CN101582157A - Adaptive spatial steganographic method based on adjacent pixel difference - Google Patents

Abstract

The invention provides an adaptive spatial steganographic method based on adjacent pixel difference. The method comprises the following two steps: embedding secret information and extracting the secret information. The method takes three continuous adjacent pixels in an image as an embedding unit and reaches the goal of hiding information by changing the value of the middle pixel in each unit. During information embedding, a steganographic algorithm always mains the size relationship of each pixel in the embedding unit, and adaptively adjusts the capacity of embedding information through a parameter according to the content of the image. Experiments show that the adaptive spatial steganographic method has higher safety and adaptive ability compared with the prior steganographic method based on pixel difference. Therefore, the adaptive spatial steganographic method has great practical significance to application fields such as covert communication and the like.

Description

A kind of self-adapting airspace steganography method based on neighbor difference

Technical field

The invention belongs to multi-media information security, the secret communication field.Be specifically related to a kind of can the embedding and the high spatial domain steganography method of security performance according to the image content information self-adaptation.

Background technology

Along with digital multimedia, the constantly universal and development of network technology mutually, the information steganography Study on Technology also is subjected to the attention of scientific research personnel and mechanism of the military etc. gradually.At present, the steganography method of Digital Image Data can be divided into latent the writing with frequency field is latent in spatial domain and writes two kinds.

Based on latent the writing of PVD (Pixel-Value Differencing) is a kind of typical spatial domain steganography method.Its main thought is to determine to embed what of secret information according to the difference of image neighbor: regional many embeddings information that neighbor difference is big, difference is little then to embed less.Though the steganography method based on PVD has certain adaptivity, on the method for present announcement, the right embedding capacity of each neighbor still depends on an artificial default parameter, and does not rely on picture material itself; And, even these methods are that the information embedding has also been carried out in 0 zone at the pixel differences, mean promptly that also the smooth region of image also is used for hiding of information, latently like this write afterwards that visual quality for images can reduce.The more important thing is, tend to destroy the magnitude relationship between neighbor in the image based on the PVD method at present,, can make that like this its security performance reduces greatly through a large amount of description of tests.

Summary of the invention

The present invention is directed to defectives such as existing method adaptive ability is poor, security performance is bad, design a spatial domain steganography method that adaptive ability is strong and safe, transmit with the secret that realizes information.

In order to realize the foregoing invention purpose, the technical scheme of employing is as follows:

A kind of self-adapting airspace steganography method based on neighbor difference comprises the embedding of secret information and extracts two steps, and the embedding of described information comprises following substep: the rotation of (11) image block; (12) view data is reset and grouping; (13) the embedding capacity of estimated image; (14) revising embedding unit intermediate pixel realization information embeds; (15) image heavily rotates the embedding with threshold value T.

The concrete grammar of described step (11) image block rotation is: at first testing image I (establish its size for M * N) is decomposed into the square of zero lap Bz * Bz size, be made as Blk (j), require Bz=3k, k ∈ N, j=1,2..., S, $S=\left[\frac{M}{\mathrm{Bz}}\right]\left[\frac{N}{\mathrm{Bz}}\right]$ Total number of presentation video piecemeal.Utilize first key to generate S angle at random then, wherein the set of the value of angle is [0,90,180,270], utilizes the angle that generates that each piecemeal in the image is turned clockwise;

Described step (12) view data is reset with the concrete grammar that divides into groups: the image that obtains in the step (11) is carried out the rearrangement of Zigzag by the order of row major, obtain a row vector.And should be divided into the embedding unit of 1 * 3 size the vector zero lap.Simultaneously, two algorithm parameter T of initialization and k, wherein T reflects the difference between adjacent two pixels, and k reflects that the maximum of single pixel embeds bit number;

The concrete grammar of the embedding capacity of described step (13) estimated image is: embed unit [g for each _i, g _I+1, g _I+2], calculate d ₁=g _I+1-g _i, d ₂=g _I+2-g _I+1, the pixel g in the middle of limiting by following four kinds of situations then _I+1Variable scope

:

Situation 1:g _I+1＜g _iAnd g _I+1＜g _I+2

if?|d ₁|＞T?and?|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[0,...,\min (g_i-T-1,g_{i+2}-T-1)]$

if?|d ₁|＞T?and?|d ₂|≤T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_{i+2}-T,0),...,\min (g_i-T-1,g_{i+2}-1)]$

if?|d ₁|≤T?and?|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_i-T,0),...,\min (g_{i+2}-T-1,g_i-1)]$

Situation 2:g _I+1＞g _iAnd g _I+1＞g _I+2

if?|d ₁|＞T?and?|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_i+T+1,g_{i+2}+T+1),...,255]$

if?|d ₁|＞T?and?|d ₂|≤T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_i+T+1,g_{i+2}+1),...,\min (g_{i+2}+T,255)]$

if?|d ₁|≤T?and?|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_{i+2}+T+1,g_i+1),...,\min (g_i+T,255)]$

Situation 3:g _i〉=g _I+1〉=g _I+2

if?|d ₁|＞T?and?|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}[g_{i+2}+T+1,...,g_i-T-1]$

if?|d ₁|＞T?and?|d ₂|≤T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[g_{i+2},...,\min (g_{i+2}+T,g_i-T-1)]$

if?|d ₁|≤T?and?|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_{i+2}+T+1,g_i-T),...,g_i]$

Situation 4:g _i≤ g _I+1≤ g _I+2

if?|d ₁|＞T?and?|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[g_i+T+1,...,g_{i+2}-T-1]$

if?|d ₁|＞T?and?|d ₂|≤T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_{i+2}-T,g_i+T+1),...,g_{i+2}]$

if?|d ₁|≤T?and?|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[g_i,...,\min (g_{i+2}-T-1,g_i+T)]$

Calculate then Wherein

The expression set Element number, if n＞0, then expression embeds unit [g _i, g _I+1, g _I+2] the embedding capacity be the n bit, all embed the embedding total volume of unit among the statistical picture I, if this capacity is greater than the length that will embed scale-of-two secret information M, then change step (14), otherwise T is reduced to T-1 with parameter, change the embedding capacity that step (13) reappraises image, if T reduces to 0, then presentation video I does not have the given secret information M of enough space embeddings.

Theoretical proof: as middle pixel g _I+1In the zone

During middle the variation, g _i, g _I+1, g _I+2Remain unchanged with the magnitude relationship of T.As embed preceding g _i≤ g _I+1≤ g _I+2, | g _I+1-g _i|＞T has g after then embedding equally _i≤ g ' _I+1≤ g _I+2, | g ' _I+1-g _i|＞T, as long as ${g^{\′}}_{i+1}\∈{\mathrm{range}}_{{g^{\′}}_{i+1}}$ In addition, can see the zone from above formula

Scope be not rely on original g _I+1Value, it only depends on g _i, g _I+2With parameter T.Therefore, in the leaching process of secret information, can obtain identical zone

Thereby, guarantee correctly to extract secret information.

Described step (14) revise to embed the concrete grammar that intermediate pixel realization information in unit embeds: each of the order traversing graph picture that generates according to second key embeds unit, as to [g _i, g _I+1, g _I+2] realize to embed, at first according to step (13) but the intermediate pixel mobility scale of determining

And embeddable information capacity n, order is extracted the n Bit data and is transferred decimal number b to from secret information M, changes g by following formula _I+1Be g ' _I+1Realize embedding:

${g^{\′}}_{i+1}=\underset{e}{\arg \min }\left\{\right|e-g_{i+1}\left|\right||e-g_i|\≡b\left(\mathrm{mod}{2}^n\right),e\∈{\mathrm{range}}_{{g^{\′}}_{i+1}}\}$

Repeating this step all is embedded into until all secret informations;

The concrete grammar that described step (15) image heavily rotates with the embedding of threshold value T is: each image block after will handling through step (14) according to first key is rotated counterclockwise, and with the parameter T of step (13) be embedded into one default, be not used in the image-region that secret information embeds, generate the latent back image of writing.

The extraction of described secret information comprises following substep: the rotation of (21) image block; (22) view data is reset and grouping; (23) extraction respectively embeds the unit secret information.

Described step (21) image block rotation and step (22) view data reset with the concrete grammar of grouping and secret information embedding step (11), (12) in corresponding way be consistent;

Described step (23) is specific as follows to the extraction operation of secret information: at first go out parameter T from the default region extraction of image, respectively embed the unit according to second cipher key sequence traversal then, all be extracted until all secret informations.If pending unit is [g _i, g ' _I+1, g _I+2], if | g ' _I+1-g _i|≤T and | g _I+2-g ' _I+1|≤T, then skip to the next unit that embeds, calculate intermediate pixel g ' otherwise embed step (13) by information _I+1Variation range , and calculate

Obtain the embedded quantity n of this embedding unit, by calculating b ≡ | g ' _I+1-g _i| (mod2 ⁿ) obtain being embedded in the decimal representation of this unit secret information, at last b is transferred to the secret information that binary number obtains the n bit.

The present invention has mainly utilized the magnitude relationship between continuous three pixels in the image, and realizes the embedding of secret information by the pixel value in the middle of changing.Can realize the embedding of secret information according to the content-adaptive of image ground own, and in telescopiny, can keep the big or small ordering relation between the original pixel of image as far as possible, make that latent imperceptibility and the security of writing the back image is all higher, thereby crucial effect is played in the secret transmission of information.

Description of drawings

Fig. 1 is embedding, the extraction schematic flow sheet of secret information of the present invention, and wherein (a) embeds process flow diagram for information, (b) is the information extraction process flow diagram;

Fig. 2 does not have repetition piecemeal and the piecemeal synoptic diagram that turns clockwise at random for image;

Fig. 3 image is reset synoptic diagram with row major order Zigzag;

Fig. 4 image is reset the back and is embedded the unit packet synoptic diagram;

Fig. 5 embeds in the unit four kinds of magnitude relationship synoptic diagram of three pixels and parameter T;

Fig. 6 image does not have the repetition piecemeal and piecemeal is rotated counterclockwise synoptic diagram at random;

Fig. 7 is under different steganography methods, and the latent difference synoptic diagram of writing back image and original image: wherein (a) is original test pattern; (b) for adopting the present invention; (c) for adopting Original PVD method; (d) for adopting the IPVD method; (e) for adopting the PVD-LSB method; (f) for adopting the Adaptive-Edge method;

Fig. 8 is for contrasting Adaptive-Edge and the present invention visual effect at smooth region: (a) be to utilize the latent contour map of writing the back regional area of Adaptive-Edge method; (b) be to utilize contour map of the present invention;

Fig. 9 is under difference embedding capacity, the latent difference synoptic diagram of writing back image and original image: wherein (a) is 5%bpp; (b) be 10%bpp; (c) be 20%bpp; (d) be 30%bpp.

Embodiment

The present invention is further illustrated below in conjunction with accompanying drawing.

Shown in Figure 1 is the algorithm flow synoptic diagram of the embedding and the extraction of secret information of the present invention.

Wherein the telescopiny of secret information comprises step: the rotation of (11) image block; (12) view data is reset and grouping; (13) the embedding capacity of estimated image; (14) revising embedding unit intermediate pixel realization information embeds; (15) image heavily rotates the embedding with threshold value T.

Each step specifies as follows:

(11) image block rotation

Suppose that image to be detected is I, its size is M * N.As shown in Figure 2, at first I is decomposed into Bz * Bz (setting Bz=6) fritter Blk (j) in zero lap zone, $j=\mathrm{1,2},...,\left[\frac{M}{\mathrm{Bz}}\right]\left[\frac{N}{\mathrm{Bz}}\right],$ Utilize a key key1 to generate then Individual angle at random, the span of angle are 0,90,180 and 270, respectively each fritter are carried out the clockwise direction rotation.

(12) view data is reset and grouping

As shown in Figure 3, the image after handling through step (11) is carried out Zigzag with row major order reset, obtain a row vector.And as shown in Figure 4, its nothing is decomposed into the embedding unit of non-overlapping 1 * 3 size.Initiation parameter T and k set T=32, k=4.In step (15), need T is embedded into the latent a certain predeterminable area (or in pixel) of writing the back image, because T=32=2 ⁵, so this auxiliary information only needs the data volume of 5 bits, the method for embedding can be utilized methods such as simple LSB substitutes.

(13) the embedding capacity of estimated image

To each embedding unit of image, as [g _i, g _I+1, g _I+2], estimate its embeddable capacity.Because embedding principle of the present invention is: only change pixel g _I+1Size, and require g _i, g _I+1, g _I+2And the magnitude relationship between the parameter T remains unchanged before and after embedding secret information., as shown in Figure 5, reach and the T relation for this reason, obtained intermediate pixel g according to 4 kinds of size orders between these three pixels _I+1But region of variation

, then by calculating

The embedding capacity that obtains this embedding unit is the n bit, adds up total embedding capacity that all embedding unit obtain image, if this capacity greater than the secret information M that will embed, then changes step (14) and carries out the embedding of secret information; Otherwise regulate parameter T=T-1, more embedded space in the release graphics picture changes total embedding capacity that (13) reappraise image again.If T reduces to 0, then key diagram looks like not have enough space embedding secret information M;

(14) revising embedding unit intermediate pixel realization information embeds

According to the embedding unit of key key2 order traversing graph picture, embed the embedding that unit neutral element size realizes secret information by changing.Embed the unit for each, as [g _i, g _I+1, g _I+2], at first determine the variable scope of intermediate pixel according to the embedding step (13) of summary of the invention secret information And embeddable bit number n, order is extracted the n Bit data and is transferred decimal number b to from secret information M then, presses following formula then to g _I+1Make amendment:

${g^{\′}}_{i+1}=\underset{e}{\arg \min }\left\{\right|e-g_{i+1}\left|\right||e-g_i|\≡b\left(\mathrm{mod}{2}^n\right),e\∈{\mathrm{range}}_{{g^{\′}}_{i+1}}\}$

Repeating (14) all is embedded into until all secret informations;

(15) image heavily rotates the embedding with threshold value T

To be rotated through each the piecemeal Blk (j) in the image of step (14) according to key key1, wherein consistent in Xuan Zhuan angle and the image block pre-service, but direction is opposite, as Fig. 2 and shown in Figure 6.And with step (13) estimated image embed the parameter T that obtains in the capacity step be embedded into one default, be not used in the image-region that secret information embeds, generate the image after latent the writing.

The extraction of secret information of the present invention comprises step: the piecemeal rotation of (21) image; (22) rearrangement of view data and grouping; (33) extract the secret information that embeds the unit.

The wherein rearrangement and the grouping of the piecemeal of step (21) image rotation and (22) view data is consistent with operating accordingly in the embedding step (11) of secret information and the step (12).

(23) extract the secret information that embeds the unit

At first extract parameter T, travel through in proper order according to key key2 then and respectively embed the unit, all be extracted until all secret informations from the latent back image predeterminable area of writing.If pending unit is [g _i, g ' _I+1, g _I+2], if | g ' _I+1-g _i|≤T and | g _I+2-g ' _I+1|≤T, illustrate that this unit does not have to hide Info, skip to and nextly embed the unit, embed step (13) calculating intermediate pixel g ' otherwise at first press secret information _I+1Variation range

And embedding capacity

By calculating b ≡ | g ' _I+1-g _i| (mod2 ⁿ), obtain the decimal representation b of n position secret information, at last b is transferred to the binary number bit stream and obtain embedding unit [g _i, g ' _I+1, g _I+2] secret information.

Below in conjunction with concrete experimental data the present invention is done further analysis explanation:

I. image collection:

In the present embodiment, utilized totally 3855 width of cloth natural images to do test, wherein image library comprises:

[NRCS]: 1543 width of cloth sizes are: 512 * 768 and 768 * 512

NRCS?Photo?Gallery(2005)：http://photogallery.nrcs.usda.gov/

[UCID]: 1338 width of cloth sizes are: 384 * 512 and 512 * 384

G.Schaefer，M.Stich，Ucid：an?uncompressed?color?image?database，in：Proceedings?of?SPIE?Electronic?Imaging，Storage?and?Retrieval?Methods?andApplications?for?Multimedia，Vol.5307，2003，pp.472-480。

And the size of taking voluntarily is 640 * 480 image 974 width of cloth.

Above all images all transfers gray level image to.

II. typical steganography method based on PVD:

【Original?PVD】：D.C.Wu，W.H.Tsai，A?steganographic?method?for?images?bypixel-value?differencing，Pattern?Recogniton?Lett.24(2003)1613-1626.

【IPVD】：X.Zhang，S.Wang，Vulnerability?of?pixel-value?differencingsteganography?to?histogram?analysis?and?modification?for?enhanced?security，PatternRecogniton?Lett.25(2004)331-339.

【PVD-LSB】：H.C.Wu，N.I.Wu，C.S.Tsai，M.S.Hwang，Imagesteganographic?scheme?based?on?pixel-value?differencing?and?LSB?replacementmethods，Proceeding?of?IEE?Inst.Elect.Eng.，Vis.Images?Signal?Process.152(5)(2005)611-615.

【Adaptive-Edge】：C.H.Yang，C.Y.Weng，S.J.Wang，H.M.Sun，Adaptivedata?hiding?in?edge?areas?of?images?with?spatial?LSB?domain?systems，IEEE?Trans.on?Information?Forensics?and?Security?3(3)(2008)488-497.

III. typical universal steganalysis feature:

【Shi78-D】：Y.Shi，G.Xuan，D.Zou，J.Gao，C.Yang，Z.Zhang，P.Chai，W.Chen，C.Chen，Image?steganalysis?based?on?moments?of?characteristic?functionsusing?wavelet?decomposition，prediction-error?image，and?neural?network，in：IEEEInt.Conf.on?Multimedia?and?Expo，2005.

【Farid72-D】：H.Farid，Detecting?hidden?messages?using?higher-orderstatistical?models，in：IEEE?Int.Conf.on?Image?Processing，Vol.2，2002，pp.II905-908.

【Moulin156-D?】：Y.Wang，P.Moulin，Optimized?feature?extraction?forlearning-based?image?steganalysis，IEEE?Trans.on?Information?Forensics?andSecurity?2(1)(2007)31-45.

【Li110-D】：B.Li，J.Huang，Y.Q.Shi，Textural?features?based?universalsteganalysis，in：Proceedings?of?the?SPIE?on?Security，Forensics，Steganography?andWatermarking?of?Multimedia，Vol.6819，2008，p.681912.

Subjective visual quality do is analyzed:

Shown in Figure 7 is to utilize different steganography methods, latent back image and the original image difference synoptic diagram write that obtains.Wherein, picture white represents that variation has taken place the pixel value size that embeds back image correspondence position.Can see that the present invention only embeds on the borderline region of image, thereby can keep the original most smooth region pixel value of image constant; Other location revision based on the PVD method then is each zone that is randomly dispersed in image.Even the difference between the image neighbor is 0, these methods have also been carried out the information embedding.But because the visual characteristic of human eye: the modification to the borderline region of image is insensitive, and is then responsive relatively to the change in image smoothing zone.Therefore, subjective vision of the present invention can be better than original method.

Shown in Figure 8 is to utilize tradition based on the latent contour map of writing the regional area of back image of PVD method (Adaptive-Edge) and the inventive method, can see that the present invention can keep the flatness in original zone.

Adaptive ability is analyzed:

Shown in Figure 9 is image is respectively at 5%bpp (bit per pixel), 10%bpp, and under 20%bpp and the 30%bpp embedding capacity, the latent synoptic diagram of writing difference between back image and the original image.Can see: main embedded location of the present invention is all on the border of image; And the inventive method can be according to the embedding capacity, utilize the boundary information of image adaptively, as under the 5%bpp situation, only embed between neighbor pixel centering, and keep peaceful skating area, other border (neighbor difference is less than T) domain information constant greater than T=32; When the embedding capacity increases, as 10%bpp, 20%bpp, 30%bpp, the value of parameter T can reduce (being respectively T=22,8,4) thereupon, means promptly that also more boundary information is utilized in the image.The experiment proved that this adaptive characteristic can increase the visual effect and the security thereof of image.

Safety Analysis:

Utilize various PVD steganography methods to generate the hidden image of different embedding capacity, 5%bppm is arranged, 10%bpp, 20%bpp and 30%bpp respectively.For each situation, wherein the data of half are as training set, and second half data are as test set.Choosing of characteristic of division utilized four kinds of typical general steganalysis methods, sorter then to adopt FLD (Fisher Linear Discriminate), and following table 1 has been listed the verification and measurement ratio under the various situations.Can see that verification and measurement ratio of the present invention is compared the reduction that has near 20% with other PVD method, that is to say that security of the present invention is better.

Table 1

Claims (8)

1, a kind of self-adapting airspace steganography method based on neighbor difference comprises the embedding of secret information and two steps of extraction of secret information, it is characterized in that the embedding of described secret information comprises following substep:

(11) image is carried out piecemeal, and each image block is carried out the clockwise angle rotation at random according to the angle that first key generates;

(12) postrotational image is arranged as a row vector by row major order, and this vector is divided into the embedding unit of 1 * 3 size in zero lap zone, two parameter T of initialization and k, wherein T reflects the difference between adjacent two pixels, and k reflects the maximum bit number that embeds of single pixel;

(13) the embedding capacity of estimated image embeds the unit to each, as [g _i, g _I+1, g _I+2], at first according to intermediate pixel g _I+1And the difference of former and later two pixels and and parameter T between relation, determine g _I+1Transformable scope

Pass through then

Obtain embedding unit [g with parameter k _i, g _I+1, g _I+2] embedding capacity n, if the total volume that image respectively embeds the unit is then revised parameter T less than the capacity that will embed secret information M, and estimated image capacity again, otherwise carry out next step;

(14) each of the order traversing graph picture that generates according to second key embeds unit, is [g as pending unit _i, g _I+1, g _I+2], according to step (13) but method is determined the mobility scale of pixel in the middle of it

And embeddable information capacity n, order is extracted the n Bit data and is transferred decimal number b to from secret information M then, changes g by following formula _I+1Be g ' _I+1Realize embedding:

${g^{\′}}_{i+1}=\underset{e}{\arg \min }\left\{\right|e-g_{i+1}\left|\right||e-g_i|\≡b\left(\mathrm{mod}{2}^n\right),e\∈{\mathrm{range}}_{{g^{\′}}_{i+1}}\}$

Repeating this step all is embedded into until all secret informations;

(15) angle that generates according to first key is rotated counterclockwise each image block, and step (13) is estimated that the parameter T that obtains is embedded in the image-region default, that be not used in the secret information embedding, generates the image that conceals after writing.

2, self-adapting airspace steganography method according to claim 1 is characterized in that the image block concrete operations of described step (11) are: at first the gray level image I of M * N size is decomposed into the image block Blk (j) of zero lap Bz * Bz size, Bz=3k, k ∈ N, j=1,2...S, wherein $S=\left[\frac{M}{\mathrm{Bz}}\right]\left[\frac{N}{\mathrm{Bz}}\right]$ Total number of presentation video piece.

3, self-adapting airspace steganography method according to claim 1 is characterized in that the anglec of rotation at random that is generated by first key of described step (11), and its span is [0 90 180 270].

4, self-adapting airspace steganography method according to claim 1 is characterized in that described step (13) determines intermediate pixel g _I+1Transformable scope

Calculate according to following four kinds of situations:

Situation 1:g _I+1＜g _iAnd g _I+1＜g _I+2

if|d ₁|＞T?and|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[0,...,\min (g_i-T-1,g_{i+2}-T-1)]$

if|d ₁|＞T?and|d ₂|≤T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_{i+2}-T,0),...,\min (g_i-T-1,g_{i+2}-1)]$

if|d ₁|≤T?and|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_i-T,0),...,\min (g_{i+2}-T-1,g_i-1)]$

Situation 2:g _I+1＞g _iAnd g _I+1＞g _I+2

if|d ₁|＞T?and|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_i+T+1,g_{i+2}+T+1),...,255]$

if|d ₁|＞T?and|d ₂|≤T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_i+T+1,g_{i+2}+1),...,\min (g_{i+2}+T,255)]$

if|d ₁|≤T?and|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_{i+2}+T+1,g_i+1),...,\min (g_i+T,255)]$

Situation 3:g _i〉=g _I+1〉=g _I+2

if|d ₁|＞T?and|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[g_{i+2}+T+1,...,g_i-T-1]$

if|d ₁|＞T?and|d ₂|≤T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[g_{i+2},...,\min (g_{i+2}+T,g_i-T-1)]$

if|d ₁|≤T?and|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_{i+2}+T+1,g_i-T),...,g_i]$

Situation 4:g _i≤ g _I+1≤ g _I+2

if|d ₁|＞T?and|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[g_i+T+1,...,g_{i+2}-T-1]$

if|d ₁|＞T?and|d ₂|≤T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[\max (g_{i+2}-T,g_i+T+1),...,g_{i+2}]$

if|d ₁|≤T?and|d ₂|＞T

${\mathrm{range}}_{{g^{\′}}_{i+1}}=[g_i,...,\min (g_{i+2}-T-1,g_i+T)]$

Wherein, d ₁=g _I+1-g _i, d ₂=g _I+2-g _I+1

5, self-adapting airspace steganography method according to claim 1 is characterized in that the definite unit [g of embedding of described step (13) _i, g _I+1, g _I+2] method that embeds capacity is: calculate

Wherein

The expression set

Element number, if n＞0, then expression embeds unit [g _i, g _I+1, g _I+2] the embedding capacity be the n bit.

6. self-adapting airspace steganography method according to claim 1, it is characterized in that described step (13) parameter modification and image volume revaluation method are: statistical picture respectively embeds total embedding capacity of unit, if this capacity is greater than the length that will embed scale-of-two secret information M, then change step (14), otherwise T is reduced to T-1 with parameter, change the embedding capacity that step (13) reappraises image, if T reduces to 0, then presentation video I does not have the given secret information M of enough space embeddings.

7, self-adapting airspace steganography method according to claim 1 is characterized in that the extraction of described secret information comprises following substep:

(21) image is carried out piecemeal, each image block is done clockwise angle rotation at random according to the angle that first key generates;

(22) image is arranged with row major order obtained a row vector, and this row vector is divided into the embedding unit of 1 * 3 size in zero lap zone;

(23) extract parameter T from the latent predeterminable area of writing the back image, the order that generates according to second key travels through and extracts the secret information that respectively embeds the unit then, all is extracted until all secret informations.

8, self-adapting airspace steganography method according to claim 7 is characterized in that described step (23) is specific as follows to the extraction operation that embeds secret information in the unit:

If the embedding unit [g of information to be extracted _i, g ' _I+1, g _I+2], if | g ' _I+1-g _i|≤T and | g _I+2-g ' _I+1|≤T then skip to next embedding unit, otherwise (13) calculates intermediate pixel g ' at first set by step _I+1Variation range

Calculate then

Obtain g ' _I+1Embedded quantity be n, calculate b ≡ again | g ' _I+1-g _i| (mod2 ⁿ), obtain the decimal representation b of n position secret information, at last b is transferred to binary number and obtain embedding unit [g _i, g ' _I+1, g _I+2] secret information.

Images