanwhile, the digital watermark adopting the scheme has better robustness under the conditions that the color carrier picture is subjected to format conversion, correction, JPEG compression and the like. In addition, the average running time of the algorithm is less than 100 milliseconds, and the actual requirement of quick digital watermarking of the color picture is met.

Drawings

FIG. 1-1 is a schematic flow chart of a color image fast blind digital watermarking method of the present invention;

FIGS. 1-2 are schematic flow diagrams of the watermark extraction method of the present invention;

fig. 2 is a color vector picture of 512 x 512 24 bits Lena;

FIG. 3 is an example of a watermark picture;

FIG. 4 is a picture of a color carrier with digital watermarking according to the present embodiment;

fig. 5 is a watermark picture of the digital watermark extracted from fig. 4 using the present scheme.

FIG. 6 is a photograph of a color carrier after altering the watermarked photograph of the color carrier;

fig. 7 is a watermark picture extracted from fig. 6;

FIG. 8 is a color carrier picture after format conversion of a digital watermarked carrier color picture;

fig. 9 is a watermark picture extracted from fig. 8;

fig. 10 is a color carrier picture obtained by JPEG lossy compression (compression quality factor selection of 20%) of a watermarked color carrier picture;

fig. 11 is a watermark picture extracted from fig. 10;

Detailed Description

The following is a detailed description of the practice of the invention:

example 1

As shown in the flow of fig. 1-1, taking a picture as an initial processing object of the present embodiment;

the first step:

determining the digital watermark information to be embedded as a binary 0/1 character string;

and a second step of:

the original digital picture to be embedded with the digital watermark is an RGB color image, which is marked as F-RGB, red, green and blue components of the F-RGB are marked as R, G, B, the size of the F-RGB is I.J, and I, J is a pixel; converting F-RGB from RGB color space to YUV color space, recording the color picture converted to YUV color space as F-YUV, and recording the brightness component, chromaticity component and concentration chromaticity component of F-YUV as Y, U, V;

referring to fig. 2, an example of a 24-bit color carrier picture Lena with a resolution of 512×512 is shown.

And a third step of:

the level of the discrete wavelet transformation is marked as L, M is selected from Y, wherein M is a pixel, M is the minimum value in I, J, the discrete wavelet transformation is carried out on the large and small pictures, Y is divided into 2L non-overlapping sub-blocks, and the size of each sub-block is M/2^L and marked as Y0; the L layer is divided into 4 sub-images, the LL sub-image is a low frequency component, the LH is a horizontal direction component, the HL is a vertical direction component, and the HH is a diagonal direction component;

referring to fig. 3, an example of a watermark picture provided by the present scheme is shown. The information to be encrypted is characters a and b, numerals 1 and 2, and ab12 is written into the black picture by white to obtain the watermark picture.

Fourth step:

dividing the components in the directions LH, HL and HH into a plurality of square blocks with side lengths of S, and respectively marking the square blocks as LH, HL and HH; traversing all square blocks with side lengths S from left to right and from top to bottom in lh, hl and hh, randomly selecting a position from a main diagonal intermediate node of each square block, marking rows and columns as r and c, and reading wavelet transformation coefficients of the r and c positions in lh, hl and hh as h, v and d;

adjusting three coefficient size relations to meet h < =v < =d; adjusting the difference value of d and h to be min_diff;

if the embedded bit character is 1, setting the vertical direction coefficient v as a diagonal coefficient d, otherwise setting the vertical direction coefficient v as a watermark coefficient h;

rewriting the adjusted h, v and d to wavelet coefficients of the positions of the c and r points; repeating the process to embed all characters;

fig. 4 shows a color carrier picture after adding a digital watermark by adopting the scheme, and fig. 5 shows a watermark picture of the digital watermark extracted by adopting the scheme.

Comparing fig. 4 and fig. 2, it can be seen that when the digital watermark is hidden in the color carrier picture by adopting the scheme, the digital watermark is well hidden in the carrier picture, and the carrier picture is not changed visually. Comparing fig. 5 and fig. 3, it can be seen that the digital watermark can be extracted very accurately by this scheme.

In order to verify the robustness of the digital watermark of the scheme, the color carrier picture added with the digital watermark is sheared or altered, subjected to format conversion and the like, and then the digital watermark is extracted.

Sixth step:

according to a watermark embedding algorithm, the maximum watermark capacity which can be embedded is recorded as SIZE, and a watermark capacity calculation formula is obtained as follows:

X＝(M–S)/pow(2，L)

SIZE＝X^2/S^2＝(M-S)^2/(4^L*S^2)；

the larger the watermark level L, the smaller the watermark capacity is embedded, and the larger the step S, the smaller the watermark amount is embedded; so the value range of L is [1, 2, 3];

referring to fig. 6, a picture obtained by altering a color carrier picture with a digital watermark is shown, and the altered picture has a part of the data content missing.

Example 2

After the processing of the image processing in embodiment 1 is completed, the watermark image extracted in fig. 6 is required. Referring to fig. 7, in an implementation such as that of fig. 1-2,

the first step:

determining the digital watermark information to be embedded as a binary 0/1 character string;

and a second step of:

the original digital picture to be embedded with the digital watermark is an RGB color image, which is marked as F-RGB, red, green and blue components of the F-RGB are marked as R, G, B, the size of the F-RGB is I.J, and I, J is a pixel; converting F-RGB from RGB color space to YUV color space, recording the color picture converted to YUV color space as F-YUV, and recording the brightness component, chromaticity component and concentration chromaticity component of F-YUV as Y, U, V;

and a third step of:

the level of the discrete wavelet transformation is marked as L, M is selected from Y, wherein M is a pixel, M is the minimum value in I, J, the discrete wavelet transformation is carried out on the large and small pictures, Y is divided into 2L non-overlapping sub-blocks, and the size of each sub-block is M/2^L and marked as Y0; the L layer is divided into 4 sub-images, the LL sub-image is a low frequency component, the LH is a horizontal direction component, the HL is a vertical direction component, and the HH is a diagonal direction component; fourth step:

dividing the components in the directions LH, HL and HH into a plurality of square blocks with side lengths of S, and respectively marking the square blocks as LH, HL and HH; traversing all square blocks with side lengths S from left to right and from top to bottom in lh, hl and hh, randomly selecting a position from a main diagonal intermediate node of each square block, marking rows and columns as r and c, and reading wavelet transformation coefficients of the r and c positions in lh, hl and hh as h, v and d; adjusting three coefficient size relations to meet h < =v < =d;

if d-v < v-h is satisfied, the embedded watermark information is character 1, otherwise, the embedded information is character 0; this process is repeated continuously to read all embedded watermark information.

Referring to fig. 8, a picture is shown after the digital watermark-added color carrier picture has been format-converted in such a way that the picture is converted from JPG to BMP of 256 colors. Fig. 9 shows a watermark picture extracted from fig. 8;

referring to fig. 10, a color carrier picture obtained by JPEG lossy compression (compression quality factor selection of 20%) of a color carrier picture added with a digital watermark is shown in fig. 11, which shows the digital watermark picture extracted from fig. 10.

The algorithm runs ten thousand times on a platform 3.40GHZ CPU,8GB RAM,Centos7,C ++, opencv2.4.9, the average embedding time of the digital watermark is 0.037414 seconds, the average extraction time is 0.015683 seconds, and the total time is 0.053,097 seconds.

In summary, the color image digital watermark adopting the scheme has better invisibility, and meets the invisibility requirement of a watermark algorithm; meanwhile, the digital watermark adopting the scheme has better robustness under the conditions that the color carrier picture is subjected to format conversion, correction, JPEG compression and the like. In addition, the average running time of the algorithm is less than 100 milliseconds, and the actual requirement of quick digital watermarking of the color picture is met.

Claims (2)

1. A color image rapid blind digital watermarking method is characterized in that: the method comprises the following steps:

the first step:

determining the digital watermark information to be embedded as a binary 0/1 character string;

and a second step of:

the original digital picture to be embedded with the digital watermark is an RGB color image, which is marked as F-RGB, red, green and blue components of the F-RGB are marked as R, G, B, the size of the F-RGB is I.J, and I, J is a pixel; converting F-RGB from RGB color space to YUV color space, recording the color picture converted to YUV color space as F-YUV, and recording the brightness component, chromaticity component and concentration chromaticity component of F-YUV as Y, U, V;

and a third step of:

the level of the discrete wavelet transformation is marked as L, M is selected from Y, wherein M is a pixel, M is the minimum value in I, J, the discrete wavelet transformation is carried out on the large and small pictures, Y is divided into 2L non-overlapping sub-blocks, and the size of each sub-block is M/2^L and marked as Y0; the L layer is divided into 4 sub-images, the LL sub-image is a low frequency component, the LH is a horizontal direction component, the HL is a vertical direction component, and the HH is a diagonal direction component;

fourth step:

dividing the components in the directions LH, HL and HH into a plurality of square blocks with side lengths of S, and respectively marking the square blocks as LH, HL and HH; traversing all square blocks with side lengths S from left to right and from top to bottom in lh, hl and hh, randomly selecting a position from a main diagonal intermediate node of each square block, marking rows and columns as r and c, and reading wavelet transformation coefficients of the r and c positions in lh, hl and hh as h, v and d;

adjusting three coefficient size relations to meet h < =v < =d; adjusting the difference value of d and h to be min_diff;

if the embedded bit character is 1, setting the vertical direction coefficient v as a diagonal coefficient d, otherwise setting the vertical direction coefficient v as a watermark coefficient h;

rewriting the adjusted h, v and d to wavelet coefficients of the positions of the c and r points; repeating the process to embed all characters;

fifth step:

combining Y, U, V component information to obtain YUV, and converting to RGB watermark picture;

sixth step:

according to a watermark embedding algorithm, the maximum watermark capacity which can be embedded is recorded as SIZE, and a watermark capacity calculation formula is obtained as follows:

X＝(M–S)/pow(2，L)

SIZE＝X^2/S^2＝(M-S)^2/(4^L*S^2)；

the larger the watermark level L, the smaller the watermark capacity is embedded, and the larger the step S, the smaller the watermark amount is embedded; therefore, the value range of L is [1, 2, 3].

2. The color image fast blind digital watermark extraction method according to claim 1, wherein: the method is characterized in that:

the first step:

determining the digital watermark information to be embedded as a binary 0/1 character string;

and a second step of:

the original digital picture to be embedded with the digital watermark is an RGB color image, which is marked as F-RGB, red, green and blue components of the F-RGB are marked as R, G, B, the size of the F-RGB is I.J, and I, J is a pixel; converting F-RGB from RGB color space to YUV color space, recording the color picture converted to YUV color space as F-YUV, and recording the brightness component, chromaticity component and concentration chromaticity component of F-YUV as Y, U, V;

and a third step of:

the level of the discrete wavelet transformation is marked as L, M is selected from Y, wherein M is a pixel, M is the minimum value in I, J, the discrete wavelet transformation is carried out on the large and small pictures, Y is divided into 2L non-overlapping sub-blocks, and the size of each sub-block is M/2^L and marked as Y0; the L layer is divided into 4 sub-images, the LL sub-image is a low frequency component, the LH is a horizontal direction component, the HL is a vertical direction component, and the HH is a diagonal direction component;

fourth step:

dividing the components in the directions LH, HL and HH into a plurality of square blocks with side lengths of S, and respectively marking the square blocks as LH, HL and HH; traversing all square blocks with side lengths S from left to right and from top to bottom in lh, hl and hh, randomly selecting a position from a main diagonal intermediate node of each square block, marking rows and columns as r and c, and reading wavelet transformation coefficients of the r and c positions in lh, hl and hh as h, v and d; adjusting three coefficient size relations to meet h < =v < =d;

if d-v < v-h is satisfied, the embedded watermark information is character 1, otherwise, the embedded information is character 0; this process is repeated continuously to read all embedded watermark information.

Images