CN104539965A - H.264/AVC compressed-domain video watermark embedding and extracting method - Google Patents

Abstract

The invention discloses an H.264/AVC compressed-domain video watermark embedding and extracting method. Binaryzation is conducted on watermark images, and binary watermark information is obtained through Arnold transformation; compressed encoding is conducted on original video by using an H.264/AVC encoder, 16 4*4 luminance blocks of each macro block are extracted out according to the H.264/AVC video encoding standard, and integer discrete cosine transformation is conducted on each 4*4 luminance block; according to the coefficient of each 4*4 luminance block obtained after integer discrete cosine transformation, each 4*4 luminance block is divided into a texture block body and a non-texture block body; according to the binary watermark information, the local energy difference of each macro block is modulated; entropy encoding is conducted on the 4*4 luminance blocks obtained after binary watermarks are embedded to obtain video streaming with the watermarks. According to the H.264/AVC compressed-domain video watermark embedding and extracting method, video information (texture and energy) is fully considered, and the robustness and invisibility of the watermarks are improved; operational complexity is low, no complex parameter choice exists, and higher real-time performance is achieved; the original video is not needed in the watermark extraction, and the H.264/AVC compressed-domain video watermark embedding and extracting method belongs to a blind watermarking method.

Description

One is compressed domain video watermark embedment and extracting method H.264/AVC

Technical field

The present invention relates to a kind of video compression territory digital watermarking, in particular one H.264/AVC compressed domain video watermark embedment and extracting method.

Background technology

Digital Video Watermarking Techniques is one of key technology of multi-media information security, is an important branch in Investigation of Information Hiding Technology field.The release of a large amount of consumer digital video product, makes the market demand of the digital product property right protection technology taking digital watermarking as important component part more urgent.How the video monitoring system Large scale construction of digitlization, networking, protect monitor message to exempt to distort, leak, palm off, make video digital watermark become the study hotspot of public safety field.

H.264/AVC be the video compression coding standard of a new generation formulated by ISO/IEC and ITU-T, no matter in compression efficiency, or be all significantly improved than video code model in the past in network adaptability.H.264/AVC standard is the highest, the most widely used coded system of compression efficiency in field of video compression.

The video digital watermark method proposed at present has:

The people such as Jing Zhang propose Robust video watermarking of H.264/AVC (a kind of robust video watermark process H.264/AVC), IEEE Transactions on Circuits and Systems II:Express Briefs, 2007,54 (2): 205-209.This method is by coming watermarked to a certain intermediate frequency DCT coefficient amendment.The method is difficult to determine modulation parameter, and after re-quantization operation, decoding is difficult to accurately watermark embedment position, location.

The people such as Ersin Esen propose Robust Video Data Hiding Using Forbidden Zone Data Hidingand Selective Embedding (a kind of robust video data hidden method based on immutable region).IEEETransactions on Circuits and Systems for Video Technology,2011,21(8):1130-1137。This method delimit watermark embedment region by controling parameters in conjunction with the selection of quantitative map and block, introduce RA coding opposing erasing.The method has good robustness and invisibility, but does not consider the characteristic of frame of video itself.

Summary of the invention

The object of the invention is to overcome the deficiencies in the prior art, provide one H.264/AVC compressed domain video watermark embedment and extracting method, for H.264/AVC digital product provides reliable watermark embedment and extraction.

The present invention is achieved by the following technical solutions, and one of the present invention is compressed domain video watermark embedding method H.264/AVC, comprises the following steps:

(1) watermarking images is carried out binaryzation, then obtain binary system watermark information w through Arnold conversion;

(2) H.264/AVC encoder is utilized to carry out compressed encoding to original video, to the macro block brightness data of I-frame video, according to H.264/AVC video encoding standard, 16 of each macro block 4x4 luminance block are extracted, according to from left to right, the arrangement of top-down order, and be designated as d successively ₀, d ₁..., d ₁₅, respectively integral discrete cosine transform is carried out to each 4x4 luminance block;

(3) according to the coefficient after each 4x4 luminance block integral discrete cosine transform, each 4x4 luminance block is divided into texture block and non-grain block;

(4) according to binary system watermark information w, the Local energy difference of each macro block is modulated;

(5) will the 4x4 luminance block entropy code after binary system watermark be embedded, obtain containing watermark video flowing.

Described step (3), is divided into the method for texture block and non-grain block to comprise the following steps each 4x4 luminance block:

(3.1) to ac coefficient after each 4x4 luminance block integral discrete cosine transform, calculate its quadratic sum and DC coefficient square difference E:

$E=\underset{i=0}{\overset{3}{\mathrm{\Σ}}}\underset{j=0}{\overset{3}{\mathrm{\Σ}}}{\left|C(i,j)\right|}^2-2*{\left|C\left(\mathrm{0,0}\right)\right|}^2$

In formula, i, j=0,1,2,3, C (i, the coefficient value at the i-th row jth row place j) obtained through integral discrete cosine transform for 4x4 luminance block, the coefficient value at the 0th row the 0th row place that C (0,0) obtains through integral discrete cosine transform for 4x4 luminance block;

(3.2) if E<0, then this 4x4 luminance block is non-texture block, if E >=0, then this 4x4 luminance block is texture block.

In described step (4), according to binary system watermark information w, the method for modulating the Local energy difference of each macro block comprises the following steps:

(4.1) if watermark information w=1, the method for modulating the DCT coefficient after each macroblock quantisation is as follows:

(4.1a) for 4x4 luminance block d ₀, d ₁, d ₂, d ₃if 4x4 luminance block is texture block, then the amplitude of the 9th coefficient after quantification and Zigzag scanning and the 15th coefficient is increased by 1; If 4x4 luminance block is non-texture block, then the amplitude of the 15th coefficient through quantizing and after Zigzag scanning increases by 1;

(4.1b) for 4x4 luminance block d ₁₂, d ₁₃, d ₁₄, d ₁₅if 4x4 luminance block is texture block, then by the 9th coefficient after quantification and Zigzag scanning and the 15th coefficient zero setting; If 4x4 luminance block is non-texture block, then by the 15th the coefficient zero setting through quantizing and after Zigzag scanning;

(4.2) if watermark information w=0, the method for modulating the DCT coefficient after each macroblock quantisation is as follows:

(4.2a) for 4x4 luminance block d ₀, d ₁, d ₂, d ₃if 4x4 luminance block is texture block, then by the 9th coefficient after quantification and Zigzag scanning and the 15th coefficient zero setting; If 4x4 luminance block is non-texture block, then by the 15th the coefficient zero setting through quantizing and after Zigzag scanning;

(4.2b) for 4x4 luminance block d ₁₂, d ₁₃, d ₁₄, d ₁₅if 4x4 luminance block is texture block, then the amplitude of the 9th coefficient after quantification and Zigzag scanning and the 15th coefficient is increased by 1; If 4x4 luminance block is non-texture block, then the amplitude of the 15th coefficient through quantizing and after Zigzag scanning is increased by 1.

One is compressed domain video watermark extracting method H.264/AVC, comprises the following steps:

1) utilize H.264/AVC decoder to containing watermark H.264/AVC video code flow decoding, to the macro block brightness data of I-frame video, according to H.264/AVC video encoding standard, 16 of each macro block 4x4 luminance block are extracted, according to from left to right, the arrangement of top-down order, and be designated as d successively ₀, d ₁..., d ₁₅, and 16 4x4 luminance block are divided into part A and part B, wherein, part A comprises d ₀, d ₁..., d ₇, part B comprises d ₈, d ₉..., d ₁₅, respectively integral discrete cosine transform is carried out to each 4x4 luminance block;

2) according to the coefficient after each 4x4 luminance block integral discrete cosine transform, each 4x4 luminance block is divided into texture block and non-grain block;

3) according to 4x4 luminance block whether texture block and the DCT coefficient after quantizing, the local energy E of 4x4 luminance block is calculated _i:

If 4x4 luminance block is texture block, E _i=u (9) × u (9)+u (15) × u (15);

If 4x4 luminance block is non-texture block, E _i=u (15) × u (15);

Wherein, u (9) and u (15) is respectively 9th and 15th DCT coefficient of 4x4 luminance block through quantizing and after Zigzag scanning, i ∈ 0,1 ..., 15};

4) according to the local energy E of 16 4x4 luminance block of each macro block _i, calculate the Local energy difference D of part A and part B:

D＝(E ₀+E ₁+E ₂+E ₃+E ₄+E ₅+E ₆+E ₇)-(E ₈+E ₉+E ₁₀+E ₁₁+E ₁₂+E ₁₃+E ₁₄+E ₁₅)；

5) if D >=0, the watermark information w'=1 extracted; If D<0, the watermark information w'=0 extracted;

6) utilize Arnold inverse transformation to be reduced by the watermark w' of extraction, recover to obtain image watermark.

The present invention has the following advantages compared to existing technology: the present invention takes into full account video information (texture, energy), improves watermark robustness and invisibility; Computational complexity is low, does not have complicated Selecting parameter, has higher real-time; Extract watermark and do not need original video, belong to blind watermark method.

Accompanying drawing explanation

Fig. 1 is the flow chart of watermark embedment of the present invention;

Fig. 2 is the flow chart of watermark extracting of the present invention.

Embodiment

Elaborate to embodiments of the invention below, the present embodiment is implemented under premised on technical solution of the present invention, give detailed execution mode and concrete operating process, but protection scope of the present invention is not limited to following embodiment.

As shown in Figure 1, the present embodiment video watermark embedding step is as follows:

Step 1, carries out binaryzation by N × N watermarking images W, production matrix W _{n × N}, obtain binary system watermark information w through Arnold conversion, number of transitions is m, and one time Arnold conversion step is as follows:

i’＝(i+j)mod N

j'＝(i+2i)mod N

In formula, i, j are matrix W _{n × N}element subscript, i ', j ' be the new subscript of matrix after Arnold conversion, i, j=0,1 ..., N-1, mod are complementation.

Step 2, utilize H.264/AVC encoder to carry out compressed encoding to original video, to the macro block brightness data of I-frame video, according to H.264/AVC video encoding standard, 16 of each macro block 4x4 luminance block are extracted, according to from left to right, the arrangement of top-down order, and be designated as d successively ₀, d ₁..., d ₁₅, respectively integral discrete cosine transform is carried out to each 4x4 luminance block.

Step 3, according to the coefficient after each 4x4 luminance block integral discrete cosine transform, each 4x4 luminance block is divided into texture block and non-grain block, and concrete steps are as follows:

3.1 calculate the quadratic sum of ac coefficient after each 4x4 luminance block integral discrete cosine transform and DC coefficient square difference E:

$E=\underset{i=0}{\overset{3}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{3}{\mathrm{\&Sigma;}}}{\left|C(i,j)\right|}^2-2*{\left|C\left(\mathrm{0,0}\right)\right|}^2$

In formula, i, j=0,1,2,3, C (i, the coefficient value at the i-th row jth row place j) obtained through integral discrete cosine transform for 4x4 luminance block, the coefficient value at the 0th row the 0th row place that C (0,0) obtains through integral discrete cosine transform for 4x4 luminance block;

If 3.2 E<0, this 4x4 luminance block is non-texture block, if E >=0, this 4x4 luminance block is texture block.

Step 4, according to binary system watermark information w, modulate the Local energy difference of each macro block, concrete steps are as follows:

If 4.1 watermark information w=1, modulate the DCT coefficient after each macroblock quantisation as follows:

4.1.1 for 4x4 luminance block d ₀, d ₁, d ₂, d ₃if 4x4 luminance block is texture block, then the amplitude of 9th DCT coefficient of 4x4 luminance block after quantification and Zigzag scanning and the 15th DCT coefficient through quantizing and after Zigzag scanning is increased by 1; If 4x4 luminance block is non-texture block, then the amplitude of 15th DCT coefficient of 4x4 luminance block through quantizing and after Zigzag scanning is increased by 1;

4.1.2 for 4x4 luminance block d ₁₂, d ₁₃, d ₁₄, d ₁₅if 4x4 luminance block is texture block, then by 9th DCT coefficient of 4x4 luminance block after quantification and Zigzag scanning and the 15th the DCT coefficient zero setting through quantizing and after Zigzag scanning; If 4x4 luminance block is non-texture block, then by 15th the DCT coefficient zero setting of 4x4 luminance block through quantizing and after Zigzag scanning;

If 4.2 watermark information w=0, modulate the DCT coefficient after each macroblock quantisation as follows:

4.2.1 for 4x4 luminance block d ₀, d ₁, d ₂, d ₃if 4x4 luminance block is texture block, then by 9th DCT coefficient of 4x4 luminance block after quantification and Zigzag scanning and the 15th the DCT coefficient zero setting through quantizing and after Zigzag scanning; If 4x4 luminance block is non-texture block, then by 15th the DCT coefficient zero setting of 4x4 luminance block through quantizing and after Zigzag scanning;

4.2.2 for 4x4 luminance block d ₁₂, d ₁₃, d ₁₄, d ₁₅if 4x4 luminance block is texture block, then the amplitude of 9th DCT coefficient of 4x4 luminance block after quantification and Zigzag scanning and the 15th DCT coefficient through quantizing and after Zigzag scanning is increased by 1; If 4x4 luminance block is non-texture block, then the amplitude of 15th DCT coefficient of 4x4 luminance block through quantizing and after Zigzag scanning is increased by 1.

Step 5, will embed the 4x4 luminance block entropy code after binary system watermark, obtain containing watermark video flowing.

As shown in Figure 2, the video watermark extracting method step of the present embodiment is as follows:

Step one, utilize H.264/AVC decoder to containing watermark H.264/AVC video code flow decoding, to the macro block brightness data of I-frame video, according to H.264/AVC video encoding standard, 16 of each macro block 4x4 luminance block are extracted, according to from left to right, the arrangement of top-down order, and be designated as d successively ₀, d ₁..., d ₁₅, be divided into part A and part B, wherein, part A comprises d ₀, d ₁..., d ₇, part B comprises d ₈, d ₉..., d ₁₅, respectively integral discrete cosine transform is carried out to each 4x4 luminance block;

Step 2, according to the coefficient after each 4x4 luminance block integral discrete cosine transform, each 4x4 luminance block is divided into texture block and non-grain block, and concrete steps are as follows:

2.1 calculate the quadratic sum of ac coefficient after each 4x4 luminance block integral discrete cosine transform and DC coefficient square difference E:

$E=\underset{i=0}{\overset{3}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{3}{\mathrm{\&Sigma;}}}{\left|C(i,j)\right|}^2-2*{\left|C\left(\mathrm{0,0}\right)\right|}^2$

In formula, i, j=0, the coefficient value at the i-th row jth row place that 1,2,3, C (i, j) obtains through integral discrete cosine transform for 4x4 luminance block, the coefficient value at the 0th row the 0th row place that C (0,0) obtains for integral discrete cosine transform;

If 2.2 E<0,4x4 luminance block is non-texture block, if E >=0,4x4 luminance block is texture block.

Step 3, according to 4x4 luminance block whether texture block and the DCT coefficient after quantizing, calculate the local energy E of 4x4 luminance block _i:

If 3.1 4x4 luminance block are texture block

E _i＝u(9)×u(9)+u(15)×u(15)；

If 3.2 4x4 luminance block are non-texture block

E _i＝u(15)×u(15)；

In formula, u (9) and u (15) is respectively the 9th DCT coefficient after quantification and Zigzag scanning of 4x4 luminance block and the 15th DCT coefficient through quantizing and after Zigzag scanning, i ∈ 0,1 ..., 15}.

Step 4, local energy E according to 16 4x4 luminance block of each macro block _i, calculate the Local energy difference D of part A and part B:

D＝(E ₀+E ₁+E ₂+E ₃+E ₄+E ₅+E ₆+E ₇)-(E ₈+E ₉+E ₁₀+E ₁₁+E ₁₂+E ₁₃+E ₁₄+E ₁₅)

If step 5 D >=0, the watermark information w'=1 extracted; If D<0, the watermark information w'=0 extracted.

Step 6, utilize Arnold inverse transformation to be reduced by the watermark w' of extraction, recover to obtain image watermark.

Compared with other watermark mode, the present invention makes full use of the visual characteristic of human eye, larger than non-grain region watermark embed strength at texture region, can not cause obvious image fault; Utilize the Local energy difference in macro block selectively watermarked, there is stronger robustness; Watermark extraction process does not need original video, belongs to blind watermatking; Judge texture block by integral discrete cosine transform and in macro block, selectively revise medium-high frequency coefficient process simple, computation complexity is low, does not need Selecting parameter and setting.

The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.

Claims (4)

1. a H.264/AVC compressed domain video watermark embedding method, is characterized in that, comprise the following steps:

(1) watermarking images is carried out binaryzation, then obtain binary system watermark information w through Arnold conversion;

(2) H.264/AVC encoder is utilized to carry out compressed encoding to original video, to the macro block brightness data of I-frame video, according to H.264/AVC video encoding standard, 16 of each macro block 4x4 luminance block are extracted, according to from left to right, the arrangement of top-down order, and be designated as d successively ₀, d ₁..., d ₁₅, respectively integral discrete cosine transform is carried out to each 4x4 luminance block;

(3) according to the coefficient after each 4x4 luminance block integral discrete cosine transform, each 4x4 luminance block is divided into texture block and non-grain block;

(4) according to binary system watermark information w, the Local energy difference of each macro block is modulated;

(5) will the 4x4 luminance block entropy code after binary system watermark be embedded, obtain containing watermark video flowing.

2. one according to claim 1 H.264/AVC compressed domain video watermark embedding method, is characterized in that, each 4x4 luminance block is divided into the method for texture block and non-grain block to comprise the following steps by described step (3):

(3.1) to ac coefficient after each 4x4 luminance block integral discrete cosine transform, calculate its quadratic sum and DC coefficient square difference E:

$E=\underset{i=0}{\overset{3}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{3}{\mathrm{\&Sigma;}}}{\left|C(i,j)\right|}^2-2*{\left|C\left(\mathrm{0,0}\right)\right|}^2$

In formula, i, j=0,1,2,3, C (i, the coefficient value at the i-th row jth row place j) obtained through integral discrete cosine transform for 4x4 luminance block, the coefficient value at the 0th row the 0th row place that C (0,0) obtains through integral discrete cosine transform for 4x4 luminance block;

(3.2) if E<0, then this 4x4 luminance block is non-texture block, if E >=0, then this 4x4 luminance block is texture block.

3. one according to claim 1 H.264/AVC compressed domain video watermark embedding method, is characterized in that, in described step (4), according to binary system watermark information w, the method for modulating the Local energy difference of each macro block comprises the following steps:

(4.1) if watermark information w=1, the method for modulating the DCT coefficient after each macroblock quantisation is as follows:

(4.1a) for 4x4 luminance block d ₀, d ₁, d ₂, d ₃if 4x4 luminance block is texture block, then the amplitude of the 9th coefficient after quantification and Zigzag scanning and the 15th coefficient is increased by 1; If 4x4 luminance block is non-texture block, then the amplitude of the 15th coefficient through quantizing and after Zigzag scanning increases by 1;

(4.1b) for 4x4 luminance block d ₁₂, d ₁₃, d ₁₄, d ₁₅if 4x4 luminance block is texture block, then by the 9th coefficient after quantification and Zigzag scanning and the 15th coefficient zero setting; If 4x4 luminance block is non-texture block, then by the 15th the coefficient zero setting through quantizing and after Zigzag scanning;

(4.2) if watermark information w=0, the method for modulating the DCT coefficient after each macroblock quantisation is as follows:

(4.2a) for 4x4 luminance block d ₀, d ₁, d ₂, d ₃if 4x4 luminance block is texture block, then by the 9th coefficient after quantification and Zigzag scanning and the 15th coefficient zero setting; If 4x4 luminance block is non-texture block, then by the 15th the coefficient zero setting through quantizing and after Zigzag scanning;

(4.2b) for 4x4 luminance block d ₁₂, d ₁₃, d ₁₄, d ₁₅if 4x4 luminance block is texture block, then the amplitude of the 9th coefficient after quantification and Zigzag scanning and the 15th coefficient is increased by 1; If 4x4 luminance block is non-texture block, then the amplitude of the 15th coefficient through quantizing and after Zigzag scanning is increased by 1.

4. one according to claim 1 H.264/AVC compressed domain video watermark extracting method, is characterized in that, comprise the following steps:

1) utilize H.264/AVC decoder to containing watermark H.264/AVC video code flow decoding, to the macro block brightness data of I-frame video, according to H.264/AVC video encoding standard, 16 of each macro block 4x4 luminance block are extracted, according to from left to right, the arrangement of top-down order, and be designated as d successively ₀, d ₁..., d ₁₅, and 16 4x4 luminance block are divided into part A and part B, wherein, part A comprises d ₀, d ₁..., d ₇, part B comprises d ₈, d ₉..., d ₁₅, respectively integral discrete cosine transform is carried out to each 4x4 luminance block;

2) according to the coefficient after each 4x4 luminance block integral discrete cosine transform, each 4x4 luminance block is divided into texture block and non-grain block;

3) according to 4x4 luminance block whether texture block and the DCT coefficient after quantizing, the local energy E of 4x4 luminance block is calculated _i:

If 4x4 luminance block is texture block, E _i=u (9) × u (9)+u (15) × u (15);

If 4x4 luminance block is non-texture block, E _i=u (15) × u (15);

Wherein, u (9) and u (15) is respectively 9th and 15th DCT coefficient of 4x4 luminance block through quantizing and after Zigzag scanning, i ∈ 0,1 ..., 15};

4) according to the local energy E of 16 4x4 luminance block of each macro block _i, calculate the Local energy difference D of part A and part B:

D＝(E ₀+E ₁+E ₂+E ₃+E ₄+E ₅+E ₆+E ₇)-(E ₈+E ₉+E ₁₀+E ₁₁+E ₁₂+E ₁₃+E ₁₄+E ₁₅)；

5) if D >=0, the watermark information w'=1 extracted; If D<0, the watermark information w'=0 extracted;

6) utilize Arnold inverse transformation to be reduced by the watermark w' of extraction, recover to obtain image watermark.