CN101610405A - A kind of is the information steganography method of carrier with the compressed video - Google Patents

Abstract

The invention belongs in the command, control, communications, and information security fields with the compressed video is the information steganography method of carrier.The characteristic of this method is to use in the implementation procedure the prior structure of vlc table in the ISO/IEC 13818-2 standard to increase/subtract value tree AST for one group, when information steganography, select for use certain leaf node among the AST to instruct the operation of embedding grammar automatically, thereby can make the compressed video file size before and after information embeds, significant change not take place, guarantee the transparency of video steganography method.Information in this method embeds and extract operation carries out in the VLC territory fully, thereby has avoided can satisfying the real-time requirement of video steganography method at the decoding of compressing video frequency flow and cataloged procedure again.

Description

A kind of is the information steganography method of carrier with the compressed video

Technical field

The invention belongs to the command, control, communications, and information security fields, specifically be based on the information steganography method in the MPEG-2 compressed video VLC territory of digital chaotic system.

Background technology

Information steganography technology (steganography) is an important branch of Information hiding (information hiding), its objective is the information secret is embedded into to send safely among the carrier and do not cause third-party suspection.The carrier that common selected conduct is hidden can be multiple digital resources such as image, audio frequency, video, document, software.Digital video is big because of its data scale, redundant degree height, and characteristics such as applied range are remarkable day by day as the latent advantage of writing aspect the carrier.In the video steganography method, information transmitter can be embedded into secret information in the disclosed video, have only the legal recipient of appointment to go out wherein information according to the prior cipher key-extraction of agreement, and other people can't aware hiding secret information, thereby realize the safe transmission of secret information.

Digital video can be divided into two classes according to coded format: the original video of uncompressed and the video behind the compressed encoding.At present the most widely used is the compressed video of motion-compensated, as meets the compressed video etc. of Moving Picture Experts Group-2.MPEG-2 reduces the redundant and spatial domain redundancy of time-domain by motion compensation, and image block is carried out dct transform and the Huffman encoding compression with the realization data.Through after the compaction algorithms, video code flow is made of the variable-length encoding (VLC) of a series of expression intra-coded macroblock and motion vector etc.

The latent writing technology of original video be similar to latent writing technology in the digital picture be spatial domain with transform domain in least significant bit (LSB) information embedding method.These class methods need be introduced drift compensation and eliminate problems such as the drift that brought by prediction and deviation accumulation.The latent information of writing of these class methods is difficult to resist the MPEG squeeze operation simultaneously, and needs a decoding and cataloged procedure again for the carrier video file that has compressed, has increased operand, has reduced the operational efficiency of method.

People such as Langelaar are published in the differential power watermarking algorithm (DEW) that proposes in the article " Optimal Differential Energy Watermarking of DCT Encoded Images and Video " of IEEE Transactions on Image Processing calendar year 2001 and force to produce an energy difference at two adjacent areas by the high frequency coefficient that removes the I frame, are used for watermarked 1 or 0.Although algorithm itself has lower computation complexity, when algorithm parameter was chosen, optimized Algorithm will all be carried out model parameter estimation to the DCT coefficient of every each frequency of frame, causes the algorithm computation complexity too high.

Design is latent on the VLC territory writes algorithm and can effectively avoid above-mentioned deficiency.The latent algorithm of writing of VLC the earliest is the mutation that LSB of Image embeds algorithm, propose in 2000 by G.C.Langelaar, be published in the article " Watermarking Digital Image and VideoData " among the IEEESignal Processing Magazine.This method determines that at first a run value is identical, and the level value difference is the subclass of 1 VLC sign indicating number, is referred to as label VLC.Secret information bit and a label VLC compare, if identical, level is constant; If different, use the secret information bit to substitute the LSB (least significant bit) of this level value.Can extract secret information by the LSB of direct reading tag VLC the recipient.This algorithm speed is very fast, and by practical application.Yet because the constraint of label VLC selection rule is stronger, only embedding information is also dangerous in label VLC, and secret information is detected easily, thereby and because the limited latent capacity of writing that has limited algorithm of label VLC kind causes the waste of bearer resource.

Lv Junxian etc. have proposed a kind of VLC territory watermarking algorithm in the article " Real-timeframe-dependent video watermarking in VLCdomain " that Signal Processing:Image Communication delivers, according to concrete certain watermark signal that will embed, revise each (run of a certain macro block, level) Zu level value makes the level mean value of this macro block be higher than or be lower than the level mean value of whole macro blocks.Have lower computation complexity on this theoretical method, but robustness is relatively poor, does not consider the problem of the original size of keeping video flowing simultaneously.

Proposed the DNW algorithm in the ICIP2004 meeting that people such as Ling Hefei propose, and proposed the IDEW algorithm at the article " based on the real-time video watermarking technology of MPEG " of " small-sized microcomputer system " magazine.Substitute energy difference in the DEW algorithm by definition energy difference ratio, as the embedded model of a bit information.Improve blocking choosing of index in the telescopiny simultaneously, reduce vision distortion.This algorithm exists and the identical problem of DEW algorithm essence, promptly owing to block the scope of application that the necessity of high frequency coefficient operation has to a certain degree limited algorithm.

In nearest research, people such as B.G.Mobasseri have provided the harmless framework " Watermarking in JPEG Bitstream " that embeds of a general compression domain in the Security andWatermarking of Multimedia meeting that SPIE holds.The principle of VLC mapping has been used in this research, and based on the design to space encoder, performance is better than simple LSB and embeds algorithm.VLC mapping be based on (run, level) encoding mechanism proposes, this thought has been used in the embedding algorithm of jpeg data stream, but since the difference of coding standard can not be applied in latent the writing of MPEG-2 compressed video.Reason is that experiment shows that B14, B15 can be very little for the information embedded space in encoding.Another problem is to be difficult to announce the VLC that decoding Fang Wei embeds information.In JPEG, can in file header, comprise a plurality of Huffman tables, to be used for dynamically generating the VLC that does not contain secret information.And can not add the Huffman table in the mpeg stream.Therefore, even there are enough space encoders to embed information, extracting them also will be very difficult.

The algorithm " Watermarking of MPEG-2 Video in Compressed DomainUsing VLC Mapping " of people in ACM Multimedia and SecurityWorkshop in 2005 such as B.G.Mobasseri used the limitation that value tree strategy has been solved embedded space in the VLC coding subsequently, but the value tree is participated in computing by the carrier video to construct, investigate promptly whether certain two pairs of VLC coding occurs simultaneously in the whole carrier video, operand is bigger, and the latent operation efficiency of writing algorithm is not high.Need rely on the bigger assisted extraction key of data volume during recipient's information extraction in this algorithm simultaneously, increase burden for the cipher key change link of algorithm.

Summary of the invention

The purpose of this invention is to provide a kind of video VLC territory steganography method based on digital chaotic system, this method can keep latent write take place before and after the video file size constant.

For achieving the above object, the present invention adopts following technical scheme: a kind of is the information steganography method of carrier with the compressed video, and the sampling of compressed video signal is represented that by variable-length encoding VLC the information telescopiny comprises the steps:

A) for the 3-dimensional digital chaos system is provided with initial key, and with any two sequence four values in three real-valued pseudo random sequences that generated, another sequence binaryzation;

B) the carrier video bit stream with compressed format is decoded as the VLC code word;

C) with the piece serial number in every frame, and it is right to choose piece in pairs according to two four value sequences;

D) use binary sequence secret information is carried out the XOR preliminary treatment;

E) according to the pretreated secret information of video sequence frame length mean allocation, to distribute to the secret information bit number and the c of each frame) in the piece logarithm chosen make comparisons, judge then that greater than the latter this video can't embed given secret information as the former, less than the latter, continue following embedding operation as the former;

F) according to vlc table B14 or B15 in the ISO/IEC 13818-2 standard, the VLC variable-length encoding is according to (run in will showing, level) code length of value and VLC code word is classified, the identical level value of run value differs 1 VLC code word and connects into a new code word, and increase/subtract according to the difference structure of two parts code length before and after this new code word value is set, each is the new code word that two VLC sign indicating numbers connect into to the leaf node in the value tree, and each is identical to the new code word front and rear part code length difference in the value tree;

G) set video sequence length variations amount Δ, be used for the change of monitor message embedding the video sequence total length;

H) for each secret information bit, with itself and c) the LSB sum of each VLC sign indicating number level value of piece centering of choosing in the step compares, as identical next information bit of embedding that then continues, as the most last two VLC that then investigate current block centering inequality, and judge the VLC code word of required modification level value according to the currency of length variations amount Δ according to AST;

I) adjust the numerical value of length variations amount Δ according to VLC code word actual modification situation;

J) f is carried out in circulation) to i) step, up to embedding all secret informations.

F) building method that increases in the step/subtract the value tree is: each VLC code word among traversal ISO/IEC 13818-2 standard vlc table B14 or the B15, ask for that to differ through the equal and level value of the corresponding run value in variable length decoding VLD operation back be 1 the right code length difference of each VLC code word, according to this difference with all VLC code words to the grouping, and the code word that guarantees identical difference is in same group, right to each code word, according to asking the corresponding front and back position in the difference operation that two code words are linked to each other respectively, constitute new VLC code word, because the non-prefix of original VLC coding, can guarantee the uniqueness of new code word, each is organized new code word and increases/subtract with one the value data tree structure is represented, wherein new code word be the leaf node that value is set.

G), h) and i) setting and the method for adjustment of sequence length variation delta is in the step: the initial value of this variation delta is set at 0, when whenever finishing once as i) behind the embedding operation of step, calculating current video code stream length changes, and the numerical value of renewal variation delta, Δ plays the effect that auxiliary tutorial message embeds simultaneously, when A＞0, need all subtract the value tree in the current VLC code word revised of needing of coupling, and when Δ＜0, need all increase the value tree in the current VLC code word revised of needing of coupling.

The recipient, the process of information extraction is:

A) will carry close video sequence and be decoded into the VLC code word;

B) method that application of aforementioned is identical and initial key generate three chaos random sequences, and select that to carry close piece right;

C) calculate to carry the close whole VLC code word of centering level value LSB with, be extracted as the secret information after the encryption;

Secret information after using binary sequence and encrypting obtains secret information by the bit XOR.

The present invention has made up many groups of different VLC according to the vlc table in the Moving Picture Experts Group-2 value has been set, the three-dimensional chaos mapping of using scrambling simultaneously produces pseudo-random number sequence, realizes the modulation of secret information to be embedded and choosing AC coefficient in the carrier MPEG video file Intra-coded blocks.To set a variable in the process and measure the big or small current change amount of video file latent writing, when variable for just or when negative, correspondence is chosen to subtract the value tree or increase the value tree is instructed follow-up embedding operation, thereby realizes the dynamic maintenance of the size of initial carrier video.Simultaneously, embed and extract operation and in the VLC territory, carry out entirely, can satisfy real-time requirement.The extraction of information is simple to operate, and does not need to transmit extra assisted extraction information.

Description of drawings

Fig. 1 is information steganography of the present invention and leaching process schematic diagram;

Fig. 2 is the macroblock coding block diagram;

Fig. 3 a, 3b are respectively signal m _iEncryption and decryption process schematic diagram;

Fig. 4 is chaos sequence encryption flow figure;

Fig. 5 a, 5b are respectively the result schematic diagram after preceding clear content of encryption and the encryption;

Fig. 6 is T _A3In a leaf node

Fig. 7 a, 7b are respectively the latent schematic diagram of writing the front and back contrast of Mobile the 1st frame;

PSNR after Fig. 8 writes for MPEG2 video sequence Mobile is latent;

Fig. 9 embeds the schematic diagram that secret information rear video sequence frame-to-frame correlation changes for order;

Figure 10 is the frame-to-frame correlation schematic diagram that embeds secret information rear video sequence with the inventive method.

Embodiment

Provided by the invention is the information steganography method of carrier with the compressed video, and the sampling of compressed video signal is represented that by variable-length encoding VLC the information telescopiny comprises the steps:

K) for the 3-dimensional digital chaos system is provided with initial key, and with any two sequence four values in three real-valued pseudo random sequences that generated, another sequence binaryzation;

L) the carrier video bit stream with compressed format is decoded as the VLC code word;

M) with the piece serial number in every frame, and it is right to choose piece in pairs according to two four value sequences;

N) use binary sequence secret information is carried out the XOR preliminary treatment;

O) according to the pretreated secret information of video sequence frame length mean allocation, to distribute to the secret information bit number and the c of each frame) in the piece logarithm chosen make comparisons, judge then that greater than the latter this video can't embed given secret information as the former, less than the latter, continue following embedding operation as the former;

P) according to vlc table B14 or B15 in the ISO/IEC 13818-2 standard, the VLC variable-length encoding is according to (run in will showing, level) code length of value and VLC code word is classified, the identical level value of run value differs 1 VLC code word and connects into a new code word, and increase/subtract according to the difference structure of two parts code length before and after this new code word value is set, each is the new code word that two VLC sign indicating numbers connect into to the leaf node in the value tree, and each is identical to the new code word front and rear part code length difference in the value tree;

Q) set video sequence length variations amount Δ, be used for the change of monitor message embedding the video sequence total length;

R) for each secret information bit, with itself and c) the LSB sum of each VLC sign indicating number level value of piece centering of choosing in the step compares, as identical next information bit of embedding that then continues, as the most last two VLC that then investigate current block centering inequality, and judge the VLC code word of required modification level value according to the currency of length variations amount Δ according to AST;

S) adjust the numerical value of length variations amount Δ according to VLC code word actual modification situation;

T) f is carried out in circulation) to i) step, up to embedding all secret informations.

F) building method that increases in the step/subtract the value tree is: each VLC code word among traversal ISO/IEC 13818-2 standard vlc table B14 or the B15, ask for that to differ through the equal and level value of the corresponding run value in variable length decoding VLD operation back be 1 the right code length difference of each VLC code word, according to this difference with all VLC code words to the grouping, and the code word that guarantees identical difference is in same group, right to each code word, according to asking the corresponding front and back position in the difference operation that two code words are linked to each other respectively, constitute new VLC code word, because the non-prefix of original VLC coding, can guarantee the uniqueness of new code word, each is organized new code word and increases/subtract with one the value data tree structure is represented, wherein new code word be the leaf node that value is set.

G), h) and i) setting and the method for adjustment of sequence length variation delta is in the step: the initial value of this variation delta is set at 0, when whenever finishing once as i) behind the embedding operation of step, calculating current video code stream length changes, and the numerical value of renewal variation delta, Δ plays the effect that auxiliary tutorial message embeds simultaneously, when Δ＞0, need all subtract the value tree in the current VLC code word revised of needing of coupling, and when Δ＜0, need all increase the value tree in the current VLC code word revised of needing of coupling.

The recipient, the process of information extraction is:

D) will carry close video sequence and be decoded into the VLC code word;

E) method that application of aforementioned is identical and initial key generate three chaos random sequences, and select that to carry close piece right;

F) calculate to carry the close whole VLC code word of centering level value LSB with, be extracted as the secret information after the encryption;

Secret information after using binary sequence and encrypting obtains secret information by the bit XOR.

Main Points at above steganography method describes in detail below:

1, AST steganography method thought

MPEG2 is the video compression coding standard of being used widely, and Digital Television, DVD have adopted the MPEG2 standard, writes carrier and has practical preferably prospect for latent with the MPEG2 video.Because video carries out playback with the speed of per second 25 frames (pal mode) usually, even passed through the video of MPEG2 compressed encoding, still there is very big redundancy, this also is the basis that carrier carries out Information hiding just with the compressed video.

In the MPEG2 compressed video, macro block is basic coding unit, and according to the difference of type of prediction, it can be divided into intra-frame macro block and non-intra-frame macro block two classes again, and the cataloged procedure of process as shown in Figure 1.As seen from the figure, according to whether adopting differential coding the data in the macro block can be divided into two types in the cataloged procedure: adopted the coefficient of differential coding, as the DC coefficient in motion vector, the intra-frame macro block; Do not pass through the coefficient of differential coding, as the AC coefficient in the intra-frame macro block and DC coefficient in the non-intra-frame macro block and AC coefficient.

To adopting the coefficient of differential coding, concealing often needs to adopt compensation technique to avoid the accumulation of error when writing; And the coefficient of process differential coding does not exist in compressed video in a large number, these coefficients are after carrying out variable length decoding, all with (run, level) right form occurs, the number of 0 value coefficient of the current summation about non-zero DCT coefficients of run value representative before in the zigzag ordering, level value is exactly the DCT coefficient after the quantification.Obviously, (run, level) to as the information embedded location time, having only the level value is can be reformed when choosing.Because based on the mpeg encoded standard, if a run value has been changed, variation has taken place in the number that promptly means 0 value DCT coefficient before this coefficient in zigzag arranges, thereby is offset when causing that the DCT coefficient is converted to the spatial domain data, obtains video image changed beyond recognition.And when having only the level value to change, the DCT coefficient positions is constant, and actual change only occurs on the amplitude of its value, may be controlled degrading only owing to information embeds the influence to video image that causes.Therefore will revise the level value in writing and do not go to revise the run value in that the VLC territory is latent.

In the AST method, choose the embedding of two 8 * 8 coefficient block implementation informations according to chaos pseudo random sequence.In order to improve the fail safe of information, will choose in 8 * 8 coefficient block the most last AC coefficient as revising object.Corresponding level value sum with the most last AC coefficient in two 8 * 8 coefficient block is that the even number interval scale embeds information bit 0, is that the odd number interval scale embeds information bit 1 and work as level value sum.Relation as current selected two coefficient block and information bit to be embedded satisfies above-mentioned agreement, and then the embedding of current information bit is finished, need not to make any modification; Relation as current selected two coefficient block and information bit to be embedded does not satisfy above-mentioned timing approximately, choosing suitable increasing/subtract to the value tree according to current video file change amount, is the modification that principle is finished the most last AC coefficient of a piece level value in two coefficient block so that the carrier file size is changed minimum.

2 chaos pseudo random number generators

Pseudorandom number generator among the present invention has been used the three-dimensional Baker mapping disturbance scheme that provides in " a kind of new three-dimensional Baker mapping of digitlization disturbance scheme " literary composition that Zhang Shijie, Liu Bin, Liu Fenlin etc. deliver in " computer application research " the 2nd phase in 2006.The limited dynamics degenerate problem that this scheme took place in the time of can overcoming chaos system and realize in computer and digital device, thus the length that produces three groups of sequence periods satisfies the practical 0-1 pseudorandom binary sequence that requires.

The platform of this digital chaotic system operation is the ordinary PC that 32 word lengths of Windows operating system have been installed, and realizes under the double-precision floating point algorithm.

Get k=t=4, w _i=h _j=0.25, a _Ij=w _ih _j=1/16, L _Ij=W _i* h _j+ H _jThree-dimensional Baker mapping equation is as follows before the compensation:

$\mathrm{\ψ}\left(y\right)=\left\{\begin{array}{cc}4y,& y\∈\left[\mathrm{0,0.25}\right)\\ 4y-1,& y\∈\left[\mathrm{0.25,0.5}\right)\\ 4y-2,& y\∈\left[\mathrm{0.5,0.75}\right)\\ 4y-3,& y\∈\left[\mathrm{0.75,1.0}\right)\end{array}\right.---\left(2\right)$

$\mathrm{\&phi;}\left(z\right)=\frac{\mathrm{<figure-callout\; id="16"\; label="z"\; filenames="A2008100500940019H1.png,A2008100500940019H2.png"\; state="\{\{state\}\}">z</figure-callout>}}{16}+\left\{\begin{array}{cc}1/16,& (x,y)\&Element;I_1{I^{\&prime;}}_1\\ 2/16,& (x,y)\&Element;I_1{I^{\&prime;}}_2\\ \&CenterDot;& \\ \&CenterDot;& \\ \&CenterDot;& \\ 15/16,& (x,y)\&Element;I_4{I^{\&prime;}}_4\end{array}\right.---\left(3\right)$

Be without loss of generality, can select an interval arbitrarily during realization, as m=n=2 and e ₁=e ₂=0.2468123.Three-dimensional Baker mapping after this adding disturbance

φ (z)) form is as follows:

$\mathrm{\&psi;}\left(x\right)=\left\{\begin{array}{cc}\mathrm{\&psi;}\left(y\right),& y\&NotElement;I_2{I^{\&prime;}}_2\\ \mathrm{\&psi;}\left(\frac{1}{e_2}\right(y-\frac{1}{4}),& y\&Element;{\mathrm{<figure-callout\; id="1"\; label="B"\; filenames="A2008100500940019H4.png,A2008100500940019H5.png"\; state="\{\{state\}\}">B</figure-callout>}}_1\mathrm{or}{B}_2\\ \mathrm{\&psi;}(\frac{2}{1-e_2}(y-\frac{1}{4}-\frac{1}{4}{e}_2)+\frac{1}{2}),& y\&Element;B_3\mathrm{or}{B}_4\end{array}\right.---\left(5\right)$

φ (z)=φ (z) (remaining unchanged) (6)

For the ease of the realization of cryptographic algorithm, need sometimes the track of chaotic maps is converted into dissimilar chaos pseudo random sequences.Wherein binary sequence can obtain by defining a threshold function table Γ:

$\mathrm{\&Gamma;}\left(x\right)=\left\{\begin{array}{cc}0& x<T\\ 1& T\&le;x\end{array}\right.,$

Wherein, T is the preset threshold such as span according to selected chaos system.The two-value chaos sequence can be expressed as { Γ (x _k), k=0,1,2 ....T in this programme is set at 0.5.Similar with the binary sequence production method, produce four values and need to set in advance three threshold values.To choose 0.25,0.5,0.75 in this programme.

3, secret information preliminary treatment

Use the pseudo random sequence of pseudorandom number generator generation the secret information sequence is carried out the XOR modulation, obtain the even information of 0-1 statistical property, improve the fail safe of secret information.

Cryptographic algorithm can be expressed as: $c_i=E_{k_i}\left(m_i\right)=m_i\&CirclePlus;k_i;$

Decipherment algorithm can be expressed as: $m_i=D_{k_i}\left(c_i\right)=c_i\&CirclePlus;k_i;$

4, increase/subtract value tree structure

Using increase and decrease is that the AST method is in the key that keeps can finishing automatically under the constant situation of MPEG-2 compressed video size the information embedding operation to the value tree.Information embeds and will cause the VLC coding to change, and the VLC sign indicating number before and after changing is called one to value, and two VLC sign indicating numbers are contacted is a code word, as the leaf node of tree data structure.This tree data structure becomes the value tree, embeds as information to cause the VLC code length elongated, then for increasing value is set; Otherwise, value is set for subtracting.Concrete is as follows to value tree make:

In the MPEG-2 compression process, VLC table that variable-length encoding operation is based on standard is tabled look-up to replace and is finished, i.e. table B14 in the ISO/IEC 13818-2 standard or table B15.If N coding arranged in the vlc table:

V＝{v ₁，v ₂，…v _N}????(7)

V wherein _iBe the long l of i _iVLC coding, i＜j is l so _i＜l _jTo v _iCarrying out embedding operation is v _iOne or more bits overturn.For example, change v _iThe k position, obtain

$v_j^k=\{{\mathrm{<figure-callout\; id="1"\; label="v\; i"\; filenames="A2008100500940019H4.png,A2008100500940019H5.png"\; state="\{\{state\}\}">v</figure-callout>}}_i,v_{i2},\&CenterDot;\&CenterDot;\&CenterDot;{\stackrel{\&OverBar;}{v}}_{\mathrm{ik}},\&CenterDot;\&CenterDot;\&CenterDot;{,v}_{{\mathrm{il}}_i}\}---\left(8\right)$

If v _i ^kExceeded space encoder, then decoder can identify and carry close VLC.Therefore embedding information can not simply be applied mechanically the thought that LSB embeds in the compressed video bit stream that has used Huffman encoding.Random modification (v _iMapping) can produce illegal VLC or violate prefix condition.

Usually the VLC table of the MPEG-2 that is used wherein comprises 113 VLC (sign bit is disregarded) for table B14.The run value is identical and the level value will not have different code lengths simultaneously in the VLC coding.We can increase or reduce 1 set to the value that the VLC coding be divided into a plurality of level variation by code length.Therefore, we are by obtaining the space encoder of an expansion to original VLC pairing:

U＝{u _ij}，i，j∈1，…，N????(9)

Wherein, u _Ij={ v _i, v _j, i ≠ j.Len (v _i-v _j) can presentation code length variation.

To in the value tree, leaf node is associated to value with VLC during VLC encodes.If the former VLC sign indicating number is shorter than the latter, then claim this set for increasing set, the tree that is produced by this set is for increasing tree.If len is (v _i-v _j)=n, this increases tree and can be defined as T so _AnWhen n=0, define this tree and be T ₀Corresponding depreciation tree can be defined with identical rule.

Table 1 has provided at T _A3In the example of some leaf nodes.At each row, by different (run, level) right, the value difference 1 of level defines the corresponding VLC sign indicating number, two VLC code lengths change into 3.According to all leaf nodes, tree can be set up accordingly in increment.Fig. 6 has shown T _A3In the example of a leaf node.

Table 1.T _A3Several VLC are to the example of value node in the tree

5, embed the action judgement

In order to guarantee the constant of video data rate, the present invention introduces an offset Δ, and the value of Δ is used for measuring the variation of current video data length.In telescopiny,, can correspondingly select suitable increasing/value tree by the guidance of Δ.When embedding information m _iThe time, use two chaos pseudo random sequences to select piece b _1iAnd b _2i, with m _iWith piece b _1iAnd b _2iThe LSB of level value sum compares among the contained VLC.If equate that then the level value of two VLC sign indicating numbers does not change.If etc., do not obtain the VLC sign indicating number of least significant end in two pieces respectively.Because the VLC sign indicating number has the characteristic of non-prefix, thereby their can increase accordingly uniquely/be matched in the depreciation tree.If corresponding T is searched in Δ＞0 ₀With increase tree; And if Δ＜0, then T ₀Subtracting tree accordingly will be searched.Two VLC be coded in increase/subtract in the tree all found after, just can the changing value of length be compared, according to the principle that keeps carrier video data size, that less operation of changing value will be performed.The telescopiny of information will be finished and the variation of video data will be limited in the scope less than a VLC code word size under this mechanism.

The leaching process of corresponding information is comparatively simple.Under the control of key and pseudo random sequence, decoder will produce identical pseudo random sequence.Calculate all VLC sign indicating number Level values in two pieces having selected and LSB, thereby and utilize the 3rd sequence to carry out the demodulation acquired information.

6 experimental results

Utilize Mobile (32 frame) and Football normal video cycle testss such as (99 frames) to experimentize, the image size of these video sequences is 720 * 486, and chroma format is YCrCb 4:2:2.At first adopting the MPEG2 video file of normal video coding/decoding system TM5 with above-mentioned sequence boil down to per second 25 frames, 10Mbps code check, is that carrier conceals and writes experiment with these MPEG2 videos then.The information that embeds can be the data of arbitrary format, and embedding information can be by complete errorless not extracting under under fire the situation at concealed MPEG2 video.Table 2 has provided the experimental result of the capacity that embeds.Fig. 7 has provided information and has embedded front and back Mobile sequence the 1st two field picture, does not have tangible vision difference.PSNR value after Fig. 8 has provided that Mobile is latent and writes, visual information are concealed and are write the image deterioration that causes and be defined within the acceptable range.The situation of Fig. 9 for video sequence being taked order embedding information cause the interframe degree of correlation to change.From Figure 10 as seen, this steganography method is mean allocation secret information bit in each frame, thereby makes the frame-to-frame correlation of former video remain unchanged substantially.

The latent hiding capacity of writing of table 2 AST

Video sequence	8 * 8 numbers of encoding in the MPEG2 code stream	The latent maximum of writing of AST is hidden capacity (bit)
			??Mobile	??195974	??97987
??Football	??688563	??344281

Claims (3)

1. one kind is the information steganography method of carrier with the compressed video, it is characterized in that, the sampling of compressed video signal is represented that by variable-length encoding VLC the information telescopiny comprises the steps:

A) for the 3-dimensional digital chaos system is provided with initial key, and with any two sequence four values in three real-valued pseudo random sequences that generated, another sequence binaryzation;

B) the carrier video bit stream with compressed format is decoded as the VLC code word;

C) with the piece serial number in every frame, and it is right to choose piece in pairs according to two four value sequences;

D) use binary sequence secret information is carried out the XOR preliminary treatment;

E) according to the pretreated secret information of video sequence frame length mean allocation, to distribute to the secret information bit number and the c of each frame) in the piece logarithm chosen make comparisons, judge then that greater than the latter this video can't embed given secret information as the former, less than the latter, continue following embedding operation as the former;

F) according to vlc table B14 or B15 in the ISO/IEC 13818-2 standard, the VLC variable-length encoding is according to (run in will showing, level) code length of value and VLC code word is classified, the identical level value of run value differs 1 VLC code word and connects into a new code word, and increase/subtract according to the difference structure of two parts code length before and after this new code word value is set, each is the new code word that two VLC sign indicating numbers connect into to the leaf node in the value tree, and each is identical to the new code word front and rear part code length difference in the value tree;

G) set video sequence length variations amount Δ, be used for the change of monitor message embedding the video sequence total length;

H) for each secret information bit, with itself and c) the LSB sum of each VLC sign indicating number level value of piece centering of choosing in the step compares, as identical next information bit of embedding that then continues, as the most last two VLC that then investigate current block centering inequality, and judge the VLC code word of required modification level value according to the currency of length variations amount Δ according to AST;

I) adjust the numerical value of length variations amount Δ according to VLC code word actual modification situation;

J) f is carried out in circulation) to i) step, up to embedding all secret informations.

2. as claimed in claim 1 is the information steganography method of carrier with the compressed video, it is characterized in that, f) building method that increases in the step/subtract the value tree is: each VLC code word among traversal ISO/IEC 13818-2 standard vlc table B14 or the B15, ask for that to differ through the equal and level value of the corresponding run value in variable length decoding VLD operation back be 1 the right code length difference of each VLC code word, according to this difference with all VLC code words to the grouping, and the code word that guarantees identical difference is in same group, right to each code word, according to asking the corresponding front and back position in the difference operation that two code words are linked to each other respectively, constitute new VLC code word, because the non-prefix of original VLC coding, can guarantee the uniqueness of new code word, each is organized new code word and increases/subtract with one the value data tree structure is represented, wherein new code word be the leaf node that value is set.

3. as claimed in claim 1 is the information steganography method of carrier with the compressed video, it is characterized in that, g), h) and i) setting and the method for adjustment of sequence length variation delta is in the step: the initial value of this variation delta is set at 0, when whenever finishing once as i) behind the embedding operation of step, calculating current video code stream length changes, and the numerical value of renewal variation delta, Δ plays the effect that auxiliary tutorial message embeds simultaneously, when Δ＞0, need all subtract the value tree in the current VLC code word revised of needing of coupling, and when Δ＜0, need all increase the value tree in the current VLC code word revised of needing of coupling.

Images