CN102364949A - H.264_advanced video coding (AVC)-video-stream-integrity-based fragile watermark authentication method - Google Patents

Abstract

The embodiment of the invention discloses an H.264_advanced video coding (AVC)-video-stream-integrity-based fragile watermark authentication method, which comprises the following steps of: replacing a corresponding codeword by adopting adaptive variable length coding and exponential Golomb coding in an H.264_AVC compressed bit stream to embed watermark information; resolving the H.264 slice-level compressed bit stream, and performing decoding to generate a feature code according to a key; resolving a macro block-level bit stream and extracting watermark; and performing integrity authentication on a video frame according to the decrypted feature code and the extracted watermark. By the method, the watermark is embedded and extracted based on the compressed bit stream, so the method is low in computational complexity and relatively more practical; after the watermark is embedded, the length of the codeword is unchanged, and video distortion is relatively lower; and the rapid extraction of the watermark can be realized, and the requirements of real-time video processing can be met.

Description

A kind of based on the method for video flowing integrated authentication fragile watermark H.264_AVC

Technical field

The present invention relates to communication technical field, relate in particular to a kind of based on the method for video flowing integrated authentication fragile watermark H.264_AVC.

Background technology

Along with the develop rapidly of multimedia network technology, the interchange of digital medium information, particularly video information has reached unprecedented degree.Yet, video information in transmission course, very easily suffer various be not intended to or have a mind to distort attack, this makes people throw doubt upon to the integrality of video information and the authenticity of content.If when distorting important contents such as relating to national security, court's proof, then may cause serious consequence.Therefore, how in network environment, authenticity, the integrality of video information content to be implemented effective protection and become the focus of multi-media information security area research in recent years.Fragile watermarking is to address this problem one of effective method as a kind of private communication technology.It embeds watermark informations such as data, image in the video information with the non mode of people, through to this watermark information.

Existing fragile watermarking algorithm based on H.264/AVC mainly is through modulation movement vector or medium-high frequency DCT coefficient embed watermark information, and the significant disadvantages of these class methods is that computation complexity is big.When watermark embeds, need earlier video bit stream to be decoded, and then combine coding standard modulation movement vector or DCT coefficient; During watermark detection, to decode to video bit stream equally earlier.

In order further to improve the practicality of video watermark, a kind of effective method is direct embed watermark information in video bit stream, promptly according to watermark information to be embedded, revises Variable Length Code (VLC) code word of code stream.Yet efficient compression performance H.264 makes it very responsive to the minor variations in the coded bit stream, and this makes realizes having very big difficulty and challenge based on the watermark in code stream territory H.264.Through revising the video flowing embed watermark of 16 * 16 coding modes in the frame, this scheme has well not sentience and less code check variation in the prior art, yet in the cataloged procedure of reality, the type macro block is relatively more rare, and watermark capacity is less relatively.Also have at present through video sequence is carried out statistical analysis; CAVLC (based on contextual adaptive variable length coding) code word space is divided into 2 groups of code word space commonly used and untapped code word space; According to the mapping relations embed watermark of waiting to embed bit and code word space, this scheme has preferably not sentience equally then, and the embedding capacity is also relatively large; Could set up the mapping regulation but must carry out statistical analysis to video sequence earlier, and bigger to the influence of code check.

Summary of the invention

The present invention proposes a kind of H.264/AVC video watermark process that shines upon based on Exp-Golomb (index Columbus) code word through to the analysis of motion vector encoder bit stream, promptly in compression bit stream H.264, through replacing corresponding code word embed watermark information.The main feature of this method is: 1) watermark embeds and extracts and is based on all that compression bit stream carries out; Computation complexity is low; Having better practicability 2) watermark do not change the length of code word after embedding; And have a less video distortion 3) can realize the rapid extraction of watermark satisfying the real-time video treatment requirement.

In order to address the above problem, the present invention proposes a kind ofly based on the method for video flowing integrated authentication fragile watermark H.264_AVC, comprising:

H.264_AVC adopting adaptive variable length coding and index Columbus through the corresponding code word embed watermark information of replacement in the compression bit stream;

Resolve H.264 chip level compression bit stream, according to key decoding generating feature sign indicating number; Resolve the watermark of macro-block level bitstream extraction;

According to the condition code of deciphering and the watermark of extraction frame of video is carried out integrated authentication.

Said adaptive variable length coding and the index Columbus of in compression bit stream H.264_AVC, adopting comprises through replacing corresponding code word embed watermark information:

The adaptive variable length coding is used for the DCT residual error coefficient, and index Columbus is used for motion vector residual error d _MvAnd side information, wherein d _MvThe residual error motion vector of representing each sub-block.

Said adaptive variable length coding and the index Columbus of in compression bit stream H.264_AVC, adopting comprises through replacing corresponding code word embed watermark information:

Utilize d _MvThe code word space of element and wait to embed the mapping relations between the bit is embedded into watermark in the bit stream of B frame and P frame.

Said method also comprises:

The coding mode of selecting 4*4 piece in the I frame is as watermark information.

Said adaptive variable length coding and the index Columbus of in compression bit stream H.264_AVC, adopting comprises through replacing corresponding code word embed watermark information:

Resolve H.264 chip level compression bit stream, if when anter is the I frame, then extracts authentication code and generate watermark information, if when anter is not the I frame, then embed watermark arrives H.264 bit stream.

The present invention proposes a kind of directly towards the video watermark algorithm of compression bit stream H.264.At first select 4 * 4 among the I_Slice, and generate authentication code, then its form with watermark is embedded in the compression bit stream of B_Slice and P_Slice according to its coding mode by key.The telescopiny of watermark is exactly the modulated process to residual error motion vector code word, and embedded location is determined by key, has strengthened the fail safe of watermark.Embedding of the present invention and leaching process are simple, quick, and have good disguise.Watermark only has a spot of influence to video quality after embedding, and to not influence of video bitrate.The watermarking algorithm that the present invention proposes has fail safe preferably, is mainly reflected in: 1) have identical code word size between the code word of mutual mapping, i.e. watermark can not impact the video file size after embedding; 2) embedding of Feature Extraction and watermark has improved the fail safe of watermark all by cipher controlled; 3) coding mode of I_Slice and motion vector all are the important informations in the encoding and decoding, in the transmission course of compression bit stream H.264/AVC, have very high reliability, can guarantee the correct extraction of watermark.

Description of drawings

In order to be illustrated more clearly in the embodiment of the invention or technical scheme of the prior art; To do to introduce simply to the accompanying drawing of required use in embodiment or the description of the Prior Art below; Obviously, the accompanying drawing in describing below only is some embodiments of the present invention, for those of ordinary skills; Under the prerequisite of not paying creative work, can also obtain other accompanying drawing according to these accompanying drawings.

Fig. 1 is the syntactic structure sketch map of the coded slice in the embodiment of the invention;

Fig. 2 is the mapping relations sketch map between the code word in the embodiment of the invention;

Fig. 3 is 9 kinds of prediction direction sketch mapes of the intra-frame 4 * 4 in the embodiment of the invention;

Fig. 4 is that the watermark in the embodiment of the invention embeds flow chart;

Fig. 5 is the watermark detection process sketch map in the embodiment of the invention.

Embodiment

To combine the accompanying drawing in the embodiment of the invention below, the technical scheme in the embodiment of the invention is carried out clear, intactly description, obviously, described embodiment only is the present invention's part embodiment, rather than whole embodiment.Based on the embodiment among the present invention, those of ordinary skills are not making the every other embodiment that is obtained under the creative work prerequisite, all belong to the scope of the present invention's protection.

The present invention proposes a kind ofly, comprising: in compression bit stream H.264_AVC, adopt adaptive variable length coding and index Columbus through replacing corresponding code word embed watermark information based on the method for video flowing integrated authentication fragile watermark H.264_AVC; Resolve H.264 chip level compression bit stream, according to key decoding generating feature sign indicating number; Resolve the watermark of macro-block level bitstream extraction; According to the condition code of deciphering and the watermark of extraction frame of video is carried out integrated authentication.

The present invention proposes a kind of H.264/AVC video watermark process that shines upon based on Exp-Golomb (index Columbus) code word through to the analysis of motion vector encoder bit stream, promptly in compression bit stream H.264, through replacing corresponding code word embed watermark information.The main feature of this method is: 1) watermark embeds and extracts and is based on all that compression bit stream carries out; Computation complexity is low; Having better practicability 2) watermark do not change the length of code word after embedding; And have a less video distortion 3) can realize the rapid extraction of watermark satisfying the real-time video treatment requirement.

H.264/AVC mainly adopt CAVLC and Exp-Golomb, the binary code word CAVLC that is encoded into syntactic element variable-length is mainly used in the DCT residual error coefficient, and Exp-Golomb is mainly used in motion vector residual error (d _Mv) and side information, like syntactic elements such as coding mode, reference frame, quantization parameters, its structure can be expressed as

[m?zeros][1][b]

In the formula, m zeros is called prefix, and by m zero the composition, b is the m bit field of the information of carrying, and the definition of m and b is following:

m＝floor(log2(n+1)) (1)

b＝n1-2m (2)

In the formula, floor () expression rounds downwards, and n representes the index of each code word, and it is shone upon according to certain mode by coding parameter k and obtains.Its objective is and give short code word, H.264/AVC will select different mapping modes according to dissimilar coding parameter k and give long code word to the value that does not often occur to the value of frequent appearance.For the motion vector residual error, k will be mapped to n by following formula

$n=\left\{\begin{array}{cc}2\left|k\right|& k\≤0\\ 2\left|k\right|-1& k>0\end{array}\right.---\left(3\right)$

Compare with the CAVLC code word, the Exp-Golomb code word is less to the dependence of video content, is more suitable for embed watermark information.But because the change of side information code word is bigger to the influence of video, the present invention selects motion vector residual error element to embed carrier as watermark.

H.264/AVC in order to prevent the error diffusion of coded data, the frame of video that a width of cloth width of cloth is encoded is divided into one or more separate sheets (slice), and encapsulates as unit bit stream.Fig. 1 is the syntactic structure of coded slice, and wherein head have defined the type when anter (Scur), and its value is suc as formula shown in (4), d _MviThe residual error motion vector of representing each sub-block

The present invention utilizes the code word space of dmv element and waits to embed the mapping relations between the bit, watermark is embedded in the bit stream of B_Slice and P_Slice.Revise the method for motion vector embed watermark with common passing through and compare, directly embed watermark has better practicality in the code stream territory.Yet H.264 the structure to code stream has strict definition, if optionally revise the serious distortion that code word can cause video, even can cause the decoder can't normal decoder.Therefore, need optionally embedding.

(1) selection of embedded location

Select amplitude d within the specific limits _MvEmbed watermark information, promptly d0≤| d _Mv|≤d1, get d0=2 among the present invention, d1=7 this mainly be because: 1) in actual coding process H.264/AVC, the residual error motion vector magnitude is general less and revise the less motion vector of amplitude and influence less to video quality and code check; 2) work as motion vector | d _Mv|=0 or 1 o'clock, do not replace code word suitably.But if the position of embed watermark with the set V represent, promptly

$V=\{V_1^p,V_2^p,V_3^p,V_4^p,...,V_i^p,...,V_N^p\}---\left(5\right)$

$V_i^p=\left\{\begin{array}{cc}1& \left|d_{\mathrm{mv}}\right|\∈[d_0,d_1]\\ 0& \left|d_{\mathrm{mv}}\right|\∉[d_0,d_1]\end{array}\right.---\left(6\right)$

In the formula; I representes the macro block sequence number; P representes the sub-piece sequence number in i the macro block, is used for controlling the watermark embedded location at inter prediction H.264/AVC, and each macro block adopts the motion compensation process based on tree type structure; The sub-piece of support from 16 * 16 to 4 * 4 size ranges, promptly each macro block can have 1～16 sub-block.In order to obtain bigger embedding capacity, get p=1 among the present invention, promptly in first sub-block that satisfies condition according to code word mapping ruler embed watermark.

(2) code word mapping ruler

Different d _MvCan be mapped to different n, thereby generate different Exp-Golomb code words.Suppose to meet the d of embedding condition _MvThe codeword set that generates does

S＝{c1，c2，ci，...，cj，...，cM} (7)

In the formula, ci representes various variable length code word, and M representes the sum of code word.Because the redundancy of Exp-Golomb code word itself is very little, direct modification ci code word embed watermark has bigger difficulty, and can cause the confusion of whole code stream.The present invention adopts the mode embed watermark information of code word mapping, promptly with cj replacement ci for example, supposes that cj representes watermark 0, and ci representes watermark 1, is that cj bit current to be embedded is 1 if treat the code word of embed watermark, then is mapped as ci to cj; Otherwise do not revise cj.

In order to guarantee not influence original code flow structure and decoded video quality after watermark embeds, the code word of being shone upon must satisfy following 2 conditions: 1) mapping back code word size does not change, i.e. watermark does not influence video bitrate after embedding; 2) between the code word of mutual mapping, close d must be arranged _MvValue, i.e. watermark does not influence video quality after embedding.According to above-mentioned mapping condition, can codeword set S be divided into two groups of S0 and S1, its mapping relations are as shown in Figure 2.From table 1, can find out the d between the code word of mutual mapping _MvDifference is less, i.e. can not cause obvious influence to video quality after the code word mapping.The mapping relations that different code word S set 0 S1 and two-value are waited to embed between the bit are shown below.

$c_i=\left\{\begin{array}{cc}{S}_0& w_k=0\\ S_1& w_k=1\end{array}\right.---\left(8\right)$

In the formula, ci representes the code word of last mapping, and wk representes k position watermark information.

Table 1 code word and d _MvCorresponding relation

d mv	N	c
			-7	14	0001111
-6	12	0001101
			-5	10	0001011
-4	8	0001001
			-3	6	00111
-2	4	00101
			2	3	00100
3	5	00110
			4	7	0001000
5	9	0001010
			6	11	0001100
7	13	0001110

(3) watermark generates

In existing algorithm, mainly be to extract condition code through the DCT coefficient, yet big in code stream domain analysis DCT coefficient calculations complexity, the present invention selects the coding mode of intra-frame 4 * 4 piece among the I_Slice as watermark information.In cataloged procedure H.264, each intra-frame 4 * 4 piece has 9 kinds to treat lectotype, promptly has 9 kinds of predictive modes as shown in Figure 3.As can be seen from the figure, they are all obtained by the adjacent pixels prediction, and just the direction of prediction is different, and direction difference is more little, and its forecasted variances is also more little.Therefore, when video was interfered or attacks, 4 * 4 block modes very likely changed.Table 2 shows behind recompile, and these data of the transition probability of each pattern (being accurate to 0.0001) are to 10 standard test sequences, nearly 100,000 4 * 4 statistics acquisitions.From table, can find out that 4 * 4 coding mode can change behind the recompile, but between the identical or close pattern of prediction direction, transition probability is generally bigger.

The transition probability of each predictive mode of table 2

mode	0	1	2	3	4	5	6	7	8
										0	0.9991	0.0003	0.0002	0.0000	0.0000	0.0000	0.0001	0.0000	0.0000
1	0.1024	0.8079	0.0285	0.0084	0.0065	0.0014	0.0173	0.0000	0.0275
										2	0.1200	0.0854	0.7051	0.0184	0.0274	0.0094	0.0092	0.0000	0.0250
3	0.1863	0.0472	0.1132	0.5751	0.0000	0.0052	0.0096	0.0496	0.0138
										4	0.0569	0.0939	0.0275	0.0089	0.7179	0.0423	0.0462	0.0064	0.0233
5	0.1030	0.0479	0.0617	0.0232	0.0559	0.6786	0.0048	0.0247	0.0090
										6	0.0571	0.1178	0.0312	0.0090	0.0603	0.0052	0.6959	0.0042	0.0192
7	0.1304	0.0493	0.0066	0.0764	0.0110	0.0214	0.0156	0.6722	0.0170
										8	0.0308	0.1073	0.0092	0.0345	0.0030	0.0000	0.0000	0.0065	0.8088

For the ease of generating watermark information; Need these 9 kinds of predictive modes are divided into two groups; The present invention is divided into 4,5 liang of groups with it and is defined as group M0 and group M1 respectively; Promptly organize 4 elements are arranged among the M0; Group have among the M1 5 elements with these 9 kinds of predictive modes according to aforesaid way total

the i.e. 126 kinds of different combinations modes of dividing into groups; In order to guarantee that authentication code has robustness preferably after under attack, best of breed should satisfy mode shifts, occur on the same group as far as possible between.According to the transition probability of each pattern in the table 2, can obtain optimum composition method by following formula

$P=\min \{{(\underset{i,j\∈M_1}{\mathrm{\Σ}}{P}_{\mathrm{ij}}-x)}^2+{(\underset{i,j\∈M_0}{\mathrm{\Σ}}{P}_{\mathrm{ij}}-y)}^2\}---\left(9\right)$

In the formula, x, y represent the number of element in M1 and the M0 group respectively, and Pij representes the transition probability of each pattern.The best packet mode that calculates is suc as formula shown in (10), and as above-mentioned analysis, mode shifts mainly occurs between the close or identical pattern, as 1 (vertically), 6 (level is downward), 8 (level upwards) all in the M group

$\left\{\begin{array}{c}{\mathrm{<figure-callout\; id="1"\; label="M"\; filenames="HSA00000610549400013.png,HSA00000610549400021.png"\; state="\{\{state\}\}">M</figure-callout>}}_1=\left\{\mathrm{0,2,3,5,7}\right\}\\ M_0=\left\{\mathrm{1,4,6,8}\right\}\end{array}\right.---\left(10\right)$

According to the packet mode of M1, M0, the generation step of authentication code is following:

1) resolving H.264 chip level compression bit stream, is I_Slice if work as anter, then carries out the 2nd) step extraction condition code; Otherwise carry out the 3rd) step.

2) select certain 4 * 4 and generate authentication code at random by key according to following formula

$C_k=\left\{\begin{array}{cc}0& \mathrm{mode}\&Element;{\mathrm{<figure-callout\; id="1"\; label="M"\; filenames="HSA00000610549400013.png,HSA00000610549400021.png"\; state="\{\{state\}\}">M</figure-callout>}}_1\\ 1& \mathrm{mode}\&Element;M_0\end{array}\right.---\left(11\right)$

3) repeat two the step finish up to I_Slice.

(4) watermark embeds

Based on above-mentioned embedding condition and mapping ruler, it is as shown in Figure 4 that watermark embeds flow process, and its concrete steps are following:

1) resolving H.264 chip level compression bit stream, is I_Slice if work as anter, then extracts authentication code according to above-mentioned steps, generates watermark information; Otherwise carry out next step, embed watermark.

2) resolve the macro-block level bit stream; If current sub-block satisfies

is true; Then wait to embed the mapping ruler between bit and the code word, carry out the code word replacement according to two-value; Otherwise this sub-piece is embed watermark not.

3) repeated for two steps, finish, then repeat to embed and finish up to video code flow if the watermark information that generates embeds.

It is thus clear that watermark embed process is only chosen suitable code word and is carried out some comparisons and replacement operation in compression bit stream H.264, do not relate to a large amount of computings, have computation complexity low with fireballing characteristics.

(5) watermark detection

Detecting operation for watermark among the present invention is simple, quick, needn't carry out complete video decode, also original media information need be provided.Fig. 5 is the frame structure of watermark detection, and its concrete steps are following.

1) resolving H.264 chip level compression bit stream, is I_Slice if work as anter, then according to certain coding mode generating feature sign indicating number C ' of 4 * 4 of key decoding _k(C ' in theory _k=C _k), extract watermark otherwise carry out next step.

2) resolve the macro-block level bit stream; If current sub-block satisfies

is true; Then extract watermark, otherwise do not extract watermark information according to following formula

$w_k^{\&prime;}=\left\{\begin{array}{cc}0& c_i\&Element;S_0\\ 1& c_i\&Element;S_1\end{array}\right.---\left(12\right)$

3) repeated for two steps and finish in order effectively frame of video to be carried out integrated authentication up to video code flow, the present invention introduces an objective evaluation index, promptly detects error rate to do

$B_{\mathrm{ER}}=\frac{N^{\&prime;}}{N}---\left(13\right)$

In the formula, N ' expression C ' _kAnd w ' _kBit number inequality, N are represented the total number of bits extracted.If B _ER＞τ, then authentification failure; Otherwise authentication success.τ is a precision threshold, and value is 0.001.

The present invention proposes a kind of directly towards the video watermark algorithm of compression bit stream H.264.At first select 4 * 4 among the I_Slice, and generate authentication code, then its form with watermark is embedded in the compression bit stream of B_Slice and P_Slice according to its coding mode by key.The telescopiny of watermark is exactly the modulated process to residual error motion vector code word, and embedded location is determined by key, has strengthened the fail safe of watermark.Embedding of the present invention and leaching process are simple, quick, and have good disguise.Watermark only has a spot of influence to video quality after embedding, and to not influence of video bitrate.The watermarking algorithm that the present invention proposes has fail safe preferably, is mainly reflected in: 1) have identical code word size between the code word of mutual mapping, i.e. watermark can not impact the video file size after embedding; 2) embedding of Feature Extraction and watermark has improved the fail safe of watermark all by cipher controlled; 3) coding mode of I_Slice and motion vector all are the important informations in the encoding and decoding, in the transmission course of compression bit stream H.264/AVC, have very high reliability, can guarantee the correct extraction of watermark.

More than to what the embodiment of the invention provided, carried out detailed introduction, used concrete example among this paper principle of the present invention and execution mode set forth, the explanation of above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, the part that on embodiment and range of application, all can change, in sum, this description should not be construed as limitation of the present invention.

Claims (5)

1. one kind based on the method for video flowing integrated authentication fragile watermark H.264_AVC, it is characterized in that, comprising:

H.264_AVC adopting adaptive variable length coding and index Columbus through the corresponding code word embed watermark information of replacement in the compression bit stream;

Resolve H.264 chip level compression bit stream, according to key decoding generating feature sign indicating number; Resolve the watermark of macro-block level bitstream extraction;

According to the condition code of deciphering and the watermark of extraction frame of video is carried out integrated authentication.

2. as claimed in claim 1 based on the method for video flowing integrated authentication fragile watermark H.264_AVC; It is characterized in that said adaptive variable length coding and the index Columbus of in compression bit stream H.264_AVC, adopting comprises through replacing corresponding code word embed watermark information:

The adaptive variable length coding is used for the DCT residual error coefficient, and index Columbus is used for motion vector residual error d _MvAnd side information, wherein d _MvThe residual error motion vector of representing each sub-block.

3. as claimed in claim 2 based on the method for video flowing integrated authentication fragile watermark H.264_AVC; It is characterized in that said adaptive variable length coding and the index Columbus of in compression bit stream H.264_AVC, adopting comprises through replacing corresponding code word embed watermark information:

Utilize d _MvThe code word space of element and wait to embed the mapping relations between the bit is embedded into watermark in the bit stream of B frame and P frame.

4. as claimed in claim 1 based on the method for video flowing integrated authentication fragile watermark H.264_AVC, it is characterized in that said method also comprises:

The coding mode of selecting 4*4 piece in the I frame is as watermark information.

5. as claimed in claim 4 based on the method for video flowing integrated authentication fragile watermark H.264_AVC; It is characterized in that said adaptive variable length coding and the index Columbus of in compression bit stream H.264_AVC, adopting comprises through replacing corresponding code word embed watermark information:

Resolve H.264 chip level compression bit stream, if when anter is the I frame, then extracts authentication code and generate watermark information, if when anter is not the I frame, then embed watermark arrives H.264 bit stream.

Images