CN101304522A - Considerable information hide method using JPEG2000 compression image as carrier - Google Patents

Abstract

The invention relates to a high capacity information hiding method using JPEG2000 compressing images as the carrier. The operation steps are: applying wavelet decomposition, wavelet coefficients quantification and bit-plane coding to images, and determining the truncation position of the code flow of all coding blocks and the lowest bit-plane of the embeddable information by rate-distortion optimization; primarily choosing a wavelet coefficient which is more than the threshold value as the embedding point, and making a redundancy estimation and adjusting the embedding point and the embedding strength; embedding the synchronous information and the hidden information into the quantified wavelet coefficient in an order of the bit plane from the lowest to the highest starting from the lowest embeddable bit-plane; making a second bit-plane coding and organizing the code flow after embedding the information. The high capacity information hiding method is based on a second coding strategy to determine the embedding position and improves the information hiding volume by self-adaptively choosing the embedding point and adjusting the embedding strength, The high capacity information hiding method has the advantages of being transparent completely for the standard JPEG2000 decoders and being able to be conveniently integrated in the JPEG2000 coders. Meanwhile, the code flow still can decode normally after being embedded with hiding information.

Description

A kind of is the considerable information hide method of carrier with the JPEG2000 compressed image

[technical field]:

The present invention relates to the digital picture is information security and the multimedia messages processing technology field that carrier carries out covert communications.

[background technology]:

The present information concealing technology is the important directions of digital information epoch information security field, the redundancy that Information Hiding Techniques utilizes Digital Media itself to be had, and human perceptual organ's physiology and psychological characteristics, to hide Info is hidden in the middle of the carrier data, and carrier signal can be text, rest image, video, audio frequency etc.Information hiding and encryption technology have marked difference, and by encrypting the content that can hide Info, the existence that then can hide Info by Information hiding combines Information hiding and can greatly improve the information security level with encryption technology.

When utilizing Information Hiding Techniques to realize covert communications, invisibility and hiding capacity are two important indicators, that is to say on the one hand to embed the abundant bit that hides Info in the carrier object, guarantee the validity of communication; On the other hand, information can not cause the carrier object appearance that perceptible variation takes place after embedding, and guarantees the crypticity of communication.

The method of hiding data in digital picture that is widely used at present is the LSB method of spatial domain, information is embedded in the least significant bit of each pixel value, the advantage of this method is simple and easy to do, hiding capacity is bigger, invisibility is good, but reliability is relatively poor, will be erased after the information via lossy compression method of embedding, so the LSB method of spatial domain can not be applied to the digital picture of common compressed format.All used the transition coding technology in the various image compression encoding international standards, the information of carrying out in transform domain embeds and can reasonablely address this problem.That use in the jpeg image compressed encoding international standard is discrete cosine transform (DCT), the entropy coding that the DCT coefficient quantization can't harm later on again, obtain final compressed bit stream, can realize that secret information embeds as long as selectively revise the low level that quantizes back DCT coefficient.JPEG2000 is the JPEG of a new generation of formulating of International Telecommunication Association (ITU), compares with old standard, and JPEG2000 has remarkable performance at aspects such as the scalability of reconstructed image quality, code stream and fault-tolerant abilitys.The JPEG2000 standard is based on wavelet transform (DWT), the information in above-mentioned DCT territory embeds strategy and can not be grafted directly in the JPEG2000 system, reason is in the JPEG2000 compression encoding process, quantize later wavelet coefficient through after the preliminary code stream of entropy coding generation, also to experience the step that a rate-distortion optimization blocks again and just can obtain final code stream, this just means if simply secret information is embedded into the low level that quantizes the back wavelet coefficient, after passing through the code stream truncation so, the secret information that embeds still can be destroyed, and makes the receiving terminal can't the complete extraction Content of Communication.

Spread spectrum often is introduced in the ability that common signals such as improving anti-lossy compression method among the information concealing method is handled means, at first the bit that hides Info that will embed is carried out spread processing, and then it is embedded in the carrier medium, receiving terminal utilizes related detecting method to obtain a result.This method relatively is fit to be applied in the digital watermark technology; functions such as digital copyright protecting are provided; and it is general and be not suitable for covert communications; this is because the spread spectrum preliminary treatment has greatly reduced Information Hiding Capacity; the testing result of spread spectrum digital watermark technology generally has only two kinds: " having " or " nothing " that is to say that it only provides the Information Hiding Capacity of a bit.

What the present invention realized is a kind of information concealing method that can embed bulk information in the JPEG2000 compressed image, and this research belongs to Tianjin natural science fund (No.06YFJMJC00700) and subsidizes research contents.This method can guarantee that embedded information remains intact in the lossy compression method process, thereby guarantee that receiving terminal correctly extracts the content of covert communications, this method has also made full use of the visual characteristic of human eye in addition, satisfy provide bigger under the prerequisite of invisibility Information Hiding Capacity to satisfy the needs of covert communications.

This partial content can be with reference to following documents and materials:

1.R.J.Anderson，and?F.A.P.Petitcolas，On?the?limits?of?steganography，IEEE?Journal?of?Selected?Area?inCommunications，1998，16.474-481

2.ISO/IEC?FCD?15444-1，JPEG2000?Part?1?Final?Committee?Draft?Version?1.0，2000

3.ISO/IEC?FCD?15444-2，JPEG2000?Part?2?Final?Committee?Draft，2000

[summary of the invention]:

The objective of the invention is to overcome the prior art above shortcomings, providing a kind of is the considerable information hide method that is applicable to covert communications of carrier with the JPEG2000 compressed image.

Provided by the invention is that the operating procedure of the considerable information hide method that is applicable to covert communications of carrier is as follows with the JPEG2000 compressed image:

The first,, original image is carried out multilayer wavelet decomposition, wavelet coefficient quantification, Bit-Plane Encoding, and rate-distortion optimization blocks according to the method for stipulating in the JPEG2000 standard.Draw the lowest order plane that wavelet coefficient in each encoding block can embed information according to the code stream disconnect position;

The second, tentatively choose greater than the wavelet coefficient of threshold value 16 as embedding point, result according to the redundancy estimation adjusts embedding point and embedment strength, removal is unsuitable for the embedding point of the information of embedding, and draws the information bit that can embed at most on each embedding point;

Three, the lowest order plane from embedding is embedded into synchronizing information and hiding Info in the wavelet coefficient after the quantification according to bit plane order from low to high;

Four, finish information and embed after, carry out Bit-Plane Encoding once more;

Five, block the result according to the rate-distortion optimization in the first step and organize final code stream.

The embedding point that hides Info and maximum embeddable information bit calculate by following redundant evaluation method:

First step utilization is handled wavelet coefficient from contrasting masking effect,

y _i＝sign(x _i)|x _i·Δ _i| ^α (1)

X wherein _iBe wavelet coefficient after the uniform quantization and the value after the following zero clearing of highest order, Δ _iBe the quantization step of current wavelet coefficient, 0＜α＜1, the following zero clearing of highest order is to obtain same result of calculation in order to allow built-in end and extraction hold, thereby guarantees that algorithm is reversible; y _iIt is the wavelet coefficient values of utilizing from contrasting after masking effect is handled;

Second step utilized the neighborhood masking effect to handle wavelet coefficient,

$z_i=\frac{y_i}{1+\left(a\underset{k\∈\mathrm{neighborhood}}{\mathrm{\Σ}}{\left|{\stackrel{\‾}{x}}_k\right|}^{\mathrm{\β}}\right)/\left|{\mathrm{\φ}}_i\right|}---\left(2\right)$

It is wavelet coefficient in the N*N scope at center that the neighborhood wavelet coefficient refers to current coefficient, | φ _i| the number of expression neighborhood wavelet coefficient; Parameter beta ∈ [0,1] is the strength control factor of masking effect, β and | φ _i| controlling the dynamics of adjusting embedment strength according to the neighborhood masking effect together; A=(1000/2 ^l) ^βBe the standardization factor, { 0,1,2,3,4} represents the level of resolution at wavelet coefficient place to l ∈;

Be the later neighborhood wavelet coefficient of uniform quantization, β gets the influence that less value can suppress the very big wavelet coefficient of minority range value in the neighborhood effectively, thus sharp edges district and texture area in can differentiate between images, z _iIt is the wavelet coefficient values of utilizing after the neighborhood masking effect is handled.For the assurance built-in end obtains same result of calculation with extracting to hold, also want zero clearing below the highest order of neighborhood wavelet coefficient, and only get the neighborhood wavelet coefficient participation calculating that range value is not less than threshold value 16.

The 3rd step, human eye is the most responsive to the noise of medium luminance area, sensitivity is along with the rising of brightness value or reduction and weaken, therefore increase by one by the weight coefficient that luminosity sensitivity determined, with l ∈ { 0,1,2,3,4} represents the level of resolution after the wavelet decomposition, and l=0 represents the highest level of resolution; With θ ∈ { 0,1,2,3} represents the subband direction, is followed successively by LL, LH, HL, HH, and wherein LL represents that level and vertical direction all are high frequency, LH represents horizontal direction low frequency, vertical direction high frequency, and HL represents horizontal direction high frequency, vertical direction low frequency, and HH represents that level and vertical direction are all high frequency; I _l ^θ(i, j) expression level of resolution l down, direction is the locational wavelet coefficient that the capable j of i is listed as in the subband of θ, then in any direction subband under the l level of resolution, coordinate is (i, the luminosity sensitivity weight coefficient Λ of wavelet coefficient j) (l, i, j) following calculating:

$\mathrm{\&Lambda;}(l,i,j)=\left\{\begin{array}{cc}2-L(l,i,j),& \mathrm{if\; L}(l,i,j)<1\\ L(l,i,j),& \mathrm{otherwise}\end{array}\right.---\left(3\right)$

Wherein median L (l, i, j) following calculating:

K is the progression of wavelet decomposition, and l=k is minimum level of resolution, wherein symbol

Expression is not more than the maximum integer of x,

It is a wavelet coefficient in the LL subband of lowest resolution level, coordinate is (i in any direction subband under it and the l level of resolution, j) wavelet coefficient correspondence identical image-region, because pixel value having been done amplitude before the wavelet decomposition is 128 DC level displacement, the dynamic range of pixel value is (128,127), so carry out normalization divided by 128 in the following formula

Obtain at last:

${z_i}^{\&prime;}=\frac{z_i}{\mathrm{\&Lambda;}(l,i,j)}---\left(5\right)$

Z ' _iIt is the wavelet coefficient values after handling according to luminosity sensitivity.Quantification redundancy on the wavelet coefficient is then estimated by following formula:

r _i＝x _i/z′ _i (6)

r _iIn fact reflected when using the uniform quantization method of the basic compressing and coding system of JPEG2000, the size of the quantification redundancy that exists on the current wavelet coefficient, under the binary system situation, every twice just expression has the redundancy of a bit, thereby just can embed hiding Info of a bit more.For the influence of control information embedding to picture quality, regulation has only the r of working as _iBe not less than at 2 o'clock and could embed and hide Info.When adjusting embedding point and embedment strength, at first get rid of r according to redundant estimation result _iIf embedding point less than 2 determines that according to following rule each embeds the bit number that hides Info that the some place can embed: then 2 ⁿ≤ r _i＜2 ^N+1, then this embedding point can embed the n bit and hides Info.

Each embeds a definite method that can embed the lowest order plane of information, after the process rate-distortion optimization obtains the code stream disconnect position of each encoding block, choose three coding passes all not intercepted (information completely just) lowest order plane, as the lowest order plane that can embed information.

To the extracting method of the hiding information of said method, the process following steps:

The first, the coding/decoding method according to the JPEG2000 standard code carries out code stream analyzing, entropy decoding, obtains the preceding wavelet coefficient of inverse quantization;

The second, the same principle primary election of built-in end embedded point during use and above-mentioned hidden method second went on foot;

Three, the information that has the bit plane of three coding passes in the code stream is complete.During the entropy decoding, find out the lowest order plan position approach that has complete information in the code stream;

Four, use that the strategy identical with built-in end carries out redundancy estimation in the above-mentioned hidden method, adjust and embed point and embedment strength, determine that just each embeds the upward wavelet coefficient bit number that can embed at most of point;

Five, extract synchronizing information, obtain the figure place that hides Info of actual embedding, continue to extract hiding Info according to synchronizing information;

Six, carry out wavelet coefficient inverse quantization and inverse wavelet transform, reconstructed image.

Scheme and principles illustrated:

By being compared with the JPEG2000 basic encoding unit, cataloged procedure of the present invention can know the action principle of understanding secondary coding, the basic encoding unit cataloged procedure of JPEG2000 as shown in Figure 1, input picture becomes the subband of different resolution and different directions earlier through wavelet decomposition, each subband is divided into encoding block through after quantizing, be unit then with the encoding block, carry out Bit-Plane Encoding according to bit plane order from high to low, generate independently compressed bit stream of each encoding block, wherein the code stream of each bit plane is made up of three coding passes again, obtains different compression ratios thereby can code stream be blocked in the position that any one coding pass finishes.According to predetermined compression ratio, determine the disconnect position of each encoding block code stream through rate-distortion optimization, at last the code stream of each encoding block intercepting is organized the compressed bit stream that constitutes entire image.Accompanying drawing 2 is compressed encoding and the information telescopiny schematic diagram in this programme, after the process rate-distortion optimization obtains the code stream disconnect position of each encoding block, directly do not organize code stream, but according to the result of rate-distortion optimization, determine that lowest order plane that encoding block is blocked to be influenced is as the lowest order plan position approach that can embed information, choose greater than the wavelet coefficient of threshold value 16 as embedding point, carry out the redundancy estimation and carry out the self adaptation adjustment embedding point and embedment strength, determine that each embeds the maximum embeddable bit numbers of point, then from embeddable lowest order plane, according to bit plane order from low to high synchronizing information and hiding Info is embedded in the wavelet coefficient after the quantification, wherein synchronizing information has been indicated the information digit of actual embedding, guarantees that correct extraction of receiving terminal hides Info.After embedding is finished, wavelet coefficient is carried out Bit-Plane Encoding again, the result of rate-distortion optimization organizes code stream when encoding according to the first time then.From schematic diagram as seen, this Information hiding scheme is embedded into information not in the bit plane that can be blocked by rate-distortion optimization by secondary coding, thereby has guaranteed the embedding information integrity.So-called secondary coding has just partly carried out twice with the entropy coding in the cataloged procedure, and processes such as other quantification as wavelet transformation, wavelet coefficient, rate-distortion optimization, code stream tissue all only need be carried out once, so computation complexity increases little.

Information extracting method as shown in Figure 3, resolve code stream, it is the form of unit that code stream is organized into the encoding block, each encoding block is carried out wavelet coefficient before entropy decoding obtains inverse quantization, in the entropy decode procedure, can know the lowest order plan position approach that has complete information in the code stream, the use self adaptation adjustment strategy identical with built-in end draws the embedment strength on each embedding point simultaneously, extracts synchronizing information at last in view of the above and hides Info.

Another characteristics of this programme are the Nonlinear Processing of utilizing wavelet coefficient before quantizing, and the luminosity sensitivity characteristics of human eye, quantification redundancy to each wavelet coefficient is estimated, accomplishes that embedment strength and the human eye masking characteristics on each embedding point combined closely.In the narration of following principle, only being for the wavelet coefficient to uniform quantization carries out redundancy estimation to the calculating of wavelet coefficient, to draw rational embedment strength on each aspect, is not that wavelet coefficient after the use processing is encoded.

Advantage of the present invention and good effect:

The present invention is directed to the basic compressing and coding system of defined in the JPEG2000 standard, designed a kind of considerable information hide method that is applicable to covert communications, this method is determined embedded location based on the secondary coding strategy, and by selecting the embedding point adaptively and adjusting the Information Hiding Capacity that embedment strength improves compressed image, on the one hand, strengthen embedded quantity at the insensitive image-region of human eye, on the other hand, at the image-region minimizing embedded quantity of human eye sensitivity.Last total embedding capacity then depends on factors such as the texture level, light and shade of concrete carrier image.This method is transparent fully for the JPEG2000 decoder of standard, can be integrated in the JPEG2000 encoder easily, but the final code stream that generates is normal decoder still, and the receiving terminal of covert communications then needs to use decode procedure of the present invention to extract to hide Info.

This scheme has two main features: one, determine the information embedded location by secondary coding; Two, self adaptation is regulated the embedment strength of each position.

[description of drawings]:

Fig. 1 is the basic coding system of JPEG2000 prescribed by standard;

Fig. 2 is the coding and the information telescopiny of this programme;

Fig. 3 is the decoding and the information extraction process of this programme;

Fig. 4 is original carrier image;

Fig. 5 is the image that hides Info;

Fig. 6 is the tower structure after 5 layers of wavelet decomposition;

Fig. 7 embeds point to choose and determine the minimum schematic diagram that embeds bit plane;

Fig. 8 is the interior wavelet coefficient of 5 * 5 neighborhoods of wavelet coefficient D;

Fig. 9 is the adjusted schematic diagram of embedment strength self adaptation;

Figure 10 is the concealed image after the embedding information;

Figure 11 is the error image before and after the embedding information;

Figure 12 is the image that hides Info that extracts and recover.

[embodiment]:

Embodiment 1:

A specific embodiment of this programme is as follows.

The initial carrier image is 512 * 512, the gray scale Lena image of every pixel 8 bits, as shown in Figure 4.Hiding Info of intending embedding is the bianry image of the Civil Aviation University of China logo shown in the accompanying drawing 5, the image size is 80 * 80, every pixel 1 bit, therefore always have in the image after hiding Info of 6400 bits need be embedded into compression, need embed in image compression and hide Info, compression ratio is 16.

The first step is carried out 5 layers of wavelet decomposition to original image, obtains tower structure as shown in Figure 6, wherein the 1st, the 2nd and the 3rd decomposition layer frequency is higher, can embed to hide Info, and the 4th and the 5th decomposition layer belongs to the low frequency important information, bigger to the picture quality influence, be not suitable for hiding Info.Wavelet coefficient is carried out uniform quantization, and choosing of each quantized subband rank distance is relevant with direction with the residing decomposition layer of subband.Quantize about wavelet decomposition, wavelet coefficient, and the details of follow-up Bit-Plane Encoding and rate-distortion optimization, in the JPEG2000 standard, have a detailed description (omiting) herein.It with each sub-band division 64 * 64 encoding block, with the encoding block is unit, according to the order from a high position to the low level, bit plane is encoded one by one, the coding of each bit plane comprises three coding passes again, the significance level of the code stream each several part that generates reduces in the past backward gradually, therefore as long as block in the position of specific coding channel end, just can obtain the code stream of different accuracy (different in other words compression ratio).If but just simply the code stream of all encoding blocks is organized into the code stream of entire image, and can not reach the purpose that 16 multiplication of voltages contract, data volume is much bigger, therefore must choose most important parts respectively and form final code stream in the code stream of each encoding block.Carry out rate-distortion optimization, determine the code stream disconnect position of each encoding block, and then draw each encoding block and can embed the lowest order plane that hides Info.As shown in Figure 6 and Figure 7, wherein in the 2nd decomposition layer HL direction subband code stream disconnect position of first encoding block at the 3rd coding pass of the 6th bit plane, here " the 6th bit plane " is to begin counting from the highest non-full zero bit plane, because 3 coding pass information completelies of the 6th bit plane, this bit plane can be used to carry and hides Info, and therefore it is defined as the lowest order plane that embedding of this encoding block hides Info.If the 6th situation that bit plane exists coding pass to be clipped should determine that so then the 5th high one bit plane is for can embed the lowest order plane;

In second step, just elect all range values in the encoding block as the embedding point more than or equal to the wavelet coefficient of threshold value 16.The information of considering embeds the low level that will change wavelet coefficient, the embedding point of selecting during in order to ensure the receiving terminal information extraction is consistent with transmitting terminal, threshold value should be 2 power, threshold value obtains big more, qualified embedding is counted out just few more, hiding capacity relatively just reduces, and also just littler to the influence of picture quality.Be that example illustrates the situation that embeds some primary election with three typical wavelet coefficients in the encoding block below, as shown in Figure 7, wavelet coefficient C is because can not be as embedding point less than threshold value; Wavelet coefficient A, B and D are chosen as the embedding point, and the bit on their the 6th bit plane is marked as embeddable, represent follow-up can the replacement them with the bit that hides Info.

Utilize formula (1) to (6) that all that select are embedded point and carry out the redundancy estimation, parameter value is: N=5, α=0.7, β=0.2.With wavelet coefficient D is example, and its quantization step is a Δ _iWavelet coefficient in=2,5 * 5 neighborhoods as shown in Figure 8, with wavelet coefficient in the horizontal LL subband of the corresponding lowest resolution in wavelet coefficient D locus be 56.The Practical Calculation process is as follows:

After the following zero clearing of wavelet coefficient D highest order, 67 become 64, and substitution (1) formula gets:

y _i＝sign(x _i)|x _i·Δ _i| ^α

＝(64×2) ⁰⁷

＝29.86

-54 ,-47,31,55,49 ,-26 ,-22,51,18 ,-19,41 ,-62 in 24 neighborhood wavelet coefficients shown in Figure 8, the wavelet coefficient that range value is not less than threshold value 16 has 12, is respectively:.-32 ,-32,16,32,32 ,-16 ,-16,32,16 ,-16,32 ,-32 will become after the following zero clearing of highest order:, substitution (2) formula gets:

$z_i=\frac{y_i}{1+\left(a\underset{k\&Element;\mathrm{neighborhood}}{\mathrm{\&Sigma;}}{\left|{\stackrel{\&OverBar;}{x}}_k\right|}^{\mathrm{\&beta;}}\right)/\left|{\mathrm{\&phi;}}_i\right|}$

$=\frac{30.83}{1+\frac{3.47\&times;22.71}{24}}$

$=7.20$

Because $1+\frac{56}{128}=1.44>1,$ So the luminosity sensitivity weight coefficient is 1.44, substitution (5) formula gets:

${z_i}^{\&prime;}=\frac{z_i}{1.44}=\frac{7.20}{1.44}=5$

By (6) formula, r _i=x _i/ z ' _i=67/5=13.4

Same process can be regarded as the r of wavelet coefficient A and B _iAs follows:

Wavelet coefficient A:1.72

Wavelet coefficient B:2.59

Because 2 ³＜13.3＜2 ⁴So, judge that wavelet coefficient D can embed 3 bits and hide Info, therefore except the 6th bit plane of mark, be embeddable with the bit labeling on the 5th bit plane of coefficient D and the 4th bit plane again;

Because 1.72＜2, judge wavelet coefficient A embeddings that be not suitable for hiding Info in view of the above, but the embedded location mark on coefficient A the 6th bit plane of being done during therefore with primary election erase, wavelet coefficient A also just no longer is the embedding point;

Because 2 ¹＜2.59＜2 ², judge that in view of the above wavelet coefficient B can embed hiding Info of 1 bit, but can not increase its embedment strength again.

Adjusted result as shown in Figure 9.

In the 3rd step, the 1st, the 2nd and the 3rd decomposition layer always has 63 encoding blocks, but the embedded location of all encoding blocks is added up the embedded capacity of each encoding block after all border note is finished.Still be not enough to embed the encoding block (this situation seldom occurs) of synchronizing information for capacity, but only its top 4 embedded locations replaced with 0, promptly embed ' 0000 ', not hide Info in this encoding block of notice extraction end.Hide Info according to the big young pathbreaker of capacity then and distribute to other encoding block.When the information of carrying out embeds, but preceding 4 embedded locations of each encoding block embed ' 1010 ' earlier, represent to hide Info in this encoding block, represent the bit number that hides Info that this encoding block is assigned to, then embed the bit that hides Info again followed by 10 bits that embed.From embedding the lowest order plane, bit plane scanned one by one from low to high, but whenever to an embedded location, promptly embeds 1 bit information, until the embedding of finishing all information when information embedded;

In the 4th step, carry out Bit-Plane Encoding again;

In the 5th step, the result who blocks according to the rate-distortion optimization that is carried out behind the Bit-Plane Encoding first time organizes final code stream.

The image that embedding hides Info with JPEG2000 standard decoder decoding back effect as shown in figure 10.The difference of two width of cloth images is very little before and after embedding, and should not directly show, the error image in the accompanying drawing 11 is that difference be multiply by the effect that shows again after 20.Error image can indicate the locus of the actual embedding that hides Info, and can see hides Info has been embedded into texture-rich or has crossed the insensitive image-region of bright dark excessively human eye, and embedding the back image does not have perceptiblely to degrade.

Information extraction process is the contrary operation of telescopiny, and difference has been narrated clear in scheme principle part, do not repeat them here.Hiding Info of extracting in this example is identical with built-in end after recovering, and effect as shown in figure 12.

Claims (4)

1, a kind of is the considerable information hide method that is applicable to covert communications of carrier with the JPEG2000 compressed image, and it is characterized in that: the operating procedure of Information hiding is as follows:

The first,, original image is carried out multilayer wavelet decomposition, wavelet coefficient quantification, Bit-Plane Encoding, and rate-distortion optimization blocks according to the method for stipulating in the JPEG2000 standard; Draw the lowest order plane that wavelet coefficient in each encoding block can embed information according to the code stream disconnect position;

The second, tentatively choose greater than the wavelet coefficient of threshold value 16 as embedding point, result according to the redundancy estimation adjusts embedding point and embedment strength, removal is unsuitable for the embedding point of the information of embedding, and draws the information bit that can embed at most on each embedding point;

Three, the lowest order plane from embedding is embedded into synchronizing information and hiding Info in the wavelet coefficient after the quantification according to bit plane order from low to high;

Four, finish information and embed after, carry out Bit-Plane Encoding once more;

Five, block the result according to the rate-distortion optimization in the first step and organize final code stream.

2, method according to claim 1 is characterized in that, embedding point that hides Info and the information bit that can embed at most calculate by following redundant evaluation method:

The first step, utilization are handled wavelet coefficient from contrasting masking effect,

y _i＝sign(x _i)|x _i·Δ _i| ^α(1)

X wherein _iBe wavelet coefficient after the uniform quantization and the value after the following zero clearing of highest order, Δ _iBe the quantization step of current wavelet coefficient, 0＜α＜1, the following zero clearing of highest order is to obtain same result of calculation in order to allow built-in end and extraction hold, thereby guarantees that algorithm is reversible, y _iIt is the wavelet coefficient values of utilizing from contrasting after masking effect is handled;

Second goes on foot, utilizes the neighborhood masking effect to handle wavelet coefficient,

$z_i=\frac{y_i}{1+\left(a\underset{k\&Element;\mathrm{neighborhood}}{\mathrm{\&Sigma;}}{\left|{\stackrel{\&OverBar;}{x}}_k\right|}^{\mathrm{\&beta;}}\right)/\left|{\mathrm{\&phi;}}_i\right|}---\left(2\right)$

It is wavelet coefficient in the N*N scope at center that the neighborhood wavelet coefficient refers to current coefficient, | φ _i| the number of expression neighborhood wavelet coefficient; Parameter beta ∈ [0,1] is the strength control factor of masking effect, β and | φ _i| controlling the dynamics of adjusting embedment strength according to the neighborhood masking effect together; Constant alpha=(1000/2 ^l) ^βBe the standardization factor, { 0,1,2,3,4} represents the level of resolution at wavelet coefficient place to l ∈;

Be the later neighborhood wavelet coefficient of uniform quantization, β gets the influence that less value can effectively suppress the very big wavelet coefficient of minority range value in the neighborhood, thus sharp edges district and texture area in can differentiate between images; z _iIt is the wavelet coefficient values of utilizing after the neighborhood masking effect is handled; For the assurance built-in end obtains same result of calculation with extracting to hold, also want zero clearing below the highest order of neighborhood wavelet coefficient, and only get the neighborhood wavelet coefficient participation calculating that range value is not less than threshold value 16;

The 3rd step, human eye are the most responsive to the noise of medium luminance area, sensitivity is along with the rising of brightness value or reduction and weaken, therefore increase by one by weight coefficient that luminosity sensitivity determined, with l ∈ { 0,1,2,3} represents the level of resolution after the wavelet decomposition, and l=0 represents the highest level of resolution; With θ ∈ { 0,1,2,3} represents the subband direction, is followed successively by LL, LH, HL, HH, and wherein LL represents that level and vertical direction all are high frequency, LH represents horizontal direction low frequency, vertical direction high frequency, and HL represents horizontal direction high frequency, vertical direction low frequency, and HH represents that level and vertical direction are all high frequency; I _l ^θ(i, j) expression level of resolution l down, direction is the locational wavelet coefficient that the capable j of i is listed as in the subband of θ, then in any direction subband under the l level of resolution, coordinate is (i, the luminosity sensitivity weight coefficient Λ of wavelet coefficient j) (l, i, j) following calculating:

$\mathrm{\&Lambda;}(l,i,j)=\left\{\begin{array}{cc}2-L(l,i,j),& \mathrm{ifL}(l,i,j)<1\\ L(l,i,j),& \mathrm{otherwise}\end{array}\right.---\left(3\right)$

Wherein median L (l, i, j) following calculating:

K is the progression of wavelet decomposition, and l=k is minimum level of resolution, wherein symbol Expression is not more than the maximum integer of x,

It is a wavelet coefficient in the LL subband of lowest resolution level, coordinate is (i in any direction subband under it and the l level of resolution, j) wavelet coefficient correspondence identical image-region, because pixel value having been done amplitude before the wavelet decomposition is 128 DC level displacement, the dynamic range of pixel value is (128,127), so carry out normalization divided by 128 in the following formula

Obtain at last:

${z_i}^{\&prime;}=\frac{z_i}{\mathrm{\&Lambda;}(l,i,j)}---\left(5\right)$

z _i' be the wavelet coefficient values after handling according to luminosity sensitivity; Quantification redundancy on the wavelet coefficient is then estimated by following formula:

r _i＝x _i/z _i′(6)

In fact ratio r i has reflected when using the uniform quantization method of the basic compressing and coding system of JPEG2000, the size of the quantification redundancy that exists on the current wavelet coefficient, under the binary system situation, every twice just represents to have the redundancy of a bit, thereby just the more than enough bit of embedding of energy hiding Info; For the influence of control information embedding to picture quality, regulation has only the r of working as _iBe not less than at 2 o'clock and could embed and hide Info; When adjusting embedding point and embedment strength, at first get rid of r according to redundant estimation result _iIf embedding point less than 2 determines that according to following rule each embeds the bit number that hides Info that the some place can embed: then 2 ⁿ≤ r _i＜2 ^N+1, then this embedding point can embed the n bit and hides Info.

3, method according to claim 1, it is characterized in that, each embeds a definite method that can embed the lowest order plane of information, after the process rate-distortion optimization obtains the code stream disconnect position of each encoding block, choose three coding passes lowest order all not intercepted, information completely just plane, as the lowest order plane that can embed information.

4, the extracting method of the information hidden of the described method of claim 1 is characterized in that the extraction step that hides Info is as follows:

The first, the coding/decoding method according to the JPEG2000 standard code carries out code stream analyzing, entropy decoding, obtains the preceding wavelet coefficient of inverse quantization;

The second, the same principle primary election of built-in end embedded point during use and claim 1 second went on foot;

Three, the information that has the bit plane of three coding passes in the code stream is complete; During the entropy decoding, find out the lowest order plan position approach that has complete information in the code stream;

Four, use that the strategy identical with built-in end carries out redundancy estimation in the claim 2, adjust and embed point and embedment strength, determine that just each embeds the upward wavelet coefficient bit number that can embed at most of point;

Five, extract synchronizing information, obtain the figure place that hides Info of actual embedding, continue to extract hiding Info according to synchronizing information;

Six, carry out wavelet coefficient inverse quantization and inverse wavelet transform, reconstructed image.

Images