CN101534441B - AVS Video Watermarking Method Based on Gray Theory and Uniform Spectrum Theory - Google Patents

Abstract

The invention discloses an AVS video watermarking method based on gray theory and uniform spectrum theory, which mainly solves the problem that the prior video watermarking method is small in embedding capacity and poor in robustness. When a watermark is embedded in an AVS video in a coding process, firstly, an original-gray image watermark is extracted by utilizing gray prediction, compressed and embedded with information content and converted into a binary series; secondly, the correlation degree among host video blocks is calculated by utilizing gray correlation, and a block at minimum correlation degree is selected as a watermark embedding position, and an embedding region is subjected to uniform spectrum processing; and finally, the parity of a quantization coefficient after uniform spectrum processing is adjusted so as to embed the watermark. When the watermark is extracted, a watermark-extracting position is determined according to a key, and a binary watermark is extracted in a mode reciprocal with embedding and is recovered into a gray watermark. Based on AVS standard, the method has the advantages of large embedding capacity and strong robustness, and can be used for the copyright protection of AVS digital multimedia products.

Description

AVS video watermark process based on gray theory and uniform spectrum theory

Technical field

The invention belongs to the multimedia messages process field, relate to the digital video watermarking method of the national audio/video encoding standard of AVS (Audio and Video Coding Standard of China), can be used for the digital multimedia product copyright protection of AVS.

Background technology

The video watermark technology has important commercial value as the copy-right protection method of digital of digital video data, is focus and the difficult point in the current digital watermark research direction.Video data is in large scale, and video watermark process will satisfy requirements such as real-time, constant bit rate, is different from image watermark method, also has the peculiar attack at video sequence, and ripe image watermark method can't directly transplanting.Existing video watermark pertinent literature has:

[1] Maneli Noorkami, Russell M.Mersereau, Compressed-Domain Video Watermarkingfor H.264, IEEE International Conference on Image Processing 2005,2:890-893. utilize in the video I frame macro block difference between adjacent 4 * 4 DC components to select the quantization parameter in the macro block to embed a little at random, only embed a watermark in each macro block as watermark as key, the method capacity is less, and robustness is waited to improve.

[2] Huang Liang, Research on the MPEG-2 Video Watermarking Scheme Based on SpreadSpectrum Technology, IEEE International Conference on Computer Engineering andTechnology, 2009,2:408-411. spread processing is carried out in watermark, watermarked on the DC coefficient of chroma video piece, the embedding capacity is little, and can't satisfy with gray level image, coloured image is the development trend of watermark.

[3] Jianquan Xie, Chunhua Yang, Dazu Huang2, Duosi Xie, A Large Capacity BlindInformation Hiding Algorithm, Proceedings of the International Symposium on ElectronicCommerce and Security, 2008,130:934-937. after having proposed entire image carried out uniform spectrum processing, watermarked on the DCT coefficient, method has increased the embedding capacity and the transparency of watermark, yet the video data volume is huge, real-time requires strong, and this method can not directly apply to video watermark.

The digital audio/video industry is one of three big parts of China's information industry, a key problem of puzzlement China multimedia service is exactly an Intellectual Property Rights Issues, Ministry of Information industry's science and technology department ratifies to have set up digital audio/video encoding and decoding technique standard operation group in June, 2002 for this reason, put forth effort to formulate national audio/video encoding standard AVS, purpose is that range of application comprises mutual medium broadband video service for mobile devices such as wireless network that develops rapidly and mobile phone provide the coding and decoding video codes and standards, multimedia mail, the multimedia service of packet network, monitoring remote video, video conference, video telephone etc.Existing video watermark process is mostly towards mpeg standard, and is at the early-stage towards the video watermark process research of AVS, and the present invention is devoted to the video watermark process research towards AVS.

Summary of the invention

The objective of the invention is to overcome the deficiency of said method, fill up, proposed a kind of AVS video watermark process, to enlarge the watermark embedded quantity, to improve robustness based on gray theory and uniform spectrum theory towards the blank of AVS video watermark process.

The technical scheme that realizes above-mentioned purpose comprises watermark embedding and watermark extracting.

One. AVS video watermark embedding grammar of the present invention comprises the steps:

(1) processing is extracted in watermark to gray level image, and the watermark information amount is compressed to 3/4 of prime information amount, and the gradation of image value that extraction obtains is carried out Binary Conversion, obtains binary watermarking information w;

(2) utilize the AVS encoder that original video is carried out compressed encoding, and extract I-frame video;

(3) the I-frame video macro block brightness data to extracting extract according to 48 * 8 luminance block of AVS standard with each macro block, according to from left to right, top-down sequence arrangement, and number consecutively is 0,1,2,3;

(4) calculate the grey correlation degree of pixel in each 8 * 8 block of pixels respectively, 8 * 8 luminance block to grey degree of association minimum are carried out uniform spectrum processing, and this piece preserved as the first key key1 in the numbering of current macro, the random sequence of using during with uniform spectrum processing is preserved as the second key key2;

(5) on the medium and low frequency quantization parameter of 8 * 8 luminance block after the uniform spectrum processing, the method for adjusting according to odd even embeds binary watermarking information w;

(6) will embed the contrary uniform spectrum processing of 8 * 8 luminance block, conversion, quantification, entropy coding after the binary system watermark, obtain containing the watermark video flowing.

Two. AVS video watermark extracting method of the present invention comprises the steps:

1) utilizes the AVS decoder to decode, and extract I-frame video containing watermark AVS video code flow;

2) I-frame video to extracting in 48 * 8 luminance block of each macro block, is determined to contain 8 * 8 luminance block of watermark according to the first key key1;

3) to containing 8 * 8 luminance block of watermark, utilize the second key key2 to pixel rearrange, the uniform spectrum processing of transform and quantization;

4) on 8 * 8 luminance block after the process uniform spectrum processing, extract watermark according to the parity of medium and low frequency quantization parameter;

5) the watermark decimal system that extracts is converted to gray-scale pixel values, and gray-scale pixel values is arranged according to the two-dimensional matrix form that extracts when handling, it is predicted, recover the gray level image watermark of expression copyright information with gray model GM (1,1).

The present invention has following advantage:

1) the present invention has contracted 1/4 because the gray level image watermark has been carried out extracting processing with the watermark information piezometric, when extracting watermark, utilizes the grey prediction method to recover the original-gray image watermark, has improved watermark and has embedded capacity;

2) the present invention has improved the method disguise owing to use 8 * 8 luminance block of grey degree of association minimum in the grey relation analysis method selecting video I frame macro block watermarked;

3) the present invention is owing to carried out uniform spectrum processing to 8 * 8 selected luminance block, reduced the similitude of neighbor in the block of pixels, the spectrum distribution homogenizing of block of pixels behind the integer transform has increased suitable watermarked big coefficient, has further improved the embedding capacity of method.

Description of drawings

Fig. 1 is a watermark embed process block diagram of the present invention;

Fig. 2 is a watermark extraction process block diagram of the present invention;

Fig. 3 is that the present invention is used for the sectional drawing of emulation testing sequence after the AVS encoding and decoding;

Fig. 4 is the gray level image watermark that the original-gray image watermark used of the present invention and grey prediction are rebuild.

Embodiment

One. the basic theory introduction

(1.1) grey relevance theory

The basic thought of grey correlation analysis is to judge according to the similarity degree of sequence curve geometry whether contact is tight, and curve is approaching more, and the degree of association between corresponding sequence is just big more, otherwise then more little.Grey relation analysis method has remedied the shortcoming that adopts the mathematical statistics method network analysis, to sample size what and have or not all suitablely equally, amount of calculation is little, and is very convenient, the situation that quantized result and The qualitative analysis are not inconsistent can not occur.According to the basic thought of grey relevance theory, the pixel value of image object as a comparison, the correlation degree between the calculating pixel, a certain block of pixels and surrounding associated degree are more little, are modified in visually more difficult discovery.

(1.2) grey prediction theory

Prediction is exactly to inquire into, infer future by means of the experience in past.Gray prediction is an important component part of gray system theory, by the processing of initial data and the foundation of gray model, finds, grasps the system development rule, and the to-be of system is made the scientific quantitative prediction.This new curve fitting technique of accumulative is incorporated in the gray model parameter Estimation, has avoided complicated matrix operation, reduced operand, the precision height does not change many character of model.Grey method based on the grey modeling theory, feature according to forecasting problem can be divided into five kinds of fundamental types, that is: ordered series of numbers prediction, catastrophe prediction, catastrophe prediction in season, topology prediction and system synthesis prediction, these five kinds of Forecasting Methodologies all are methods important and commonly used in the regionl development research.

Grey GM (1,1) modelling desired data is few, amount of calculation is little, has obtained using widely, and the step of using gray model GM (1,1) to carry out the ordered series of numbers prediction is:

Make that original series is X ⁽⁰⁾(k), forecasting sequence is X ' ⁽⁰⁾(k), k=1,2,3...n;

Provide X ⁽¹⁾=(x ⁽¹⁾(t ₁), x ⁽¹⁾(t ₂) ... x ⁽¹⁾(t _n)), wherein $x^{\left(1\right)}\left(t_k\right)=\underset{m=1}{\overset{k}{\mathrm{\Σ}}}{x}^{\left(0\right)}\left(t_m\right);$

Ask X ⁽¹⁾MEAN sequence Z ⁽¹⁾:

Z ⁽¹⁾＝(z ⁽¹⁾(t ₁)，z ⁽¹⁾(t ₂)，…z ⁽¹⁾(t _n))；

z ⁽¹⁾(t _k)＝0.5x ⁽¹⁾(t _k)+0.5x ⁽¹⁾(t _k-1)，k＝1，2，3...n；

Order: $C=\underset{k=2}{\overset{n}{\mathrm{\Σ}}}{z}^{\left(1\right)}\left(t_k\right);$

$D=\underset{k=2}{\overset{n}{\mathrm{\Σ}}}{x}^{\left(0\right)}\left(t_k\right);$

$E=\underset{k=2}{\overset{n}{\mathrm{\Σ}}}{z}^{\left(1\right)}\left(t_k\right)x^{\left(0\right)}\left(t_k\right);$

$F=\underset{k=2}{\overset{n}{\mathrm{\Σ}}}{z}^{\left(1\right)}{\left(t_k\right)}^2;$

$a=\frac{\mathrm{CD}-(n-1)E}{(n-1)F-C^2};$

$b=\frac{\mathrm{DF}-\mathrm{CE}}{(n-1)F-C^2};$

GM (1,1) expression formula is: $u=\frac{1-0.5\×a}{1+0.5\×a};$ $v=\frac{b-a\×x^{\left(0\right)}\left(1\right)}{1+0.5\×a};$

Finally obtain forecasting sequence: X ' ⁽⁰⁾(k)=u ^K-2* v; K=1,2,3...n;

(1.3) uniform spectrum theory

In transform domain, the main component visual of image concentrates on low frequency part, detail section is embodied in the medium-high frequency part, watermarked generally all be chosen in transform domain in, in the low frequency coefficient, but along with the increase of watermark capacity and the restriction of invisibility, watermark just must be embedded in the high frequency coefficient, makes the robustness of method obviously reduce.Image spectrum Energy distribution inequality is owing to neighbor self-similarity in the image causes, if with the pixel of host image scramble at random, eliminate the self-similarity between pixel, then the spectrum energy of image just can be distributed in the whole frequency component relatively fifty-fifty behind the scramble, so just, finish the uniform spectrum processing process, so just can select suitable low frequency coefficient watermarked, increased watermark capacity, encrypt watermark simultaneously, improved the fail safe of method.

Two. the related symbol explanation

W: gray scale watermarking images;

W: binary watermarking;

C: refer to the coded quantization coefficient;

Ww: the binary watermarking that extracts;

WE: the gray value matrix of arranging according to the extraction mode;

W ': extract the gray scale watermarking images after handling;

B0, b1, b2, b3: each brightness 8*8 piece piecemeal is obtained 44 * 4;

h _i: b0, b1, b2, the grey incidence coefficient of b3, i=0,1,2,3;

H _i: four 8 * 8 brightness data pieces grey degree of association separately, i=0,1,2,3;

Key1: the numbering of 8 * 8 luminance block of grey degree of association minimum in each macro block;

Key2: the used length of uniform spectrum processing is 64 random sequence.

Three. AVS video watermark embedding grammar of the present invention

With reference to Fig. 1, AVS video watermark embedding step of the present invention is as follows:

Step 1, W extracts preliminary treatment to the gray scale watermarking images, and the watermark information amount is compressed to 3/4 of prime information amount, and the gradation of image value that remains is carried out Binary Conversion, obtains binary watermarking information w, and concrete steps are as follows:

1.1 the gray level image watermark is divided into some 1 * 8 one-dimension array w _i, number i is determined by the actual size of watermarking images:

w _i＝[w _i(1)，w _i(2)，w _i(3)，w _i(4)，w _i(5)，w _i(6)，w _i(7)，w _i(8)]；

1.2 is the 7th of each one-dimension array of 1 * 8 and the 8th position 0, only keeps first six digits gray-scale pixels information, obtains W ' and is: [w _i(1), w _i(2), w _i(3), w _i(4), w _i(5), w _i(6), 0,0];

1.3 the W ' gray value that obtains after extract handling is carried out decimal system conversion, obtain binary watermarking information w, it is original 3/4 making the watermark information amount that will embed like this, has improved the embedding amount of information.

Step 2 utilizes the AVS encoder that original video is carried out compressed encoding, and extracts I-frame video.

In the process of AVS coding, frame of video is divided into three types of I, B, P, and according to the sequence arrangement of IBBPBBP, the base unit of coding is 16 * 16 macro block, the present invention extracts I-frame video in the mode that extracts all I frame macro blocks in to original video AVS cataloged procedure.

Step 3 is done grey correlation analysis to the brightness data of I frame macro block, with 8 * 8 uniform spectrum processing of degree of association minimum.

3.1 48 * 8 luminance block of each macro block of I frame are extracted, according to from left to right, top-down sequence arrangement, and number consecutively is 0,1,2,3;

3.2 according to from top to bottom, order piecemeal from left to right obtains 44 * 4 sub-piece, is designated as b0, b1, b2, b3 to each 8 * 8 block of pixels; With b0, b1, b2, the pixel transitions among the b3 is that length is 16 one-dimensional sequence, is designated as:

X _i＝(x _i(1)，x _i(2)，…，x _i(16))，i＝0，1，2，3；

3.3 calculate b0, b1, b2, the grey incidence coefficient of each sequence of b3, concrete steps are as follows:

3.3.1 ask b0, b1, b2, the initial value picture of each sequence of b3:

X _i′＝X _i/x _i(1)＝(x′ _i(1)，x′ _i(2)，…，x′ _i(16))，i＝0，1，2，3；

3.3.2 ask b0, b1, b2, the difference sequence of b3:

Δ _ij＝(Δ _ij(1)，Δ _ij(2)，…，Δ _ij(16))；

Wherein, Δ _Ij(k)=| x ' _i(k)-x ' _j(k) |, i=0,1,2,3; J=0,1,2,3; I ≠ j; K=1,2,3...16;

3.3.3 ask the maximum and the minimum value of difference sequence, be respectively:

$M_i=\underset{j}{\max }\underset{k}{\max }{\mathrm{\Δ}}_{\mathrm{ij}}\left(k\right);$

$m_i=\underset{j}{\min }\underset{k}{\min }{\mathrm{\Δ}}_{\mathrm{ij}}\left(k\right);$

I=0 wherein, 1,2,3; K=1,2,3...16;

3.3.4: the grey incidence coefficient h that asks each sequence _i:

$h_i=\underset{j=1,j\≠1}{\overset{4}{\mathrm{\Σ}}}\underset{k=1}{\overset{16}{\mathrm{\Σ}}}{\mathrm{\γ}}_{\mathrm{ij}}\left(k\right)$ i＝0，1，2，3；k＝1，2，3...16；

${\mathrm{\γ}}_{\mathrm{ij}}\left(k\right)=\frac{m_i+{\mathrm{\ζM}}_i}{{\mathrm{\Δ}}_{\mathrm{ij}}\left(k\right)+{\mathrm{\ζM}}_i},$ ζ∈(0，1)，i＝0，1，2，3；j＝0，1，2，3；i≠j；

Wherein get ζ=0.5 among the present invention;

3.4 with the grey incidence coefficient summation of four sequences of each 8 * 8 block of pixels, obtain the grey degree of association H of this 8 * 8 block of pixels, $H=\underset{i=1}{\overset{4}{\mathrm{\Σ}}}{h}_i,$ I=0,1,2,3; Be used as the first key key1 under the numbering preservation of 8 * 8 block of pixels in current macro with grey degree of association minimum;

3.5 to use length be 64 random sequence to the pixel of the piece of this ash degree of association minimum rearrange, conversion, quantification, finish uniform spectrum processing, and used random sequence preserved as the second key key2.

Step 4, on through the medium and low frequency quantization parameter of 8 * 8 luminance block after the uniform spectrum processing, embed binary watermarking information w, owing to adjust the quantization parameter parity with watermarked method according to watermark information, it is less to the change of quantization parameter to compare additive method, can improve the transparency of algorithm, the present invention adopts this method watermarked, and embedding formula is:

if?c％2≠w，c＝c+q；

Wherein c refers to the quantization parameter of the watermark information to be embedded selected, and q is a positive integer, can regulate watermark embed strength.Because the gray value of image is through after the Binary Conversion, highest order is a most significant bit, thereby the present invention embeds the watermark information of highest order on DC coefficient.

Step 5 to the contrary uniform spectrum processing of 8 * 8 luminance block behind the embed watermark information, is carried out the AVS coding again, obtains containing the watermark video flowing, and concrete steps are as follows:

5.1 the random sequence when 8 * 8 luminance block that contain watermark are used rearrangement is aligned to original position to each pixel, finishes contrary uniform spectrum operation;

5.2, carry out conversion, quantification, entropy coding, obtain containing the watermark video flowing to 8 * 8 luminance block after the contrary uniform spectrum processing.

Four. AVS video watermark extracting method of the present invention

With reference to Fig. 2, digital watermarking leaching process of the present invention is as follows:

Step 1 is utilized the AVS decoder to decode to containing watermark AVS video code flow, and is extracted I-frame video.

In the AVS standard, the base unit of decoding is 16 * 16 macro block, and the present invention extracts I-frame video in the mode that extracts all I frame macro blocks in to AVS coded video bit stream process.

Step 2 in each macro block of I-frame video, is determined to contain 8 * 8 luminance block of watermark according to the first key key1.

With 48 * 8 luminance block of each macro block according to from left to right, top-down sequence arrangement, and number consecutively is 0,1,2,3, if the numbering of preserving among the numbering of a certain 8 * 8 luminance block and the first key key1 is consistent, then definite this piece is the luminance block that contains watermark.

Step 3, to containing 8 * 8 luminance block of watermark, utilize the second key key2 to pixel rearrange, the uniform spectrum processing of transform and quantization.

Step 4 on through 8 * 8 luminance block after the uniform spectrum processing, is extracted binary watermarking, and concrete formula is:

ww＝c％2；

Wherein c contains the quantization parameter of watermark 8 * 8 luminance block after the decoding, and it is surplus that c%2 is that c gets 2.

Step 5 is converted to gray-scale pixel values with the binary watermarking ww decimal system that extracts, and the two-dimensional matrix form when it is expanded the back according to reservation process is arranged.

5a is converted to the binary watermarking ww decimal system that extracts gray value and is arranged as one-dimensional sequence wg, the one-dimension array wgb=[wgb (1) that wg is divided into i individual 1 * 6, wgb (2), wgb (3), wgb (4), wgb (5), wgb (6)], concrete number i is determined by the actual size of watermarking images, each array wgb of 1 * 6 is expanded to wgb ':

wgb＝[wgb(1)，wgb(2)，wgb(3)，wgb(4)，wgb(5)，wgb(6)，0，0]；

Mode when 5b handles all wgb ' according to extraction is arranged as two-dimensional matrix wE.

Step 6 is predicted the gray-scale pixels value matrix wE that arranges by the extraction mode with gray model GM (1,1) intension type, recovers the gray level image watermark of expression copyright information.

Effect of the present invention can illustrate by following experiment simulation:

1 simulated conditions

Experiment selects for use the foreman.cif sequence as test video sequence 1, and the carphone.yuv sequence is as test video sequence 2, and size is that gray level image-Xi electrograph sheet of 128 * 128 experimentizes as watermark, and the experiment software environment is visualC++6.0.

2 simulation results

Emulation experiment requires to test to the transparency, real-time, the constant bit rate of video watermark process proposed by the invention, test result Y-PSNR PSNR, bit rate Bit rate, scramble time Encoding time are as shown in table 1, and original video and the sectional drawing that contains the watermark video are as shown in Figure 3.

Emulation experiment is tested the robustness of the video watermark process that the present invention proposes, a series of attack tests have been designed, comprise that weight contracts, the spiced salt add make an uproar, Gauss add make an uproar, shear in the rotation, Gauss's low-pass filtering, mean filter, Wiener filtering, frame, frame rotation, frame are average, frame reorganization etc., and be as shown in table 2 through the NC (normalizated correlation coefficient) of the watermark of extracting after the various attack and original watermark.

The gray level image watermark that original-gray image watermark and grey prediction are rebuild as shown in Figure 4.

By table 1 as seen, watermarked after, the Y-PSNR of cycle tests 1 has become 33.46dB by 34.07dB, 0.61dB has only descended, the Y-PSNR of cycle tests 2 has become 34.44dB by 34.81dB, the 0.37dB that only descended, and the visible transparency of the present invention is better; The bit rate of cycle tests 1 has become 194.03kbit/s by 191.12kbit/s, only increased 2.91kbit/s, the bit rate of cycle tests 2 has become 191.90kbit/s by 188.94kbit/s, has only increased 2.96kbit/s, has satisfied the requirement of constant bit rate; 1 scramble time of cycle tests was increased to 53.434 seconds by 51.539 seconds, had prolonged 1.895 seconds, and 2 scramble times of cycle tests were increased to 49.612 seconds by 48.092 seconds, had prolonged 1.52 seconds, had satisfied real-time and had required to have satisfied the real-time requirement.

By table 2 as seen, watermark of extracting after the process various attack and original watermark normalizated correlation coefficient NC value have shown that all greater than 0.67 the present invention has good robustness.

With reference to Fig. 3, wherein Fig. 3 a, 3c are the sectional drawing of test video sequence 1,2 after the AVS encoding and decoding, Fig. 3 b, 3d are the sectional drawing that test video sequence 1,2 is decoded after AVS coding and interpolation watermark again, from Fig. 3 a and 3b as seen, both are no significant difference visually, Fig. 3 c and Fig. 3 d do not have tangible vision difference yet, have shown that the present invention has the good transparency.

With reference to Fig. 4, wherein Fig. 4 a is the original-gray image watermark, and Fig. 4 b is the gray level image watermark that grey prediction is rebuild, and visible grey prediction method still can clearly be rebuild original image after having compressed 1/4 data volume.

Table 1 test video is in the variation of watermarked back coding parameter

Table 2 test video is watermark NC value after various attack

Claims (4)

1. a kind of AVS video watermark embedding method based on gray theory and uniform spectrum theory, comprises the steps: (1) Extract the grayscale image watermark, compress the watermark information amount to 3/4 of the original information amount, and perform binary conversion on the extracted image grayscale value to obtain binary watermark information w; (2) Utilize national audio and video encoding standard AVS encoder to carry out compression encoding to original video, and extract 1 frame video; (3) For the macroblock luminance data of the extracted I frame video, extract 4 8 × 8 luminance blocks of each macroblock of the I frame according to the AVS video coding standard, according to the left-to-right, top-down Arranged in sequence and numbered as 0, 1, 2, 3; (4) Calculate the gray correlation degree of the pixels in each 8×8 pixel block respectively, and perform uniform spectrum processing of rearranging, transforming and quantizing the 8×8 brightness block with the smallest gray correlation degree, and put the block in the current macro The number of the block is saved as the first key key1, and the random sequence used in uniform spectrum processing is saved as the second key key2; (5) On the mid-low frequency quantization coefficients of the 8×8 luminance block processed by the uniform spectrum, the binary watermark information w is embedded according to the parity adjustment method; (6) For the 8×8 luminance block embedded with the binary watermark, first use the random sequence when reordering, arrange each pixel to the original position, and complete the inverse uniform spectrum operation; then perform the inverse uniform spectrum processing on the 8×8 The brightness block is transformed, quantized, and entropy encoded to obtain a watermarked video stream. 2. video watermark embedding method according to claim 1, wherein said in the step (4) calculates respectively the degree of gray correlation of the pixel in each 8 * 8 pixel block, is to divide the pixel in the 8 * 8 block according to from From left to right, from top to bottom, it is divided into four system behavior sequences, and the gray correlation coefficient of each sequence is calculated; then they are summed to obtain the gray correlation degree of the 8×8 block. 3. a kind of AVS video watermark extracting method based on gray theory and uniform spectrum theory, comprises the steps: 1) Utilize the national audio and video coding standard AVS decoder to decode the watermarked AVS video code stream, and extract I frame video; 2) For the extracted I-frame video, in the 4 8×8 brightness blocks of each macroblock, determine the 8×8 brightness block containing the watermark according to the first key key1; 3) For the 8×8 luminance block containing the watermark, use the second key key2 to perform uniform spectrum processing of rearranging, transforming and quantizing the pixels; 4) On the 8×8 luminance block processed by the uniform spectrum, the watermark is extracted according to the parity of the middle and low frequency quantization coefficients; 5) convert the extracted watermark decimal into grayscale pixel values, expand and arrange according to the two-dimensional matrix form during extraction; 6) Use the gray model to predict the gray-scale pixel values arranged according to the extraction method, and recover the gray-scale image watermark representing the copyright information. 4. The method for extracting video watermark according to claim 3, wherein in the 4 8×8 luminance blocks of each macroblock described in step 2), determine the 8×8 luminance block containing the watermark according to the first key key1 The luminance block is to arrange the four 8×8 luminance blocks of each macroblock in order from left to right and top to bottom, and number them as 0, 1, 2, 3 in turn. If a certain 8×8 luminance If the number of the block is consistent with the number stored in the first key key1, it is determined that the block is a luma block containing a watermark.

Images