CN104853215B - The video steganography method kept based on motion vector local optimality - Google Patents
The video steganography method kept based on motion vector local optimality Download PDFInfo
- Publication number
- CN104853215B CN104853215B CN201510184721.7A CN201510184721A CN104853215B CN 104853215 B CN104853215 B CN 104853215B CN 201510184721 A CN201510184721 A CN 201510184721A CN 104853215 B CN104853215 B CN 104853215B
- Authority
- CN
- China
- Prior art keywords
- motion vector
- steganography
- secret information
- vector
- video
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
Abstract
The present invention relates to a kind of video steganography methods kept based on motion vector local optimality.This method establishes a kind of criterion of motion vector local optimality and the searching algorithm of optimal candidate motion vector, herein on basis, cost computation model and preferably efficient steganography code are disturbed by constructing reasonable motion vector, to establish private communication channel to complete the insertion of the steganography of secret information, and then complete the extraction process of secret information.This method preferably uses STC steganography code.This method is a kind of based on the holding of motion vector local optimality, the motion vector field video steganography method for possessing higher Embedding efficiency and capable of being effective against existing steganalysis attack, especially suitable for the higher covert communications scene of security requirement.
Description
Technical field
The present invention relates to a kind of video steganography (Video Steganography) methods, and in particular to one kind is based on movement
The highly concealed type video steganography method and its application in covert communications, this method that vector local optimality is kept belong to information
Information hiding subdomains in security technology area.
Background technique
Present information concealing technology mainly include steganography (Steganography), steganalysis (Steganalysis) with
And digital watermarking (Digital Watermarking).How steganography is mainly studied by secret information embedded images, video, audio
Equal digital multimedia documents are to achieve the purpose that covert communications；Steganalysis mainly uses the methods of machine learning, pattern-recognition
Steganography classification judgement is carried out to file to be measured.
Under normal circumstances, the design needs of steganographic algorithm comprehensively consider and weigh following several respects element:
1) not sentience: carrier before and after the steganography undistinguishable for human body can not only rely on the mankind
Audio visual judge whether file to be measured passes through steganography；
2) robustness: steganography file bear a degree of noise jamming and multi-media processing (if compressed, cut out,
Size change over etc.) after, it still is able to the secret information for accurately restoring wherein to be embedded into；
3) embedding capacity: under the premise of guaranteeing not sentience and robustness to a certain degree, as much as possible to carrier
Middle insertion secret information；
4) Embedding efficiency: under the premise of embedding information amount is certain, the modification number or degree to carrier are reduced as far as possible；
5) safety: even if possessing the priori knowledge about steganographic algorithm, existing steganalysis method also can not be with foot
Enough high probability distinguish general carrier and the file Jing Guo the algorithm steganography.
With making rapid progress for Internet technology, the booming and height of the tertiary industry such as video on demand, video calling
Effect video encoding standard is weeded out the old and bring forth the new, and video relies on its visual performance abundant and powerful information propagation capabilities, just gradually
Replace image, becomes current most popular communication media and more suitable steganography carrier, therefore, video steganography also results in
The extensive concern of this field scholar.
Video steganography can generally be divided into airspace (Spatial Domain) video steganography and compression domain (Compressed
Domain) video steganography: the former is by modifying the original pixel value of video frame directly to be embedded in secret information；The latter is then in video
Cataloged procedure in introduce steganography disturbance so that compressed encoding and steganography insertion can carry out simultaneously.Compressed domain video steganography root
According to the difference of embedded domain, can be divided into based on motion vector (Motion Vector, MV) video steganography (bibliography:
C.Xu,X.Ping,and T.Zhang.Steganography in compressed video stream.In Proc.1st
Int.Conf.Innov.Comput.,Inf.Control,vol.1:269–272,2006.；H.Aly.Data hiding in
motion vectors of compressed video based on their associated prediction
error.IEEE Transactions on Information Forensics and Security,vol.6(1):14–18,
2006.；Y.Cao,X.Zhao,D.Feng,and R.Sheng.Video steganography with perturbed
motion estimation.In Proc.13th Int.Conf.IH,vol.6958:193–207,2011.；Y.Yao,
W.Zhang,N.Yu,and X.Zhao.Defining embedding distortion for motion vector-based
video steganography.Multimedia Tools and Applications,Article in Press,2014,
DOI:10.1007/s11042-014-2223-8.), the video based on transformation coefficient (Transform Coefficient) is hidden
It writes, the video steganography based on intra prediction mode (Intra Prediction Mode), be based on inter-frame forecast mode (Inter
Prediction Mode) video steganography and be based on this several class of the video steganography of entropy coding (Entropy Coding).
By patent consulting, existing related application situation is as follows in the field of the present invention:
(1) " video based on the disturbance of macro block division mode is hidden for the Chinese patent that number of patent application is 201410233838.5
Write method " discloses one kind based on H.264/AVC video encoding standard, and by modification macro block division mode to be embedded in secret letter
The video steganography method of breath.The patented method in telescopiny preferred steganography code STC (Syndrome Trellis Code,
Verify grid code) and WPC (Wet Paper Code, l Water Paper coding) double-deck private communication channel of building, to effectively increase insertion
Capacity and Embedding efficiency.The patented method belongs to the video steganography based on inter-frame forecast mode, is not directed to motion vector
Disturbance modification, therefore, the patent application and general object of the present invention, mentality of designing and specific implementation are significantly different.
(2) Chinese patent " a kind of video steganography side based on motion vector that number of patent application is 201310275143.9
Method " discloses a kind of video steganography method based on motion vector.The patented method carry out secret information steganography insertion when,
Current video sequence is encoded first, obtains the motion vector that wherein all P frames include, and count respectively according to preset distortion function
The steganography insertion cost of resulting each horizontal motion vector and vertical component is calculated, then all horizontal components are formed using STC
Carrier carry out insertion disturbance, and cost is embedded according to the steganography that steganography result adjusts vertical component, then uses STC pairs again
The carrier of all vertical component compositions carries out insertion disturbance, and then recompiles this according to the steganography result of horizontal and vertical component
Video sequence obtains final steganography video file.Distortion function defined in the patented method has comprehensively considered motion vector
Airspace, the residual error that generates of relativity of time domain and estimation, enable the motion vector modified as to keep adjacent as possible
Domain statistical property, while reducing steganography and disturbing the influence generated to video coding performance (code rate, fidelity), therefore, the patent
Method can be effective against motion vector field steganalysis feature MVRB and ADB_MV (bibliography: Y.Cao, X.Zhao, and
D.Feng.Video steganalysis exploiting motion vector reversion-based
features.IEEE Signal Processing Letters,vol.19(1):35-38,2012.；Y.Su,C.Zhang,
and C.Zhang.A video steganalytic algorithm against motion-vector-based
steganography.Signal Processing,vol.91(8):1901-1909,2011.).However, due to the patent side
Method is still traditional " carry out adding one to a certain component of motion vector or subtract one " to the modification mode of motion vector, i.e.,
(mv'x,mv'y)∈{(mvx+1,mvy),(mvx-1,mvy),(mvx,mvy+1),(mvx,mvy- 1) } (wherein (mvx,mvy) and
(mv'x,mv'y) respectively indicate the original and process modified motion vector of steganography), it is therefore possible to destroy the office of motion vector
Portion's optimality (Local Optimality), therefore can not resist the most effective steganalysis feature AoSO's in current motion vector domain
Attack (bibliography: K.Wang, H.Zhao, and H.Wang.Video Steganalysis Against Motion
Vector-Based Steganography by Adding or Subtracting One Motion Vector
Value.IEEE Transactions on Information Forensics and Security,vol.9(5):741-
751,2014.).As a comparison, the angle that the present invention is kept from motion vector local optimality devises completely new reasonable
It is embedded in cost function and motion vector perturbation scheme, the attack of AoSO can be effective against, be a kind of highly concealed type communication party
Method, therefore, the patent application and general object of the present invention, mentality of designing and implementation are significantly different.
Summary of the invention
The object of the present invention is to provide one kind to be kept based on motion vector local optimality, possess higher Embedding efficiency
And the motion vector field video steganography method of existing steganalysis attack can be effective against.
The present invention compares other motion vector field video steganography methods, establishes a kind of sentencing for motion vector local optimality
The searching algorithm of calibration standard and optimal candidate motion vector (Best Candidate MV) passes through building herein on basis
Reasonable motion vector disturbance cost computation model and preferably efficient steganography code, to establish private communication channel to complete secret letter
The steganography of breath is embedded in.As it can be seen that method proposed by the present invention is different from previous any video steganography method, especially suitable for safety
The more demanding covert communications scene of property.
According to investigation, overwhelming majority motion vector field video steganography method does not use steganography code to improve insertion effect at present
Rate and safety, moreover, all motion vector field video steganography methods can not resist AoSO's (referring to background technique)
Attack.Against the above deficiency and disadvantage, improvement of the invention are as follows: firstly, it is preferred that and apply suitable steganography code, not only mention
High Embedding efficiency and safety, also effectively reduce steganography disturbance influence caused by video coding performance；Secondly, abandoning
Traditional motion vector modifies mode, influence of the steganography operation to safety and video coding performance is comprehensively considered, by reasonable
Efficient searching algorithm filters out most suitable local optimum motion vector as final modification as a result, so as to be effective against
The attack of the motion vector fields steganalysis method such as AoSO.
Specifically, The technical solution adopted by the invention is as follows:
A kind of video steganography method kept based on motion vector local optimality, wherein the telescopiny packet of secret information
Include following steps:
1) it pre-processes: in current video frame, to each motion vector (mvx,mvy), calculate LSB (mvx+mvy) obtain two
System number 0 or 1, and write all calculated results as vector x, wherein LSB () expression takes lowest bit position (Least
Significant Bit) operation, mvxAnd mvyRespectively indicate the horizontal and vertical component of motion vector；
2) construct private communication channel: regarding gained vector x in step 1) as the carrier of private communication channel, according to preferred steganography code
Working principle, the parameter appointed in advance according to communicating pair generates parity check matrix H, and combines preset motion vector
Disturbance cost computation model is ultimately constructed to be gone out to meet the x' of Hx'=m, and wherein m indicates secret letter to be embedded in the private communication channel
Breath, x' indicate modified carrier vector；
3) it can determine in current video frame there is which motion vector to need to modify according to step 2)；
4) current video frame is encoded according to result obtained as above, and according to preset optimal candidate motion vector sought algorithm
Each motion vector to be modified is substituted for corresponding optimal candidate motion vector respectively, and then is completed to the secret of private communication channel
Confidential information insertion；
5) according to the frame coded sequence of video file, step 1) is repeated to 4) until secret information is complete to remaining every frame
Portion's insertion finishes, and obtains final steganography video file.
Further, the above method further includes the extraction process of secret information, comprising the following steps:
1) current video frame is decoded, to each motion vector (mvx,mvy), calculate LSB (mvx+mvy) obtain binary number 0
Or 1, and write all calculated results as vector x, wherein LSB () expression takes lowest bit bit manipulation；
2) according to preferred steganography code working principle, generate even-odd check according to the parameter that communicating pair is appointed in advance
Matrix H, calculates Hx, and acquired results are the secret information bit extracted；
3) according to the frame decoding sequence of video file, step 1) is repeated to 2) until secret information is complete to remaining every frame
It extracts and finishes in portion.
On the basis of above scheme, the present invention is further improved, i.e., has carried out preferably to the steganography code of use.Embedding
Enter the private communication channel established in process (Fig. 1), using steganography code Syndrome Trellis Code (STC verifies grid code) into
The insertion of row secret information extracts secret information from the private communication channel according to the working principle of STC in extraction process (Fig. 2).
About being summarized as follows for STC:
If x=(x1,x2,…,xn)TFor original binary carrier vector；X'=(x'1,x'2,…,x'n)TFor by steganography
Carrier vector after insertion；It the real vector that is positive and indicates the element x in xiIt is modified as 1-xi
Cost value；H is parity matrix；M is secret information vector.Then STC can construct the x' for meeting Hx'=m, and make
It is minimum to obtain steganography insertion cost, wherein function δ meets
It, can by estimation (Motion Estimation) in the Inter prediction module (Fig. 3) of video compression coding
To obtain the corresponding motion vector of original block (Original Block), it is directed toward the best prediction block in reference frame
(Prediction Block).The residual block (Residual Block) then obtained is by converting (Transformation)
After quantization (Quantization) operation, result not only needs to carry out entropy coding (Entropy) and obtains binary code stream, also
It needs to obtain reconstructive residual error block by inverse quantization (Rescale) and inverse transformation (Inverse Transformation) operation
(Reconstructed Residual Block), and finally obtain the corresponding reconstructed block (Reconstructed of original block
Block).Since quantization operation is that information detracts irreversible process, reconstructed block and original block be not identical, the phase of the two
It is determined like degree by motion-estimation precision and quantization operation step-length collective effect.
When video encoder carries out estimation to original block to be encoded, usually using rate-distortion optimization (Rate
Distortion Optimization, RDO) compromise between model choosing comprehensively code rate and the distortion factor, in motion vector sought
Selection possesses the minimum i.e. J of RDO functional value in rangemvThe motion vector of=D+ λ R as the corresponding optimal movement of this block to be encoded to
Amount, wherein D indicates the distortion factor between the prediction block that motion vector is directed toward and original block, needed for R presentation code motion vector
Bit number, λ indicate the parameter between adjusting D and R, and the λ the big, more tend to the distortion of low bit- rate height, smaller, more tend to high code
Rate low distortion.
Based on the brief introduction above to inter prediction encoding overall flow and estimation, now to movement proposed by the present invention
The criterion of vector local optimality, motion vector disturbance cost computation model and optimal candidate motion vector sought algorithm are done
As described in detail below.
[1] a kind of criterion of motion vector local optimality:
For motion vector MV0,0=(mvx,mvy), if its prediction block being directed toward is Pred (MV0,0), by inter-prediction
Encoding obtained reconstructed block is Rec (MV0,0).Calculating matrix Pred (MV0,0) and Rec (MV0,0) between absolute error and (Sum
Of Absolute Difference, SAD), i.e. sad0,0=SAD (Rec (MV0,0),Pred(MV0,0)).Similarly, for MV0,0
8 adjacent motion vector MV of surroundingi,j=(mvx+i,mvy+ j) (- 1≤i, j≤1) (Fig. 4), respectively obtain the pre- of their directions
Survey block Pred (MVi,j), and calculate sadi,j=SAD (Rec (MV0,0),Pred(MVi,j)).If sadi,jIn (- 1≤i, j≤1)
Minimum value is sad0,0, then claim motion vector MV0,0Has local optimality.
Note:Wherein, Mi,jAnd Ni,jPositioned at (i, j) in respectively matrix M and N
Element；
[2] a kind of computation model of motion vector disturbance cost:
To motion vector MV0,0=(mvx,mvy) both horizontally and vertically upper 4 adjacent motion vector, that is, MV-1,0=
(mvx-1,mvy)、MV1,0=(mvx+1,mvy)、MV0,-1=(mvx,mvy- 1) and MV0,1=(mvx,mvy+ 1) (Fig. 4) carries out part
Optimal sex determination, if thering is N (0≤N≤4) a motion vector to have local optimality in them, to motion vector MV0,0It carries out
The disturbance cost of steganography insertion are as follows:
Wherein, Ω={ MV-1,0,MV1,0,MV0,-1,MV0,1, i.e., the disturbance cost of motion vector depends on its contiguous range
The quantity of interior local optimum motion vector and the variation degree of RDO functional value；
[3] a kind of searching algorithm of optimal candidate motion vector:
If according to steganography insertion as a result, it is desirable to motion vector MV0,0=(mvx,mvy) modify, then it successively executes
Following steps select most suitable local optimum motion vector as final modification result from all candidate motion vectors:
A) respectively to MV0,0Four motion vector MV in contiguous range-1,0=(mvx-1,mvy)、MV1,0=(mvx+1,
mvy)、MV0,-1=(mvx,mvy- 1) and MV0,1=(mvx,mvy+ 1) local optimum sex determination is carried out；
B) by MV-1,0, MV1,0, MV0,-1And MV0,1In local optimum motion vector composition candidate motion vector set close Ψ,
If Ψ non-empty, go to step e), otherwise continues to execute step c)；
C) region of search of a motion vector, size (the candidate motion vector quantity for including) and final search delimited
Time is directly proportional, it should be chosen according to practical computing resource or ability, in addition, for any Candidate Motion in the region
Vector M V'=(mvx',mvy'), it is necessary to meet LSB (mvx+mvy)≠LSB(mvx'+mvy')；
D) for each candidate motion vector in the region of search, local optimum sex determination is carried out respectively, is determined wherein
The motion vector of all local optimums, and by their composite set Ψ；
E) for each local optimum motion vector in Ψ, its RDO functional value is calculated, selection possesses minimum RDO function
The local optimum motion vector of value is as final modification result.
Based on content described above, being protected based on motion vector local optimality using steganography code STC proposed by the present invention
The video steganography method held, wherein the telescopiny of secret information include the following steps (unless otherwise specified, following steps by
Computer executes):
1) in current video frame, by each motion vector (mvx,mvy) horizontal and vertical component summation after take minimum ratio
Special position, i.e. LSB (mvx+mvy), binary number 0 or 1 is obtained, and all calculated results are formed into binary system carrier vector x；
2) in current video frame, for each motion vector MV, cost is disturbed according to preset motion vector and calculates mould
Its steganography disturbance cost Cost (MV) is calculated in type, and all calculated results are formed vector Γ；
3) private communication channel is constructed: according to the working principle of steganography code STC, first according to the preassigned load factor of communicating pair
α and parameter matrix Η constructs parity matrix ΗSTC, reconstruct out and meet ΗSTCX'=m and make steganography be embedded in generation
Valence D (x, x', Γ) reaches the smallest x', wherein m indicates that secret information to be embedded in private communication channel, x' are indicated by hidden
The carrier vector modified after the insertion of channel steganography；If the binary number on x and x' corresponding position is not identical, then it represents that corresponding
Original motion vector needs to modify, and then indicates not needing to modify if they are the same；
4) according to the modification of carrier vector in private communication channel as a result, for each motion vector for needing to modify, according to pre-
If optimal candidate motion vector sought algorithm be substituted for the smallest local optimum fortune of RDO functional value in candidate motion vector
Moving vector；
5) current video frame is encoded according to result obtained as above to complete the steganography of secret information and be embedded in；
6) according to the frame coded sequence of video file, step 1) is repeated to 5) until secret information is complete to remaining every frame
Portion's insertion finishes, and obtains final steganography video file.
Further, included the following steps (unless otherwise specified, using the extraction process of the secret information of steganography code STC
Following steps are executed by computer):
1) current video frame is decoded, to each motion vector (mvx,mvy) calculate LSB (mvx+mvy) obtain binary number 0
Or 1, and write all calculated results as binary vector x；
2) secret information in private communication channel is extracted: preassigned negative according to communicating pair according to the working principle of STC
Load rate α and parameter matrixConstruct parity matrix ΗSTC, calculate ΗSTCX, acquired results are the secret information extracted
Bit；
3) according to the frame decoding sequence of video file, step 1) is repeated to 2) until secret information is complete to remaining every frame
It extracts and finishes in portion.
Video steganography method of the invention has the beneficial effect that correlative technology field:
1) Embedding efficiency is high, and steganography disturbance is small.Current almost all of motion vector field video steganography method is not all embedding
It uses steganography code to improve Embedding efficiency during entering, thus steganography operation can not be effectively reduced to video coding performance etc.
Caused by disturb so that the compression efficiency of video and fidelity become negatively affected；In contrast, the preferred steganography code of the present invention
STC, so as to complete the steganography insertion of secret information with higher Embedding efficiency, in other words, in the identical condition of load factor
Under, the present invention will be far fewer than other motion vector field video steganography methods, so as to effective to the modification number of carrier video
Steganography is reduced to disturb and keep the coding efficiency of video.
2) it can be effective against steganalysis, have higher-security.Firstly, present invention employs steganography code STC raisings
Embedding efficiency, reduces the modification number and degree to carrier video, to effectively reduce steganography operation to Video coding
It is disturbed caused by performance etc., so that steganography video and the original video undistinguishable in visual perception；Secondly, the present invention is got rid of
Traditional motion vector modification mode has been abandoned, has used preset optimal candidate motion vector sought algorithm to protect in telescopiny
Card is by the local optimality of modification motion vector, so as to be effective against the most powerful steganalysis spy in current motion vector domain
Levy the attack of AoSO.Therefore, the present invention compares other motion vector field video steganography methods, possesses higher steganography safety.
3) customized motion vector disturbs cost computation model.The present invention is by modification motion vector to carry out secret
The steganography of information is embedded in, it is therefore desirable to carry out effective evaluation to the disturbance cost of motion vector.Different evaluation mechanism may be led
It causes different steganography to be embedded in effect and the safety of steganographic algorithm is had an impact.The present invention is adjacent by calculating current motion vector
The variation degree of the quantity of local optimum motion vector and RDO functional value is within the scope of domain to evaluate disturbance generation of this motion vector
Valence, steganography insertion when preferentially modify in contiguous range there are more local optimum motion vectors and RDO functional value variation it is smaller
Those of motion vector.As it can be seen that can to guarantee that video coding performance does not generate as far as possible in steganography telescopiny larger by the present invention
It disturbs and motion vector to be modified can preferentially be changed to the local optimum motion vector (if present) in its contiguous range
To shorten steganography time-consuming.In addition to this, for different application scenarios, the present invention can customize different motion vector disturbance generations
Valence computation model, therefore possess higher flexibility.
It 4) can be by adjusting region of search size to compromise between steganographic algorithm performance and insertion time-consuming.In this hair
In bright preset optimal candidate motion vector sought algorithm, transported when local optimum is not present in motion vector contiguous range to be modified
When moving vector, a region of search can be delimited and exhaustive search wherein goes out the smallest local optimum motion vector of RDO functional value
As final modification result.Under normal circumstances, the area of region of search is bigger, i.e. the quantity of candidate motion vector is more, finally
Obtained best local optimum motion vector is more ideal, it also requires longer search is time-consuming.Therefore, steganography insertion is being carried out
When, demand of the current application scene to time cost and safety can be comprehensively considered, region of search size is carried out flexible
Regulation is to be bonded actual compromise in time-consuming make between algorithm performance of insertion.
Detailed description of the invention
Fig. 1 is secret information insertion schematic diagram of the invention；
Fig. 2 is that secret information of the invention extracts schematic diagram；
Fig. 3 is video interprediction encoding frame flow diagram；
Fig. 4 is motion vector (mvx,mvy) and its contiguous range in relative position schematic diagram between motion vector；
Fig. 5 is secret information insertion flow chart of the invention；
Fig. 6 is that secret information of the invention extracts flow chart.
Specific embodiment
Below by specific embodiment and in conjunction with attached drawing 5-6, the invention will be further described.
Before carrying out the insertion of secret information steganography using the present invention, the data that can be first embedded in needs are encrypted, and are obtained
To random binary data stream.It is proposed by the present invention based on motion vector local optimality keep video steganography method it is secret
Confidential information is embedded in process as shown in figure 5, concrete operations details is as follows:
1) video frame for needing to carry out steganography insertion is obtained, for each motion vector (mv in the framex,mvy), it obtains
Lowest bit position, that is, LSB (mv of the sum of its horizontal and vertical componentx+mvy), and all calculated results are formed into binary system carrier
Vector x；
2) for each motion vector MV in the video frame0,0=(mvx,mvy), generation is disturbed according to preset motion vector
Valence computation model obtains its steganography insertion cost: according to preset motion vector local optimality criterion, counting MV0,0It is adjacent
Four motion vector MV within the scope of domain-1,0=(mvx-1,mvy)、MV1,0=(mvx+1,mvy)、MV0,-1=(mvx,mvy- 1) and
MV0,1=(mvx,mvy+ 1) have the quantity N (0≤N≤4) of the motion vector of local optimality in, and be based on rate-distortion optimization mould
It includes MV that type, which calculates,0,0The RDO functional value J of five motion vectors insideMV(MV∈{MV0,0,MV-1,0,MV1,0,MV0,-1,
MV0,1), and then MV is calculated0,0Steganography be embedded in cost
Wherein, Ω={ MV-1,0,MV1,0,MV0,-1,MV0,1, and just by the steganography of each motion vector insertion cost composition
Real vector Γ；
3) private communication channel and the thereto secret letter of insertion are established using steganography code Syndrome Trellis Code (STC)
Breath.The load factor α and parameter matrix appointed in advance according to communicating pair firstEven-odd check square needed for establishing STC
Battle array ΗSTC, then using vector x obtained by step 1) as carrier needed for STC, and vector Γ obtained by step 2) is combined, using STC structure
It produces and meets ΗSTCX'=m and make overall steganography insertion cost D (x, x', Γ) the smallest x', wherein m is indicated by hidden
The secret information of channel insertion, x' indicates the carrier vector after steganography is embedded in, if the binary number on x and x' corresponding position
Difference then illustrates that corresponding original motion vector needs to modify, on the contrary then do not need to modify；
4) according to step 3) as a result, for the motion vector MV for needing to modify each in the video frame0,0=(mvx,
mvy), it modifies according to preset optimal candidate motion vector sought algorithm to it: by MV0,0Four fortune in contiguous range
Moving vector MV-1,0=(mvx-1,mvy)、MV1,0=(mvx+1,mvy)、MV0,-1=(mvx,mvy- 1) and MV0,1=(mvx,mvy+1)
In local optimum motion vector form set Ψ, if Ψ non-empty, the smallest local optimum of RDO functional value is selected in Ψ to move
Vector is as final modification as a result, if Ψ is sky, i.e. MV-1,0,MV1,0,MV0,-1,MV0,1Do not have local optimality, then delimit
One candidate motion vector region of search (its scale need according to practical computing resource and algorithm institute's allowed time at
This determination) and all local optimum motion vectors in the region are formed into set Ψ, and then select RDO functional value in Ψ minimum
Local optimum motion vector as final modification result.
5) video frame is encoded according to the result of step 4), that is, completes to be embedded in the secret information of private communication channel；
6) according to the frame coded sequence of video file, step 1) is repeated to 5) until secret information is complete to remaining every frame
Portion's insertion finishes, and obtains final steganography video file.
The secret information of the video steganography method proposed by the present invention kept based on motion vector local optimality extracts stream
Journey is as shown in fig. 6, concrete operations details is as follows:
1) video frame for needing to extract secret information is obtained, for each motion vector (mv in the framex,mvy), it obtains
Lowest bit position, that is, LSB (mv of the sum of its horizontal and vertical componentx+mvy), and all calculated results are formed into binary vector
x；
2) secret information in private communication channel is extracted.The load factor α and ginseng appointed in advance according to communicating pair first
Matrix numberParity matrix Η needed for establishing Syndrome Trellis Code (STC)STC, reuse step 1) institute
It obtains vector x and calculates ΗSTCX, calculated result are the secret information bit extracted；
3) according to the frame decoding sequence of video file, step 1) is repeated to 2) until secret information is complete to remaining every frame
It extracts and finishes in portion.
It can be seen that in insertion process of the invention from the above specific embodiment, according to the needs of application scenarios, both
Different motion vector disturbance cost computation models can be customized to influence execution speed, safety and the steganography of steganographic algorithm
The compressed encoding performance (code rate, fidelity etc.) of video, can also adjust the region of search size of optimal candidate motion vector with
Compromise between algorithm performance and insertion time-consuming, therefore the present invention has stronger adaptability and flexibility.
It is a kind of highly concealed type motion vector field video steganography method in order to which it is proposed by the present invention to protrude explanation, use is following
Configuration prepares video sample collection and carries out steganalysis experiment:
1) YUV sequence: it is 300 frames that by internet, to collect to obtain 50 resolution ratio, which be CIF (352 × 288), length,
YUV420 sequence.
2) video encoder: using x264 open source video encoder preparation compression video sample, in order to reduce time overhead,
Setting coding class is basic class Baseline.
3) compress video parameter: bit rate (bitrate) is 0.5mb/s, 1mb/s, 3mb/s or 10mb/s, and frame per second is
30fps。
4) steganography parameter: the load factor (payload) of carrier video is 0, and the load factor of steganography video is 0.25 or 0.5.
5) prepared by carrier video sample collection: according to the above configuration instruction, 4 carriers are prepared in 50 YUV420 files
Video sample collection is denoted as 0.5mb/s_0,1mb/s_0,3mb/s_0,10mb/ according to the form of " bitrate_payload " respectively
s_0。
6) prepared by steganography video sample collection: according to the above configuration instruction, 8 steganography are prepared in 50 YUV420 files
Video sample collection, is denoted as 0.5mb/s_0.25,0.5mb/s_0.5,1mb/s_0.25,1mb/s_0.5,3mb/s_0.25 respectively,
3mb/s_0.5,10mb/s_0.25,10mb/s_0.5.
7) steganalysis configures: extracting AoSO feature to carrier video sample collection and steganography video sample collection, is divided into 8 groups of spies
Collection closes and carries out steganalysis respectively, i.e., { 0.5mb/s_0,0.5mb/s_0.25 }, { 0.5mb/s_0,0.5mb/s_0.5 },
{ 1mb/s_0,1mb/s_0.25 }, { 1mb/s_0,1mb/s_0.5 }, { 3mb/s_0,3mb/s_0.25 }, { 3mb/s_0,3mb/s_
0.5 }, { 10mb/s_0,10mb/s_0.25 }, { 10mb/s_0,10mb/s_0.5 }；It is any to select for each of the above characteristic set
It takes the feature of wherein 40 pairs of videos for classifier training, obtains the support vector machines based on Gaussian kernel and polynomial kernel
(Support Vector Machine, SVM), the feature of remaining 10 pairs of videos repeat this process 100 for steganography classification judgement
It is secondary and the data obtained is averaged.
According to the above experimental configuration, the results are shown in Table 1 for gained steganalysis, it can be seen that final average detected is just
True rate is all very low (only one group of accuracy is more than 60%), and therefore, the method for the present invention can be fully against current motion vector domain
The attack of most effective steganalysis feature AoSO, is a kind of highly concealed type video steganography method, especially suitable for safety
More demanding covert communications scene.
Table 1 carries out the average detected accuracy of steganalysis using AoSO feature to the proposed method of the present invention
The above embodiments are merely illustrative of the technical solutions of the present invention rather than is limited, the ordinary skill of this field
Personnel can be with modification or equivalent replacement of the technical solution of the present invention are made, without departing from the spirit and scope of the present invention, this
The protection scope of invention should subject to the claims.
Claims (8)
1. a kind of video steganography method kept based on motion vector local optimality, which is characterized in that the insertion of secret information
Process the following steps are included:
1) in current video frame, to each motion vector (mvx,mvy), calculate LSB (mvx+mvy) binary number 0 or 1 is obtained,
And write all calculated results as vector x, wherein LSB expression takes lowest bit bit manipulation；
2) using gained vector x as the carrier of private communication channel, according to the working principle of the steganography code of selection, according to communicating pair thing
The parameter first arranged generates parity check matrix H, and combines preset motion vector disturbance cost computation model ultimately constructed out
Meet the x' of Hx'=m, wherein m indicates secret information to be embedded in the private communication channel, x' indicate modified carrier to
Amount, Hx' expression parity check matrix H and vector x ' be multiplied；
3) motion vector for needing to modify in current video frame is determined according to step 2)；
4) current video frame is encoded according to result obtained as above, and is distinguished according to preset optimal candidate motion vector sought algorithm
Each motion vector to be modified is substituted for corresponding optimal candidate motion vector, and then completes to believe the secret of private communication channel
Breath insertion；
5) according to the frame coded sequence of video file, step 1) is repeated to 4) until secret information is all embedding to remaining every frame
Enter to finish, obtains final steganography video file.
2. the method as described in claim 1, which is characterized in that further include the extraction process of secret information, its step are as follows:
1) current video frame is decoded, to each motion vector (mvx,mvy), calculate LSB (mvx+mvy) binary number 0 or 1 is obtained,
And write all calculated results as vector x, wherein LSB expression takes lowest bit bit manipulation；
2) according to the working principle of the steganography code of selection, parity matrix is generated according to the parameter that communicating pair is appointed in advance
H, calculates the product Hx of parity check matrix H and vector x, and acquired results are the secret information bit extracted；
3) according to the frame decoding sequence of video file, step 1) is repeated to 2) until secret information is all mentioned to remaining every frame
It takes complete.
3. method according to claim 1 or 2, which is characterized in that the steganography code is STC steganography code.
4. method according to claim 1 or 2, which is characterized in that the preset motion vector disturbs cost computation model
Are as follows:
To motion vector MV0,0=(mvx,mvy) both horizontally and vertically upper 4 adjacent motion vector, that is, MV-1,0=(mvx-
1,mvy)、MV1,0=(mvx+1,mvy)、MV0,-1=(mvx,mvy- 1) and MV0,1=(mvx,mvy+ 1) local optimality is carried out to sentence
It is fixed, if there is N number of motion vector to have local optimality in them, wherein 0≤N≤4, then to motion vector MV0,0It is embedding to carry out steganography
The disturbance cost entered are as follows:
Wherein, Ω={ MV-1,0,MV1,0,MV0,-1,MV0,1,Indicate motion vector MV0,0RDO functional value, JMVIndicate five
The RDO functional value of a motion vector, MV ∈ { MV0,0,MV-1,0,MV1,0,MV0,-1,MV0,1, i.e. the disturbance cost of motion vector takes
Certainly in the variation degree of the quantity of local optimum motion vector and RDO functional value in its contiguous range.
5. method as claimed in claim 4, which is characterized in that set and need the motion vector modified as MV0,0=(mvx,
mvy), then the preset optimal candidate motion vector sought algorithm are as follows:
A) respectively to MV0,0Four motion vector MV in contiguous range-1,0=(mvx-1,mvy)、MV1,0=(mvx+1,mvy)、
MV0,-1=(mvx,mvy- 1) and MV0,1=(mvx,mvy+ 1) local optimum sex determination is carried out；
B) by MV-1,0, MV1,0, MV0,-1And MV0,1In local optimum motion vector composition candidate motion vector set close Ψ, if Ψ
Non-empty then gos to step e), otherwise continues to execute step c)；
C) region of search of a motion vector delimited, size and final search time are directly proportional, according to practical computing resource
Or ability is chosen；For any candidate motion vector MV'=(mv in the regionx',mvy'), meet LSB (mvx+mvy)
≠LSB(mvx'+mvy')；
D) for each candidate motion vector in the region of search, local optimum sex determination is carried out respectively, and determination is wherein owned
The motion vector of local optimum, and by their composite set Ψ；
E) for each local optimum motion vector in Ψ, its RDO functional value is calculated, selection possesses minimum RDO functional value
Local optimum motion vector is as final modification result.
6. method as claimed in claim 5, which is characterized in that the method for the local optimum sex determination are as follows:
For motion vector MV0,0=(mvx,mvy), if its prediction block being directed toward is Pred (MV0,0), by inter prediction encoding
Obtained reconstructed block is Rec (MV0,0)；Calculating matrix Pred (MV0,0) and Rec (MV0,0) between absolute error and, i.e. sad0,0
=SAD (Rec (MV0,0),Pred(MV0,0))；Similarly, for MV0,08 adjacent motion vector MV of surroundingi,j=(mvx+i,mvy
+ j), wherein -1≤i, j≤1, respectively obtain the prediction block Pred (MV that they are directed towardi,j), and calculate sadi,j=SAD (Rec
(MV0,0),Pred(MVi,j))；If sadi,jIn minimum value be sad0,0, then claim motion vector MV0,0Has local optimality.
7. method as claimed in claim 6, which is characterized in that the telescopiny using the secret information of STC steganography code includes
Following steps:
1) in current video frame, by each motion vector (mvx,mvy) horizontal and vertical component summation after take lowest bit
Position, i.e. LSB (mvx+mvy), binary number 0 or 1 is obtained, and all calculated results are formed into binary system carrier vector x；
2) in current video frame, for each motion vector MV, cost computation model meter is disturbed according to preset motion vector
Calculation obtains its steganography disturbance cost Cost (MV), and all calculated results are formed vector Γ；
3) according to the working principle of steganography code STC, first according to the preassigned load factor α of communicating pair and parameter matrixStructure
Build parity matrix ΗSTC, reconstruct out and meet ΗSTCX'=m and make steganography insertion cost D (x, x', Γ) reach most
Small x', wherein m indicates that secret information to be embedded in private communication channel, x' indicate to be repaired after being embedded in by private communication channel steganography
The carrier vector changed；If the binary number on x and x' corresponding position is not identical, then it represents that corresponding original motion vector needs are repaired
Change, then indicates not needing to modify if they are the same；
4) according to the modification of carrier vector in private communication channel as a result, for each motion vector for needing to modify, according to preset
Optimal candidate motion vector sought algorithm be substituted for the smallest local optimum of RDO functional value in candidate motion vector move to
Amount；
5) current video frame is encoded according to result obtained as above to complete the steganography of secret information and be embedded in；
6) according to the frame coded sequence of video file, step 1) is repeated to 5) until secret information is all embedding to remaining every frame
Enter to finish, obtains final steganography video file.
8. the method for claim 7, which is characterized in that the extraction process using the secret information of STC steganography code includes
Following steps:
1) current video frame is decoded, to each motion vector (mvx,mvy) calculate LSB (mvx+mvy) binary number 0 or 1 is obtained, and
Write all calculated results as binary vector x；
2) secret information in private communication channel is extracted: according to the working principle of STC, according to the preassigned load factor of communicating pair
α and parameter matrixConstruct parity matrix ΗSTC, calculate ΗSTCX, acquired results are the secret information ratio extracted
It is special；
3) according to the frame decoding sequence of video file, step 1) is repeated to 2) until secret information is all mentioned to remaining every frame
It takes complete.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510184721.7A CN104853215B (en) | 2015-04-17 | 2015-04-17 | The video steganography method kept based on motion vector local optimality |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510184721.7A CN104853215B (en) | 2015-04-17 | 2015-04-17 | The video steganography method kept based on motion vector local optimality |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104853215A CN104853215A (en) | 2015-08-19 |
| CN104853215B true CN104853215B (en) | 2018-12-28 |
Family
ID=53852514
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510184721.7A Expired - Fee Related CN104853215B (en) | 2015-04-17 | 2015-04-17 | The video steganography method kept based on motion vector local optimality |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104853215B (en) |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105979269B (en) * | 2016-06-03 | 2019-06-04 | 中国科学院信息工程研究所 | Motion vector field video steganography method based on novel insertion cost |
| CN110149457B (en) * | 2019-05-08 | 2021-05-04 | 中山大学 | Multi-objective optimization H.264 video steganography method with constraint conditions |
| CN110502646A (en) * | 2019-08-02 | 2019-11-26 | 广州多益网络股份有限公司 | A kind of multimedia file purification method, device, storage medium and terminal device |
| CN112312138B (en) * | 2020-08-17 | 2022-04-19 | 中国科学技术大学 | Video steganography method for inhibiting inter-frame residual signal distortion propagation |
| CN112738355B (en) * | 2020-12-24 | 2021-10-26 | 华南理工大学 | Non-degraded HEVC video steganography method capable of resisting deep learning network detection |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101583033A (en) * | 2009-06-05 | 2009-11-18 | 中山大学 | Method for protecting H.264 video data by using robust watermarks |
| CN102685500A (en) * | 2012-05-08 | 2012-09-19 | 合肥工业大学 | H.264 large-capacity information hiding method based on information variable-length packet structure |
| US20130011004A1 (en) * | 2011-07-08 | 2013-01-10 | Frederic Lefebvre | Method for watermark detection using reference blocks comparison |
| CN103108188A (en) * | 2013-03-01 | 2013-05-15 | 武汉大学 | Video steganalysis method based on partial cost non-optimal statistics |
-
2015
- 2015-04-17 CN CN201510184721.7A patent/CN104853215B/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101583033A (en) * | 2009-06-05 | 2009-11-18 | 中山大学 | Method for protecting H.264 video data by using robust watermarks |
| US20130011004A1 (en) * | 2011-07-08 | 2013-01-10 | Frederic Lefebvre | Method for watermark detection using reference blocks comparison |
| CN102685500A (en) * | 2012-05-08 | 2012-09-19 | 合肥工业大学 | H.264 large-capacity information hiding method based on information variable-length packet structure |
| CN103108188A (en) * | 2013-03-01 | 2013-05-15 | 武汉大学 | Video steganalysis method based on partial cost non-optimal statistics |
Non-Patent Citations (1)
| Title |
|---|
| 《一种隐写分析盲性的评价及提高方法》;孙磊等;《计算机应用与软件》;20120930;第29卷(第9期);7-9,36 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104853215A (en) | 2015-08-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104853215B (en) | The video steganography method kept based on motion vector local optimality | |
| Park | Edge-based intramode selection for depth-map coding in 3D-HEVC | |
| Yao et al. | Defining embedding distortion for motion vector-based video steganography | |
| Li et al. | CNN based post-processing to improve HEVC | |
| Egilmez et al. | Graph-based transforms for inter predicted video coding | |
| TW201220855A (en) | Image encoding method and apparatus, image decoding method and apparatus, and programs therefor | |
| CN107197297B (en) | Video steganalysis method for detecting steganalysis based on DCT coefficient steganalysis | |
| CN105933711B (en) | Neighborhood optimum probability video steganalysis method and system based on segmentation | |
| CN108965887B (en) | Video information hiding method and device based on inter-block decoupling | |
| CN106162195A (en) | A kind of 3D HEVC deep video information concealing method based on single depth frame internal schema | |
| CN106664411A (en) | Motion-compensated partitioning | |
| CN109819260A (en) | Video steganography method and device based on the fusion of multi-embedding domain | |
| CN105979269B (en) | Motion vector field video steganography method based on novel insertion cost | |
| CN105915916B (en) | Video steganalysis method based on the estimation of motion vector distortion performance | |
| Wang et al. | A fast intra prediction algorithm for 360-degree equirectangular panoramic video | |
| CN107155112A (en) | A kind of compressed sensing method for processing video frequency for assuming prediction more | |
| WO2015093909A1 (en) | Method and apparatus for predicting video signal using predicted signal and transform-coded signal | |
| CN106101714A (en) | One and the tightly coupled H.264 video information hiding method of compression encoding process | |
| Fang et al. | 3dac: Learning attribute compression for point clouds | |
| Fu et al. | Efficient depth intra frame coding in 3D-HEVC by corner points | |
| Saberi et al. | An efficient data hiding method using the intra prediction modes in HEVC | |
| CN108353193B (en) | Method and apparatus for processing video data based on multiple graph-based models | |
| CN107040786A (en) | A kind of H.265/HEVC video steganalysis method adaptively selected based on space-time characteristic of field | |
| Liu et al. | Adaptive video data hiding with low bit-rate growth based on texture selection and ternary syndrome-trellis coding | |
| Jia et al. | A layered embedding-based scheme to cope with intra-frame distortion drift in IPM-based HEVC steganography |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| EXSB | Decision made by sipo to initiate substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20181228 Termination date: 20190417 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |