Wavelet transform-based digital watermarking algorithm in YUV domain

Abstract

The application provides a YUV domain digital watermarking algorithm based on wavelet transform, which comprises the following steps: separating an original image to obtain an embedding channel; performing two-stage wavelet transform on the embedding channel; calculating a low-frequency pixel average value LL in a second low-frequency subband; obtaining the watermarking intensity size embedded in the embedding channel according to the corresponding relationship between the LL and the watermarking intensity amplitude; embedding the watermark in a first high-frequency subband to obtain a watermark-carrying second high-frequency subband; performing inverse wavelet transform to obtain a watermark-carrying channel; and merging the channels to obtain a watermark-carrying image. The two-stage wavelet transform only needs 2 rows of buffer zones and addition and subtraction shift operations, and has low cost. Different intensity curves are designed for the YUV three channels according to the visual characteristics of the YUV three channels, and the optimal image quality can be obtained by embedding the watermark in any one channel of the YUV. The load is high, and long information can be hidden. The robustness is high, and the high accuracy can still be obtained under a strong noise transmission channel.

Description

Technical Field

[0001] The invention belongs to the technical field of integrated circuit manufacturing, and in particular relates to a YUV domain digital watermarking algorithm based on wavelet transform. Background Art

[0002] The development of digital technology and networking has enabled the convenient and rapid dissemination of digital information over the Internet. However, while fast and accurate digital transmission provides convenience, it also poses new challenges, such as copyright infringement, piracy, and arbitrary tampering of digital products. Digital watermarking technology is a new direction in information security technology. It is a novel technique that can protect copyright and authenticate the source and integrity of digital products in open network environments. It has broad application prospects in tamper identification, hierarchical data access, data tracking and detection, commercial and video broadcasting, service payment for online digital media, and e-commerce authentication. The wavelet transform, known as the "mathematical microscope," has excellent time-frequency characteristics and consistency with the human visual system. After the inverse wavelet transform, the added watermark information is almost evenly distributed throughout the image. This prevents certain levels of noise and filtering from interfering with the hidden information, thereby greatly improving the robustness to clipping, noise, and filtering.

[0003] YUV is a low-bandwidth color encoding model widely used in image transmission. Based on different sampling methods, YUV has been derived into different formats to meet varying bandwidth requirements. Compared to traditional RAW or RGB domains, YUV watermarks do not undergo image post-processing and are only affected by the transmission channel, making them a more reliable medium. However, traditional watermarking algorithms only embed specific channels, resulting in a low watermark payload and low watermark image quality. These algorithms fail to consider the individual characteristics of the Y, U, and V channels and lack optimization for these characteristics. Summary of the Invention

[0004] The present invention aims to provide a YUV domain digital watermarking algorithm based on wavelet transform. This algorithm utilizes a two-stage wavelet transform, requiring only a two-line buffer and addition, subtraction, and shift operations, resulting in low cost. Different intensity curves are designed for the three YUV channels to address the visual characteristics of YUV. Embedding a watermark in any YUV channel yields superior watermark image quality. This algorithm offers high payload, enabling the concealment of longer messages. The Y channel boasts twice the concealment capacity of the U / V channels, enabling it to handle higher payload requirements. Furthermore, it exhibits high robustness, maintaining high accuracy even in noisy transmission channels.

[0005] The present invention provides a YUV domain digital watermarking algorithm based on wavelet transform, comprising:

[0006] S1. Separate the original image to obtain an embedded channel, where the embedded channel is any one of the three YUV channels obtained by separating the original image; and embed the watermark into only one of the three YUV channels.

[0007] S2. Performing a first-level Haar wavelet transform on the embedded channel to obtain four first-level sub-bands, wherein the four first-level sub-bands include: a first-level low-frequency sub-band, a first-level vertical mid-high frequency sub-band, a first-level horizontal mid-high frequency sub-band, and a first-level high frequency sub-band;

[0008] S3, performing a secondary Haar wavelet transform on the primary high frequency sub-band to obtain four first and second level sub-bands; the four first and second level sub-bands include: first and second level high frequency sub-bands;

[0009] S4. Performing a secondary Haar wavelet transform on the primary low-frequency sub-band to obtain four second-secondary sub-bands; the four second-secondary sub-bands include: a second-secondary low-frequency sub-band;

[0010] S5. Calculate the low-frequency pixel average LL in the second-level low-frequency sub-band, and obtain the watermark intensity alpha of the watermark embedded in the embedded channel according to the corresponding relationship between the low-frequency pixel average LL and the watermark intensity amplitude Z.

[0011] S6. Embed the watermark into the first secondary high frequency sub-band according to the calculated watermark strength to obtain a watermark-carrying secondary high frequency sub-band;

[0012] S7. Perform inverse wavelet transform on all sub-bands including the watermarked secondary high-frequency sub-band to obtain a watermarked channel; merge the watermarked channel with the two channels without watermarks to obtain a watermarked image.

[0013] Furthermore, step S2 specifically includes: performing a 2-dimensional Haar wavelet transform on the embedded channel with a 2*2 pixel block as a unit; A0, A1, A2, and A3 are respectively the pixel values ​​of the four pixels of the 2*2 pixel block; L0, L1, L2, and L3 are respectively the coefficients corresponding to the four primary subbands in sequence; the calculation method is:

[0014] L0=A0+A1+A2+A3; L1=A0+A1−A2−A3;

[0015] L2=A0−A1+A2−A3; L3=A0−A1−A2+A3.

[0016] Furthermore, step S3 specifically includes: performing a one-dimensional Haar wavelet transform on the first-level high-frequency subband with a 1*4 pixel block as a unit; the four first and second-level subbands are: the first and second-level low-frequency subbands, the first and second-level vertical medium and high-frequency subbands, the first and second-level horizontal medium and high-frequency subbands, and the first and second-level high-frequency subbands; B0, B1, B2, and B3 are respectively the pixel values ​​of the four pixels of the 1*4 pixel block in the first-level high-frequency subband; L30, L31, L32, and L33 are respectively the coefficients corresponding to the four first and second-level subbands; the calculation method is:

[0017] L30=B0+B1+B2+B3; L31=B0+B1−B2−B3;

[0018] L32=B0−B1+B2−B3; L33=B0−B1−B2+B3.

[0019] Furthermore, step S4 specifically includes: performing a one-dimensional Haar wavelet transform on the primary low-frequency subband with a 1*4 pixel block as a unit; the four second-level subbands are: the second-level low-frequency subband, the second-level vertical medium-high frequency subband, the second-level horizontal medium-high frequency subband, and the second-level high frequency subband; C0, C1, C2, and C3 are respectively the pixel values ​​of the four pixels of the 1*4 pixel block in the primary low-frequency subband; L00, L01, L02, and L03 are respectively the coefficients corresponding to the four second-level subbands; the calculation method is:

[0020] L00=C0+C1+C2+C3; L01=C0+C1−C2−C3;

[0021] L02=C0−C1+C2−C3; L03=C0−C1−C2+C3.

[0022] Furthermore, the first-level low-frequency subband is m rows * 4n columns of pixels; the second-level low-frequency subband obtained by the Haar wavelet transform is an m-row * n-column coefficient matrix; the L00 represents any coefficient in the coefficient matrix; the actually calculated low-frequency band pixel average value LL = L00 / 16.

[0023] Furthermore, step S5 specifically includes:

[0024] According to experiments or experience, k+1 typical values ​​of the low-frequency pixel average value ranging from small to large are obtained, and k intervals or segments are correspondingly divided;

[0025] Obtaining the watermark intensity amplitudes of the k+1 Y channel embedded watermarks corresponding to the k+1 typical values ​​one by one;

[0026] The watermark intensity amplitudes of k+1 U channels and V channels embedded in the watermark corresponding to the k+1 typical values ​​are obtained, wherein the watermark intensity amplitudes of the U channels and the V channels embedded in the watermark corresponding to the same typical value are equal; and the watermark intensity amplitude of the Y channel corresponding to the same typical value is smaller than the watermark intensity amplitude of the U channel or the V channel.

[0027] Furthermore, in step S5, the watermark intensity amplitude Z is calculated according to the following formula:

[0028] ,

[0029] Where x0-x k represents k+1 typical values; when calculating the Y channel, the y0-y k represents the watermark intensity amplitude of k+1 of the Y channels; when calculating the U channel and the V channel, the y0-y k represents the watermark intensity amplitude of k+1 of the U channel and the V channel; n is any integer number between 1 and k;

[0030] x0<the calculated low frequency band pixel average value LL<x k When LL is calculated based on its value, the interpolation method is used to insert LL into the corresponding interval (x n-1 , x n ); by two points (x n-1 ,y n-1 ) and (x n ,y n ) Determine a straight line, LL is the horizontal coordinate corresponding to the straight line, and the vertical coordinate of the point LL corresponding to the straight line is calculated to be the watermark intensity amplitude Z.

[0031] Furthermore, in step S5, the maximum change in the pixel value of the pixel in the time domain caused by the watermark embedding in the pixel in the embedded channel is set to Δp, and the corresponding maximum watermark intensity value ΔN=Δp*16 in the frequency domain is calculated; and the watermark intensity alpha=Z*ΔN is calculated.

[0032] Furthermore, step S6 specifically includes:

[0033] Divide the first secondary high-frequency sub-band into 2*2 blocks, set the watermark embedding mode corresponding to each watermark bit to P, where P includes a 2*2 matrix composed of -1 / 1 elements, and the sum of all elements is 0; embed the watermark bit by bit into each block in the first secondary high-frequency sub-band to obtain the watermarked secondary high-frequency sub-band; L33 is the coefficient of the first secondary high-frequency sub-band, and L33' is the coefficient of the watermarked secondary high-frequency sub-band;

[0034] L33' = L33 + P * alpha (if Bit=1);

[0035] L33' = L33- P * alpha (if Bit=0).

[0036] Furthermore, step S7 specifically includes:

[0037] S71, performing a one-dimensional inverse wavelet transform: performing a one-dimensional Haar wavelet inverse transform on the first secondary low-frequency sub-band, the first secondary vertical mid-high frequency sub-band, the first secondary horizontal mid-high frequency sub-band, and the watermarked secondary high frequency sub-band to obtain an inverse-transformed first-level high frequency sub-band;

[0038] B0', B1', B2', B3' are respectively the pixel values ​​of the 4 pixels of the 1*4 pixel block in the inverse transformed first-level high frequency sub-band;

[0039] B0'=L30+L31+L32+L33'; B1'=L30+L31−L32−L33';

[0040] B2'=L30−L31+L32−L33'; B3'=L30−L31−L32+L33'.

[0041] Furthermore, step S7 specifically includes:

[0042] S72, performing a two-dimensional inverse wavelet transform: performing a two-dimensional Haar wavelet inverse transform on the primary low-frequency sub-band, the primary vertical mid-high frequency sub-band, the primary horizontal mid-high frequency sub-band, and the inverse-transformed primary high frequency sub-band to obtain the watermark channel;

[0043] A0', A1', A2', and A3' are the pixel values ​​of the four pixels of the 2*2 pixel block in the watermark channel; L3' is the coefficient of the inverse transformed first-level high-frequency subband; L0 is the coefficient of the first-level low-frequency subband, L1 is the coefficient of the first-level vertical mid-high frequency subband; and L2 is the coefficient of the first-level horizontal mid-high frequency subband.

[0044] A0'=L0+L1+L2+L3'; A1'=L0+L1−L2−L3';

[0045] A2'=L0−L1+L2−L3'; A3'=L0−L1−L2+L3'.

[0046] Furthermore, the method further includes: watermark extraction; the watermark extraction includes:

[0047] The YUV channel of the watermark-carrying image is separated to obtain the watermark-carrying channel; the watermark-carrying channel is any one of a watermark-carrying Y channel, a watermark-carrying U channel and a watermark-carrying V channel;

[0048] A primary wavelet transform is performed, and the watermark-carrying channel is subjected to a 2-dimensional Haar wavelet transform in 2*2 pixel blocks to obtain four primary extraction subbands; the four primary extraction subbands include a primary extraction high-frequency subband;

[0049] A secondary wavelet transform is performed, and the primary extraction high-frequency subband is subjected to a 1-dimensional Haar wavelet transform in 1*4 pixel blocks to obtain four secondary extraction subbands; the four secondary extraction subbands include a secondary extraction high-frequency subband;

[0050] The coefficients in the secondary extraction high-frequency subband are blocked to obtain coefficient matrix blocks; the watermark bit values of each coefficient matrix block are calculated and extracted in sequence, and all the extracted watermark bit values Bit are spliced to obtain a final watermark; in each coefficient matrix block, if the result of matrix point multiplication of the coefficient matrix block and the watermark embedding mode P is >0, then Bit = 1; if the result of matrix point multiplication of the coefficient matrix block and the watermark embedding mode P is <0, then Bit = 0.

[0051] Compared with the prior art, the present application has the following beneficial effects:

[0052] The application provides a YUV domain digital watermarking algorithm based on wavelet transform, comprising the following steps: S1, separating an original image to obtain an embedded channel, wherein the embedded channel is any one of YUV three channels obtained by separating the original image; watermark embedding is only performed on one of the YUV three channels; S2, performing one-level haar wavelet transform on the embedded channel to obtain four one-level subbands, wherein the four one-level subbands comprise a one-level low-frequency subband, a one-level vertical middle-high-frequency subband, a one-level horizontal middle-high-frequency subband and a one-level high-frequency subband in sequence; S3, performing two-level haar wavelet transform on the one-level high-frequency subband to obtain four first two-level subbands; the four first two-level subbands comprise a first two-level high-frequency subband; S4, performing two-level haar wavelet transform on the one-level low-frequency subband to obtain four second two-level subbands; the four second two-level subbands comprise a second two-level low-frequency subband; S5, calculating a low-frequency segment pixel average value LL in the second two-level low-frequency subband, and obtaining a watermark intensity size alpha of the embedded channel embedding the watermark according to a corresponding relationship between the low-frequency segment pixel average value LL and a watermark intensity amplitude Z; S6, embedding the watermark into the first two-level high-frequency subband according to the calculated watermark intensity size to obtain a watermark-carrying two-level high-frequency subband; S7, performing wavelet inverse transform on all subbands containing the watermark-carrying two-level high-frequency subband to obtain a watermark-carrying channel; and combining the watermark-carrying channel and two channels without carrying the watermark to obtain a watermark-carrying image.

[0053] The application adopts two-level wavelet transform, only needs 2 row buffer zones and addition and subtraction shift operations, and has low cost. Different intensity curves are designed for YUV three channels according to YUV visual characteristics, and optimal image quality can be obtained by embedding the watermark in any one of the YUV channels. The load is high, long information can be hidden, and the hiding capacity of the U / V channel is twice that of the Y channel, so that higher load demand can be met. The robustness is high, and high accuracy can still be obtained under a strong noise transmission channel. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A YUV domain digital watermarking algorithm flowchart based on wavelet transform is shown in the figure.

[0055] Figure 2 A one-level wavelet transform diagram in the algorithm is shown in the figure.

[0056] Figure 3 A two-level wavelet transform diagram of the one-level high-frequency subband HH1 in the algorithm is shown in the figure.

[0057] Figure 4 A two-level wavelet transform diagram of the one-level low-frequency subband LL1 in the algorithm is shown in the figure.

[0058] Figure 5This is the watermark strength table in the algorithm of the present invention.

[0059] Figure 6 This is the curve of the watermark intensity amplitude Z with respect to the low-frequency pixel average value LL in the algorithm of the present invention.

[0060] Figure 7 Schematic diagram of embedding a watermark into the first and second-level high frequency sub-bands HH2a in the algorithm of the present invention to obtain the watermarked second-level high frequency sub-bands HH2a'.

[0061] Figure 8 Schematic diagram of 1-dimensional wavelet inverse transform in the algorithm of the present invention.

[0062] Figure 9 Schematic diagram of 2D wavelet inverse transform in the algorithm of the present invention.

[0063] Figure 10 Schematic diagram of the first-level wavelet transform of the watermark channel C' in the algorithm of the present invention.

[0064] Figure 11 Schematic diagram of the second-level wavelet transform of the first-level extracted high-frequency subband HH1t in the algorithm of the present invention.

[0065] Figure 12 Schematic diagram of watermark extraction accuracy under lossless transmission channel.

[0066] Figure 13 Schematic diagram of watermark extraction accuracy under strong noise channel transmission.

[0067] Figure 14 This is a first schematic diagram of the image quality of the original image, the watermarked Y channel, the watermarked U channel, and the watermarked V channel.

[0068] Figure 15 This is a second schematic diagram of the image quality of the original image, the watermarked Y channel, the watermarked U channel, and the watermarked V channel. DETAILED DESCRIPTION

[0069] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description. It should be noted that the drawings are in a very simplified form and are not to exact scale, and are only used for the purpose of conveniently and clearly illustrating the embodiments of the present invention.

[0070] For ease of description, some embodiments of the present application may use spatially relative terms such as "above," "below," "top," and "below" to describe the relationship between one element or component and another (or other) elements or components as shown in the drawings of the embodiments. It should be understood that in addition to the orientations described in the drawings, spatially relative terms are also intended to include different orientations of the device during use or operation. For example, if the device in the drawings is turned over, the elements or components described as being "below" or "beneath" other elements or components will subsequently be positioned as being "above" or "above" the other elements or components. The terms "first," "second," and the like below are used to distinguish between similar elements and are not necessarily used to describe a specific order or chronological sequence. It should be understood that these terms used in this manner are interchangeable where appropriate.

[0071] The embodiment of the present invention provides a YUV domain digital watermarking algorithm based on wavelet transform, such as Figure 1 Shown, including:

[0072] S1. Separate the original image to obtain an embedded channel, where the embedded channel is any one of the three YUV channels obtained by separating the original image; and embed the watermark into only one of the three YUV channels.

[0073] S2, the embedded channel is subjected to a first-level Haar wavelet transform to obtain four first-level sub-bands, which include: a first-level low-frequency sub-band, a first-level vertical mid-high frequency sub-band, a first-level horizontal mid-high frequency sub-band, and a first-level high frequency sub-band;

[0074] S3, performing a second-level Haar wavelet transform on the first-level high-frequency sub-band to obtain four first and second-level sub-bands; the four first and second-level sub-bands include: first and second-level high-frequency sub-bands;

[0075] S4, performing a secondary Haar wavelet transform on the primary low-frequency sub-band to obtain four secondary secondary sub-bands; the four secondary secondary sub-bands include: a secondary secondary low-frequency sub-band;

[0076] S5. Calculate the low-frequency pixel average value LL in the second-level low-frequency sub-band, and obtain the watermark intensity alpha of the watermark embedded in the embedded channel according to the corresponding relationship between the low-frequency pixel average value LL and the watermark intensity amplitude Z;

[0077] S6. Embed the watermark into the first and second high frequency sub-bands according to the calculated watermark strength to obtain the watermark-carrying second high frequency sub-band;

[0078] S7. Perform inverse wavelet transform on all sub-bands including the watermarked secondary high-frequency sub-band to obtain a watermarked channel; merge the watermarked channel with the two channels without watermarks to obtain a watermarked image.

[0079] The following describes in detail the steps of the YUV domain digital watermarking algorithm based on wavelet transform according to an embodiment of the present invention with reference to the accompanying drawings.

[0080] Step S1: Separate the original image to obtain a channel C to be embedded. Channel C is any one of the three YUV channels obtained by separating the original image. The watermark is embedded in only one of the three YUV channels. During the entire watermark embedding operation, only one of the three YUV channels is embedded. Embedding any two or all three channels with a watermark will affect image quality.

[0081] Original images can be in RAW or RGB format. Images can be represented using different color spaces, the most common of which are RGB (Red, Green, Blue) and YUV (Luminance-Chroma) color spaces. RGB color space is one of the most common color representations. In RGB color space, the color of an image is composed of the intensities of three channels: red (R), green (G), and blue (B). The value of each channel typically ranges from 0 to 255, representing the brightness of that color channel. YUV color space is a color space used to represent the brightness and chromaticity of an image. The luma channel (Y) represents the brightness of the image, while the chromaticity channels (U and V) represent color differences. Because the human eye is more sensitive to changes in brightness and relatively insensitive to changes in chromaticity, by converting the image to YUV format, the luma (Y) and chromaticity (U, V) components are clearly separated, making it easier to extract image features.

[0082] Convert the original image's RGB format to YUV format according to the calculation formula, and separate the Y channel, U channel, and V channel. The formula is as follows:

[0083] Y=0.299×R+0.587×G+0.114×B;

[0084] U=0.564×(BY); V=0.713×(RY).

[0085] Among them, Y represents the luminance channel, U and V both represent the chrominance channels, and R, G, and B represent the red, green, and blue of the image, respectively. By calculating through the above formula, the YUV three-channel can be obtained. The present invention is based on a mathematical model. The Y component has the largest weight in the final mapping to the RGB domain, and the U / V components have similar weights. The present invention is based on a physical model. The human eye is more sensitive to the luminance Y component, and the chrominance U / V components do not significantly affect the visual effect.

[0086] Step S2: Figure 2As shown, a first-level wavelet transform is performed on the embedded channel C, and a first-level Haar wavelet transform is performed to obtain four first-level sub-bands. The four first-level sub-bands include: a first-level low-frequency sub-band LL1, a first-level vertical mid-high frequency sub-band HL1, a first-level horizontal mid-high frequency sub-band LH1, and a first-level high frequency sub-band HH1. The wavelet transform can decompose the image into multiple resolutions. Each level of wavelet transform obtains four sub-bands, namely the low-frequency sub-band LL, the vertical mid-high frequency sub-band HL, the horizontal mid-high frequency sub-band LH, and the high frequency sub-band HH. The Haar wavelet transform is a signal processing method based on wavelet analysis, which can decompose the signal into multiple sub-signals of different frequencies.

[0087] Specifically, the embedded channel C is subjected to a 2D Haar wavelet transform with 2*2 pixel blocks as units to obtain four primary subbands. For example, the embedded channel C is a 64*64 (64 rows by 64 columns) pixel matrix, and the coefficient matrices of the four primary subbands are each 32*32. The calculation method is as follows:

[0088] A0, A1, A2, and A3 are the pixel values ​​of the four pixels in the 2*2 pixel block embedded in channel C. L0 is the coefficient of the first-level low-frequency subband LL1. L0 can be understood as the output value (also called coefficient) of the first-level low-frequency subband LL1 obtained by performing a 2D Haar wavelet transform on the embedded channel C. All these coefficients constitute the first-level low-frequency subband LL1. L1 is the coefficient of the first-level vertical mid-high frequency subband HL1; L2 is the coefficient of the first-level horizontal mid-high frequency subband LH1; and L3 is the coefficient of the first-level high frequency subband HH1.

[0089] L0=A0+A1+A2+A3; L1=A0+A1−A2−A3;

[0090] L2=A0−A1+A2−A3; L3=A0−A1−A2+A3.

[0091] Step S3: Figure 3 As shown, a two-level wavelet transform is performed on the first-level high-frequency subband HH1 using 1*4 pixel blocks as units to obtain four first- and second-level subbands: the first- and second-level low-frequency subband LL2a, the first- and second-level vertical mid- and high-frequency subband HL2a, the first- and second-level horizontal mid- and high-frequency subband LH2a, and the first- and second-level high-frequency subband HH2a. The calculation method is as follows:

[0092] B0, B1, B2, and B3 are the pixel values ​​of the 4 pixels in the 1*4 pixel block in the first-level high-frequency subband HH1; L30 is the coefficient of the first and second-level low-frequency subband LL2a; L31 is the coefficient of the first and second-level mid- and high-frequency subband HL2a in the vertical direction; L32 is the coefficient of the first and second-level mid- and high-frequency subband LH2a in the horizontal direction; L33 is the coefficient of the first and second-level high-frequency subband HH2a.

[0093] L30=B0+B1+B2+B3; L31=B0+B1−B2−B3;

[0094] L32=B0−B1+B2−B3; L33=B0−B1−B2+B3.

[0095] This invention uses a two-stage wavelet transform: the first stage uses a two-dimensional transform (2x2 blocks), and the second stage uses a one-dimensional transform (1x4 blocks). The two-dimensional and one-dimensional wavelet transforms require only two buffer rows and addition, subtraction, and shift operations, resulting in low cost and highly robust watermarking.

[0096] Step S4: Figure 4 As shown, a two-level wavelet transform is performed, and the first-level low-frequency subband LL1 is subjected to a 1-dimensional Haar wavelet transform with a 1*4 pixel block as a unit to obtain four second-level subbands. Exemplarily, the first-level low-frequency subband LL1 is, for example, a 32*32 (32 rows multiplied by 32 columns) pixel matrix, and the four second-level subbands are each a 32*8 coefficient matrix. The four second-level subbands are the second-level low-frequency subband LL2b, the second-level vertical mid-high frequency subband HL2b, the second-level horizontal mid-high frequency subband LH2b, and the second-level high frequency subband HH2b. The calculation method is as follows:

[0097] C0, C1, C2, and C3 are the pixel values ​​of the 4 pixels in the 1*4 pixel block in the first-level low-frequency sub-band LL1; L00 is the coefficient of the second-level low-frequency sub-band LL2b; L01 is the coefficient of the second-level middle and high-frequency sub-band HL2b in the vertical direction; L02 is the coefficient of the second-level middle and high-frequency sub-band LH2b in the horizontal direction; L03 is the coefficient of the second-level high-frequency sub-band HH2b.

[0098] L00=C0+C1+C2+C3; L01=C0+C1−C2−C3;

[0099] L02=C0−C1+C2−C3; L03=C0−C1−C2+C3.

[0100] S5. Calculate the low-frequency pixel average LL in the second-level low-frequency sub-band LL2b. Based on the correspondence between the low-frequency pixel average LL and the watermark intensity amplitude Z, obtain the watermark intensity alpha embedded in each of the three YUV channels. Specifically, calculate the low-frequency pixel average LL: LL = L00 / 16. Since L00 = C0 + C1 + C2 + C3, that is, L00 is the superposition of the four pixels in the first-level low-frequency sub-band LL1, and each pixel in the first-level low-frequency sub-band LL1 is the superposition of the four pixels A0, A1, A2, and A3 embedded in channel C, L00 is equivalent to including the 16 pixels embedded in channel C. L00 in the frequency domain divided by 16 is the low-frequency pixel average LL in the time domain. In this example, the parameter value in the frequency domain is divided by 16 to obtain the corresponding parameter value in the time domain, and the parameter value in the time domain is multiplied by 16 to obtain the corresponding parameter value in the frequency domain.

[0101] Figure 5 is the watermark strength table; Figure 6 It is the curve of watermark intensity amplitude Z with respect to the low frequency band pixel average value LL. Figure 5 The watermark strength table in the figure shows typical values ​​obtained based on a large number of experiments. Figure 5 The specific values ​​in the table are not limited and can be adjusted according to actual conditions. Figure 5 The first column shows k+1 (for example, k+1 is 8) typical low-frequency pixel average values ​​LL (x0-x7); the value range of the low-frequency pixel average value LL is: 0~255. Figure 5 The second column shows the watermark intensity amplitudes Z of the 8 Y channel embedded watermarks corresponding to the 8 typical low-frequency pixel average values ​​LL. Figure 5 The third column shows the watermark intensity amplitudes Z of the eight U and V channels corresponding to the eight typical low-frequency pixel average values ​​LL. The watermark intensity amplitudes Z of the U and V channels corresponding to the same low-frequency pixel average value LL are equal. The watermark intensity amplitude of the Y channel corresponding to the same typical value is smaller than that of the U or V channel.

[0102] The calculation formula of the watermark intensity amplitude Z with respect to the low-frequency pixel average value LL is as follows:

[0103] ,

[0104] It can be seen from the above formula that when the calculated low-frequency pixel average value LL ≤ x0 (x0 is 4, for example), the corresponding watermark intensity amplitude of the Y channel embedded watermark is a constant value y0 (y0 is 0.16, for example), and the corresponding watermark intensity amplitude of the U channel and V channel embedded watermark is a constant value y0 (y0 is 0.33, for example).

[0105] When x0<the calculated low-frequency pixel average value LL<x7, the watermark intensity amplitude Z embedded in the watermark can be understood as a short line segment with a certain slope. For example, according to the size of the low-frequency pixel average value LL calculated by LL= L00 / 16, LL is interpolated into a corresponding interval (x n-1 , x n ); For example, if the calculated LL is 25, 25 is inserted into the interval (x1, x2), x1 is 12, and x2 is 28. n-1 ,y n-1 ) and (x n ,y n ) to determine a straight line, for example, two points (x1, y1) and (x2, y2) determine the straight line of the example, and the vertical coordinate of the LL point corresponding to the straight line is calculated as the watermark intensity amplitude Z. For the Y channel, take Figure 5 y in the second column n-1 and y n ; For U channel and V channel, take Figure 5 y in the third column n-1 and y n For the Y channel, the first vertical coordinate corresponding to the horizontal coordinate of 25 on the first straight line determined by the points (12, 0.22) and (28, 0.30) is calculated. The first vertical coordinate is the watermark intensity amplitude Z of the Y channel embedded watermark corresponding to the low-frequency pixel average LL, for example, 25. For the U / V channel, the second vertical coordinate corresponding to the horizontal coordinate of 25 on the second straight line determined by the points (12, 0.42) and (28, 0.48) is calculated. The second vertical coordinate is the watermark intensity amplitude Z of the U / V channel embedded watermark corresponding to the low-frequency pixel average LL, for example, 25.

[0106] Low frequency pixel average LL≥x k (x k For example, when the value is 188, the watermark intensity amplitude of the corresponding Y channel embedded watermark is a constant value y k (y k For example, 0.63), the corresponding U channel and V channel embedded watermark intensity amplitude is a constant value y k (y k For example, 0.99). The nonlinear watermark strength model of the present invention is applicable to the three channels of Y / U / V.

[0107] Figure 6It can be seen that the watermark intensity amplitude of the Y channel embedded watermark corresponding to the same low-frequency pixel average value LL is smaller than the watermark intensity amplitude of the U channel and V channel embedded watermarks. The watermark intensity amplitude of the Y channel embedded watermark is smaller, that is, the Y channel embedding adopts a strong constraint watermark intensity model, the Y channel embedded watermark is weak and not easy to be seen by the human eye. The watermark intensity amplitude of the U channel and V channel embedded watermark is larger, that is, the U / V channel embedding adopts a wide constraint watermark intensity model, and the U / V channel embedded watermark is relatively strong.

[0108] According to the low frequency pixel average LL Figure 5 The intervals and calculation formulas in the watermark intensity table are interpolated to calculate the ordinate, which is the watermark intensity amplitude Z. Assuming the maximum change in the pixel value of a pixel in the time domain caused by watermark embedding is Δp, calculate the corresponding maximum watermark intensity value in the frequency domain ΔN = Δp * 16; calculate the watermark intensity alpha = Z * ΔN.

[0109] Step S6: Figure 7 As shown, based on the calculated watermark strength, the watermark is embedded into the first and second high-frequency sub-bands HH2a, resulting in the watermarked second high-frequency sub-band HH2a'. Specifically, the watermark embedding pattern corresponding to each watermark bit is set to P. P can be any sequence of -1, 1, with the number of -1 and 1 being equal. P can also be a 2*2 matrix composed of elements of -1 / 1, where the sum of all elements is 0. An example of P is the following 2*2 matrix:

[0110] .

[0111] For example, the first and second high frequency sub-bands HH2a are divided into 2*2 blocks, and the watermark is embedded bit by bit into each block of the first and second high frequency sub-bands HH2a. L33' is the coefficient of the watermarked second high frequency sub-band HH2a'. The watermark information is, for example, a combination of binary bits 0 and 1. In each block,

[0112] L33' = L33 + P * alpha (if Bit=1);

[0113] L33' = L33 - P * alpha (if Bit=0).

[0114] Since the human eye has different perceptions of changes in high and low frequency components, it usually has a high perception of low frequencies and relatively weak perception of high frequency details. Therefore, the watermark is embedded in the high frequency sub-band to ensure the robustness and concealment of the steganographic algorithm.

[0115] S7, perform inverse wavelet transform on all sub-bands containing the watermarked secondary high-frequency sub-band HH2a' to obtain the watermarked channel; merge the three YUV channels of the watermarked image to obtain the watermarked image. Figure 8 As shown, S71, perform a one-dimensional inverse wavelet transform: perform a one-dimensional Haar wavelet inverse transform on the first and second low-frequency sub-bands LL2a, the first and second vertical medium and high-frequency sub-bands HL2a, the first and second horizontal medium and high-frequency sub-bands LH2a, and the watermarked second high-frequency sub-band HH2a' obtained by the wavelet transform of the first high-frequency sub-band HH1 to obtain the inverse transformed first high-frequency sub-band HH1'.

[0116] B0', B1', B2', B3' are the pixel values ​​of the 4 pixels of the 1*4 pixel block in the inverse transformed primary high frequency sub-band HH1'; L33' is the coefficient of the watermarked secondary high frequency sub-band HH2a'.

[0117] B0'=L30+L31+L32+L33'; B1'=L30+L31−L32−L33';

[0118] B2'=L30−L31+L32−L33'; B3'=L30−L31−L32+L33'.

[0119] like Figure 9 As shown, S72, perform a 2D inverse wavelet transform: perform a 2D inverse Haar wavelet transform on the first-level low-frequency sub-band LL1, the first-level vertical mid-high frequency sub-band HL1, the first-level horizontal mid-high frequency sub-band LH1 and the inverse-transformed first-level high frequency sub-band HH1' to obtain the watermark channel C', which is calculated as follows:

[0120] A0', A1', A2', A3' are the pixel values ​​of the 4 pixels in the 2*2 pixel block in the watermark channel C' respectively; L3' is the coefficient of the inverse transform of the first-level high-frequency subband HH1'.

[0121] A0'=L0+L1+L2+L3'; A1'=L0+L1−L2−L3';

[0122] A2'=L0−L1+L2−L3'; A3'=L0−L1−L2+L3'.

[0123] The above method can be used to obtain a watermark channel. During the entire watermark embedding operation, only one of the three YUV channels is embedded. The specific channel to be embedded is set according to actual needs. The watermark channel and the two unwatermarked channels are merged to obtain a watermarked image. Specifically, the watermark channel and the two unwatermarked channels can be converted to RGB channels and then merged to obtain the watermarked image.

[0124] The YUV domain digital watermark algorithm based on wavelet transform of the present invention also includes: watermark extraction; watermark extraction includes:

[0125] like Figure 10As shown, the YUV channels of the watermarked image are separated to obtain the watermarked channel C'; the watermarked channel C' is any one of the watermarked Y channel, the watermarked U channel and the watermarked V channel.

[0126] Perform a first-level wavelet transform and perform a 2D Haar wavelet transform on the watermark channel C' in units of 2*2 pixel blocks to obtain four first-level extraction subbands. The four first-level extraction subbands are the first-level extraction low-frequency subband LL1t, the first-level extraction vertical mid-high frequency subband HL1t, the first-level extraction horizontal mid-high frequency subband LH1t, and the first-level extraction high frequency subband HH1t. The calculation method is as follows:

[0127] A0', A1', A2', and A3' are the pixel values ​​of the four pixels in the 2*2 pixel block in the watermark channel C' respectively; T0 is the coefficient of the first-level extraction of the low-frequency subband LL1t, T1 is the coefficient of the first-level extraction of the middle and high-frequency subband HL1t in the vertical direction; T2 is the coefficient of the first-level extraction of the middle and high-frequency subband LH1t in the horizontal direction; T3 is the coefficient of the first-level extraction of the high-frequency subband HH1t.

[0128] T0=A0'+A1'+A2'+A3'; T1=A0'+A1'−A2'−A3';

[0129] T2=A0'−A1'+A2'−A3'; T3=A0'−A1'−A2'+A3'.

[0130] like Figure 11 As shown, a two-level wavelet transform is performed, and the first-level high-frequency subband HH1t is subjected to a one-dimensional Haar wavelet transform with 1*4 pixel blocks as units to obtain four second-level extracted subbands; namely, the second-level extracted low-frequency subband LL2c, the second-level extracted vertical mid-high frequency subband HL2c, the second-level extracted horizontal mid-high frequency subband LH2c, and the second-level extracted high frequency subband HH2c. The calculation method is as follows:

[0131] D0, D1, D2, and D3 are the pixel values ​​of the 4 pixels in the 1*4 pixel block in the high-frequency subband HH1t extracted in the first level; T30 is the coefficient of the low-frequency subband TT2c extracted in the second level; T31 is the coefficient of the middle and high-frequency subband HT2c extracted in the vertical direction in the second level; T32 is the coefficient of the middle and high-frequency subband TH2c extracted in the horizontal direction in the second level; T33 is the coefficient of the high-frequency subband HH2c extracted in the second level.

[0132] T30=D0+D1+D2+D3; T31=D0+D1−D2−D3;

[0133] T32=D0−D1+D2−D3; T33=D0−D1−D2+D3.

[0134] Divide the coefficients in the secondary extracted high-frequency subband HH2c into blocks to obtain coefficient matrix blocks; for example, divide them into 2*2 blocks, each block being a 2*2 coefficient matrix. Calculate and extract the watermark bit values ​​corresponding to each coefficient matrix block in sequence, and concatenate all the extracted watermark bit values ​​to obtain the final watermark. In each coefficient matrix block, if the result of the matrix dot product operation between the coefficient matrix block and the watermark embedding mode P is greater than 0, then Bit = 1; if the result of the matrix dot product operation between the coefficient matrix block and the watermark embedding mode P is less than 0, then Bit = 0.

[0135] Figure 12 Schematic diagram of watermark extraction accuracy under lossless transmission channel. Figure 12 As shown in the figure, under lossless transmission channels, the bit error rate (BER) of the three YUV channel watermarks is 0. The lower the BER, the higher the accuracy of the extracted watermark information. Under lossless transmission channels, the accuracy of the three YUV channel watermark extraction reaches 100%.

[0136] Figure 13 This is a diagram showing the accuracy of watermark extraction under strong noise channel transmission. Figure 13 As shown in the figure, under strong noise (SNR < 20dB) channel transmission, the accuracy of watermark extraction in the U / V channel can reach 100%, and the accuracy of the Y channel is close to 100%. This method has good robustness to different YUV data transmission conditions.

[0137] Figure 14 This is a first schematic diagram of the image quality of the original image, the watermarked Y channel, the watermarked U channel, and the watermarked V channel. Figure 15 The second schematic diagram of the image quality of the original image, the watermarked Y channel, the watermarked U channel and the watermarked V channel. Figure 14 and Figure 15 As shown, there is no visual difference between the images in the watermarked Y channel, U channel, and V channel, and the image quality is high. The peak signal-to-noise ratio (PSNR) of the watermark images embedded in different channels is higher than 38dB. The Y / U / V channels all maintain high image quality (PSNR>38dB indicates high quality).

[0138] In summary, the application provides a YUV domain digital watermarking algorithm based on wavelet transform, comprising: S1, separating an original image to obtain an embedded channel, the embedded channel being any one of YUV three channels obtained by separating the original image; watermark embedding is only embedded in one of the YUV three channels; S2, performing one-level haar wavelet transform on the embedded channel to obtain four one-level subbands, the four one-level subbands comprising, in sequence, a one-level low-frequency subband, a one-level vertical middle-high-frequency subband, a one-level horizontal middle-high-frequency subband and a one-level high-frequency subband; S3, performing two-level haar wavelet transform on the one-level high-frequency subband to obtain four first two-level subbands; the four first two-level subbands comprising a first two-level high-frequency subband; S4, performing two-level haar wavelet transform on the one-level low-frequency subband to obtain four second two-level subbands; the four second two-level subbands comprising a second two-level low-frequency subband; S5, calculating a low-frequency segment pixel average value LL in the second two-level low-frequency subband, and obtaining a watermark intensity size alpha of the embedded channel embedding the watermark according to a corresponding relationship between the low-frequency segment pixel average value LL and a watermark intensity amplitude Z; S6, embedding the watermark into the first two-level high-frequency subband according to the calculated watermark intensity size, to obtain a watermark-carrying two-level high-frequency subband; S7, performing wavelet inverse transform on all subbands containing the watermark-carrying two-level high-frequency subband to obtain a watermark-carrying channel; and combining the watermark-carrying channel and two channels without carrying the watermark to obtain a watermark-carrying image.

[0139] The application adopts two-level wavelet transform, only needs 2 row buffer and addition and subtraction shift operation, and has low cost. Different intensity curves are designed for YUV three channels according to YUV visual characteristics, and optimal image quality can be obtained by embedding the watermark in any one of the YUV channels. The load is high, can hide longer information, and has twice hiding capacity of U / V channels in Y channel, and can cope with higher load demand. The robustness is high, and still has higher accuracy rate under a strong noise transmission channel.

[0140] The embodiments in the specification are described in a progressive manner, and each embodiment focuses on the difference from other embodiments. The same or similar parts of each embodiment can be referred to each other.

[0141] The above description is only a description of the preferred embodiments of the application, and does not limit the scope of the application. Any person skilled in the art can make possible changes and modifications to the technical solutions of the application without departing from the spirit and scope of the application. Therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the application, without departing from the technical solutions of the application, are within the protection scope of the application.

Claims

1. A YUV domain digital watermarking algorithm based on wavelet transform, characterized in that: include: S1. Separate the original image to obtain an embedded channel, where the embedded channel is any one of the three YUV channels obtained by separating the original image; The watermark is embedded in only one of the three YUV channels; S2. Performing a first-level Haar wavelet transform on the embedded channel to obtain four first-level sub-bands, wherein the four first-level sub-bands include: a first-level low-frequency sub-band, a first-level vertical mid-high frequency sub-band, a first-level horizontal mid-high frequency sub-band, and a first-level high frequency sub-band; S3, performing a secondary Haar wavelet transform on the primary high frequency sub-band to obtain four first and second level sub-bands; the four first and second level sub-bands include: first and second level high frequency sub-bands; S4. Performing a secondary Haar wavelet transform on the primary low-frequency sub-band to obtain four second-secondary sub-bands; the four second-secondary sub-bands include: a second-secondary low-frequency sub-band; S5. Calculate the low-frequency pixel average LL in the second-level low-frequency sub-band, and obtain the watermark intensity alpha of the watermark embedded in the embedded channel according to the corresponding relationship between the low-frequency pixel average LL and the watermark intensity amplitude Z. S6. Embed the watermark into the first secondary high frequency sub-band according to the calculated watermark strength to obtain a watermark-bearing secondary high frequency sub-band; S7. Perform inverse wavelet transform on all sub-bands including the watermarked secondary high-frequency sub-band to obtain a watermarked channel; merge the watermarked channel with the two channels without watermarks to obtain a watermarked image.

2. The YUV domain digital watermarking algorithm based on wavelet transform according to claim 1, characterized in that: Step S2 specifically includes: performing a 2D Haar wavelet transform on the embedded channel with a 2*2 pixel block as a unit; A0, A1, A2, and A3 are the pixel values ​​of the four pixels in the 2*2 pixel block; L0, L1, L2, and L3 are the coefficients corresponding to the four primary subbands in sequence; the calculation method is: L0=A0+A1+A2+A3; L1=A0+A1−A2−A3; L2=A0−A1+A2−A3; L3=A0−A1−A2+A3.

3. The YUV domain digital watermarking algorithm based on wavelet transform according to claim 1, characterized in that: Step S3 specifically includes: performing a one-dimensional Haar wavelet transform on the first-level high-frequency subband with a 1*4 pixel block as a unit; the four first and second-level subbands are: the first and second-level low-frequency subbands, the first and second-level vertical medium and high-frequency subbands, the first and second-level horizontal medium and high-frequency subbands, and the first and second-level high-frequency subbands; B0, B1, B2, and B3 are respectively the pixel values ​​of the four pixels of the 1*4 pixel block in the first-level high-frequency subband; L30, L31, L32, and L33 are respectively the coefficients corresponding to the four first and second-level subbands; the calculation method is: L30=B0+B1+B2+B3; L31=B0+B1−B2−B3; L32=B0−B1+B2−B3; L33=B0−B1−B2+B3.

4. The YUV domain digital watermarking algorithm based on wavelet transform according to claim 1, characterized in that: Step S4 specifically includes: performing a one-dimensional Haar wavelet transform on the primary low-frequency subband with a 1*4 pixel block as a unit; the four second-level subbands are: the second-level low-frequency subband, the second-level vertical medium-high frequency subband, the second-level horizontal medium-high frequency subband, and the second-level high frequency subband; C0, C1, C2, and C3 are respectively the pixel values ​​of the four pixels of the 1*4 pixel block in the primary low-frequency subband; L00, L01, L02, and L03 are respectively the coefficients corresponding to the four second-level subbands; the calculation method is: L00=C0+C1+C2+C3; L01=C0+C1−C2−C3; L02=C0−C1+C2−C3; L03=C0−C1−C2+C3.

5. The YUV domain digital watermarking algorithm based on wavelet transform according to claim 4 is characterized in that: The first-level low-frequency subband is m rows * 4n columns of pixels; the second-level low-frequency subband obtained by the Haar wavelet transform is an m-row * n-column coefficient matrix; the L00 represents any coefficient in the coefficient matrix; the actual calculated low-frequency band pixel average value LL = L00 / 16.

6. The YUV domain digital watermarking algorithm based on wavelet transform according to claim 1, characterized in that: Step S5 specifically includes: According to experiments or experience, k+1 typical values ​​of the average value of pixels in the low-frequency band are obtained from small to large intervals, and k intervals or segments are correspondingly divided; Obtaining the watermark intensity amplitudes of the k+1 Y channel embedded watermarks corresponding to the k+1 typical values ​​one by one; The watermark intensity amplitudes of k+1 U channels and V channels embedded in the watermark corresponding to the k+1 typical values ​​are obtained, wherein the watermark intensity amplitudes of the U channels and the V channels embedded in the watermark corresponding to the same typical value are equal; and the watermark intensity amplitude of the Y channel corresponding to the same typical value is smaller than the watermark intensity amplitude of the U channel or the V channel.

7. The YUV domain digital watermarking algorithm based on wavelet transform according to claim 6, characterized in that: In step S5, the watermark intensity amplitude Z is calculated according to the following formula: ； Among them, x0-x k represents k+1 typical values; when calculating the Y channel, the y0-y k represents the watermark intensity amplitude of k+1 said Y channels; when calculating said U channel and said V channel, said y0-y k represents the watermark intensity amplitude of k+1 of the U channel and the V channel; n is any integer number between 1 and k; x0<the calculated low frequency band pixel average value LL<x k When LL is calculated based on its value, the interpolation method is used to insert LL into the corresponding interval (x n-1 , x n ); by two points (x n-1 ,y n-1 ) and (x n ,y n ) Determine a straight line, LL is the horizontal coordinate corresponding to the straight line, and the vertical coordinate of the point LL corresponding to the straight line is calculated to be the watermark intensity amplitude Z.

8. The YUV domain digital watermarking algorithm based on wavelet transform according to claim 1, characterized in that: In step S5, the maximum change in the pixel value of the pixel in the embedded channel caused by the watermark embedding in the time domain is set to Δp, and the corresponding maximum watermark intensity value in the frequency domain is calculated as ΔN=Δp*16; and the watermark intensity alpha=Z*ΔN is calculated.

9. The YUV domain digital watermarking algorithm based on wavelet transform according to claim 3, characterized in that: Step S6 specifically includes: Divide the first secondary high-frequency sub-band into 2*2 blocks, set the watermark embedding mode corresponding to each watermark bit to P, where P includes a 2*2 matrix composed of -1 / 1 elements, and the sum of all elements is 0; embed the watermark bit by bit into each block in the first secondary high-frequency sub-band to obtain the watermarked secondary high-frequency sub-band; L33 is the coefficient of the first secondary high-frequency sub-band, and L33' is the coefficient of the watermarked secondary high-frequency sub-band; L33' = L33 + P * alpha, if Bit=1; L33' = L33- P * alpha, if Bit=0.

10. The YUV domain digital watermarking algorithm based on wavelet transform according to claim 9, characterized in that: Step S7 specifically includes: S71, performing a one-dimensional inverse wavelet transform: performing a one-dimensional Haar wavelet inverse transform on the first secondary low-frequency sub-band, the first secondary vertical mid-high frequency sub-band, the first secondary horizontal mid-high frequency sub-band, and the watermarked secondary high frequency sub-band to obtain an inverse-transformed first-level high frequency sub-band; B0', B1', B2', B3' are respectively the pixel values ​​of the 4 pixels of the 1*4 pixel block in the inverse transformed first-level high frequency sub-band; B0'=L30+L31+L32+L33'; B1'=L30+L31−L32−L33'; B2'=L30−L31+L32−L33'; B3'=L30−L31−L32+L33'.

11. The YUV domain digital watermarking algorithm based on wavelet transform according to claim 10, characterized in that: Step S7 specifically also includes: S72, performing a two-dimensional inverse wavelet transform: performing a two-dimensional Haar wavelet inverse transform on the primary low-frequency sub-band, the primary vertical mid-high frequency sub-band, the primary horizontal mid-high frequency sub-band, and the inverse-transformed primary high frequency sub-band to obtain the watermark channel; A0', A1', A2', and A3' are the pixel values ​​of the four pixels of the 2*2 pixel block in the watermark channel; L3' is the coefficient of the inverse transformed first-level high-frequency subband; L0 is the coefficient of the first-level low-frequency subband, L1 is the coefficient of the first-level vertical mid-high frequency subband; and L2 is the coefficient of the first-level horizontal mid-high frequency subband. A0'=L0+L1+L2+L3'; A1'=L0+L1−L2−L3'; A2'=L0−L1+L2−L3'; A3'=L0−L1−L2+L3'.

12. The YUV domain digital watermarking algorithm based on wavelet transform according to claim 9, characterized in that: Also includes: Watermark extraction; The watermark extraction includes: Separating the YUV channel of the watermarked image to obtain the watermarked channel; the watermarked channel is any one of the watermarked Y channel, the watermarked U channel, and the watermarked V channel; Performing a first-level wavelet transform, performing a 2-dimensional Haar wavelet transform on the watermark channel with a 2*2 pixel block as a unit to obtain four first-level extraction subbands; the four first-level extraction subbands include: a first-level extraction high-frequency subband; Performing a secondary wavelet transform, performing a one-dimensional Haar wavelet transform on the primary extracted high-frequency subband with a 1*4 pixel block as a unit to obtain four secondary extracted subbands; the four secondary extracted subbands include: a secondary extracted high-frequency subband; The coefficients in the secondary extracted high-frequency sub-band are divided into blocks to obtain coefficient matrix blocks; the watermark bit value of each coefficient matrix block is calculated and extracted in sequence, and all the extracted watermark bit values ​​Bit are spliced ​​to obtain the final watermark; in each coefficient matrix block, if the result of the matrix dot multiplication operation of the coefficient matrix block and the watermark embedding mode P is greater than 0, then Bit = 1; if the result of the matrix dot multiplication operation of the coefficient matrix block and the watermark embedding mode P is less than 0, then Bit = 0.

Images