CN1885340A - Reversible data concealing method based on integer wavelet statistic characteristics - Google Patents

Reversible data concealing method based on integer wavelet statistic characteristics Download PDF

Info

Publication number
CN1885340A
CN1885340A CNA2006100363867A CN200610036386A CN1885340A CN 1885340 A CN1885340 A CN 1885340A CN A2006100363867 A CNA2006100363867 A CN A2006100363867A CN 200610036386 A CN200610036386 A CN 200610036386A CN 1885340 A CN1885340 A CN 1885340A
Authority
CN
China
Prior art keywords
coefficient
peak
band
histogram
frequency sub
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.)
Pending
Application number
CNA2006100363867A
Other languages
Chinese (zh)
Inventor
梁小萍
黄继武
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Sun Yat Sen University
National Sun Yat Sen University
Original Assignee
National Sun Yat Sen University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by National Sun Yat Sen University filed Critical National Sun Yat Sen University
Priority to CNA2006100363867A priority Critical patent/CN1885340A/en
Publication of CN1885340A publication Critical patent/CN1885340A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Editing Of Facsimile Originals (AREA)
  • Image Processing (AREA)

Abstract

The invention relates to a reversible data hiding method, belonging to the multimedia information safety technique, wherein said method comprises: hiding data and checking data; the first step comprises: 1, correcting histogram; 2, decomposing integral wavelet; 3, selecting embedded mode; 4, selectively modifying low-frequency factor; 5, modifying high-frequency factor; 6, rebuilding integral wavelet; 7, judging pixel value overflow; and the second step comprises: 1, decomposing integral wavelet; 2, selectively extracting; 3, extracting and recovering high-frequency factor; 4, returning low-frequency factor; 5, rebuilding integral wavelet; 6, returning histogram. The invention uses the character that the integral wavelet high-frequency sub-band factor of natural gray image is similar with gauss distribution, to modify the histogram of high-frequency sub-band and the low-level plane of low-frequency sub-band, to embed data, to hide information, to return the host image without aberration.

Description

Reversible data concealing method based on integer wavelet statistic characteristics
Technical field
The present invention relates to a kind of multi-media information security method, specifically is a kind of at natural gray image, based on the reversible data concealing method of integer wavelet statistic characteristics.
Background technology
Existing data hidden algorithm embeds certain data volume in Digital Media, also introduce subtle distortion simultaneously, and this all is suitable under many application scenarios, comprises copyright protection, content authentication and secret communication.But for particular application such as military affairs, medical science, courts, in Digital Media, introduce permanent distortion do not allow often, undesirable, and on the network the exigent buyer of Digital Media of quality is also wished to obtain undistorted master Digital Media, this can seek help from the hiding reversible data technology and solve this imbalance between supply and demand.Hiding reversible data be with Information hiding in Digital Media, after the information of hiding is extracted out, can undistortedly recover fully as the Digital Media of host signal.
By retrieving domestic and international prior art, find that relevant technical literature has following 13 pieces:
[1]J.M.Barton，“Method and apparatus for embeddingauthentication information within digital data，”U.S.Patent5,646,997，1997.
[2]J.Fridrich，M.Goljan，and R.Du，“Lossless dataembedding-new paradigm in digital watermarking，”EURASIP J.Appl.Sig.Process.，vol.2002，no.02，pp.185-196，Feb.2002.
[3]Guorong Xuan；Shi，Y.Q.；Ni，Z.C.；Jidong Chen；Chengyun Yang；Yizhan Zhen；Junxiang Zheng，“High capacity lossless data hidingbased on integer wavelet transform，”Proc.ISCAS，vol.2，pp.II29-32，May.2004.
[4]Mehmet Utku Celik，Gaurav Sharma，Ahmet Murat Tekalp，Eli Saber.“Lossless generalized-LSB data embedding，”IEEE Trans.ImageProcess.，vol.14，n.2，pp.253-266，Feb.2005.
[5]C.W.Honsinger，P.Jones，M.Rabbani，and J.C.Stoffel，“Lossless recovery of an original image containing embeddeddata，”US Patent：6，278，791，2001.
[6]M.Goljan，J.Fridrich，and R.Du，“Distortion-free dataembedding，”Proceedings of 4th Information Hiding Workshop，pp.27-41，Pittsburgh，PA，April 2001.
[7]C.De Vleeschouwer，J.F.Delaigle，and B.Macq，“Circularinterpretation of bijective transformations in losslesswatermarking for media asset management，”IEEE Trans.Multimedia，vol.5，no.1，pp.97-105，Mar.2003.
[8]Z.Ni，Y.Q.Shi，N.Ansari，W.Su，Q.Sun，X.Lin，“Robustlossless image data hiding，”IEEE Int.Conf.on Multimedia andExpo(ICME)，vol.3，pp.2199-2202，2004.
[9]J.Tian，“Reversible data embedding using a differenceexpansion，”IEEE Trans.Circuits Syst.Video Technol.，vol.13，no.8，pp.890-896，Aug.2003.
[10]Adnan M.Alattar，“Reversible watermark using the differenceexpansion of a generalized integer transform，”IEEE Trans.ImageProcess.，vol.13，n 8，pp.1147-1156，Aug.2004.
[11]Z.Ni，Y.Q.Shi，N.Ansari and W.Su，“Reversible DataHiding，”IEEE Int.Symposium on Circuits and Systems，Bangkok，Thailand，May 2003.
[12]A.Van Leest，M.Van der Veen，F.Bruekers.“Reversible imagewatermarking，”IEEE Int.Conf.on Image Processing，vol.2，pp.731-734，2003.
[13]Y.Q.Shi，Z.Ni，D.Zou，C.Liang and G.Xuan，″Lossless datahiding：Fundamentals，algorithms and applications，″Proceedings ofIEEE International Symposium on Circuits and Systems，vol.II，pp.33-36，Vancouver，Canada，May 2004.
The appearance of hiding reversible data algorithm can be traced back to the invention (document [1]) of Barton in 1997 the earliest.Existing hiding reversible data algorithm can be divided into hiding reversible data algorithm that relies on compress technique and hiding reversible data algorithm two classes that are independent of compress technique roughly according to whether adopting lossless compressiong to realize.The side information of the extra channel of needs that the demoder in the two class algorithms has is auxiliary just to be correctly decoded, and can not realize blind extraction, and the extra side information that then do not need that has just can be correctly decoded, and can realize blind extraction.
Early stage hiding reversible data generally belongs to hiding reversible data algorithm one class that relies on compress technique, adopt the thought of lossless compress to realize, exactly the partial bit stream on the carrier media spatial domain or on the transform domain is carried out lossless compress, " vacating space " gives the information that really will embed, and packed data and embedding information are formed the position that bit stream is embedded into the preceding bit stream of compression again.The invention of Barton (document [1]) is used for digital multimedia, as jpeg image and MPEG video, adopts lossless compressiong that the partial bit stream of carrier is carried out lossless compress, and vacating space is used to embed authentication information.The method of Fridrich (document [2]) is carried out lossless compress in the image spatial domain with the low bit-planes of pixel value, and vacating space is used to embed authentication information.The method of Xuan (document [3]) is image to be carried out integer wavelet decompose, and the low bit-planes of coefficient of detail subbands is carried out lossless compress, and vacating space embeds information.The method of Xuan can obtain than the bigger embedding capacity of preceding two kinds of methods, but invisibility is not poor slightly, just covers the still differentiated of close image and host image on visual effect.The method of Celik (document [4]) has adopted more efficiently lossless compression method, and LSB (lowest bit position) expanded to general LSB (generalized-LSB), performance increases, and is to rely on method representative in the reversible data concealing method of compress technique, better performances.These algorithms utilize the redundance of low bit-planes data to embed data as far as possible, but because the data randomness of low bit-planes is strong, redundance is not high, so the embedding capacity of these algorithms is not high, therefore embedding capacity even be negative value concerning many images is difficult to practicality.Often compress higher bit-planes data in order to improve the embedding capacity, but this can cause serious vision distortion.
Belong to hiding reversible data algorithm one class that is independent of compress technique, the invention (document [5]) of Honsinger is arranged the earliest, the method of employing mould 256 both can have been avoided the benefit of pixel value on the image spatial domain, also can recover host image, have more serious salt-pepper noise but cover close image at receiving end.Thereafter method has RS (document [6]), the method (document [7]) based on the patchwork theory, piecemeal-variance displacement method (document [8]), DE method (document [9] and document [10]) and spatial domain histogram displacement method (document [11] and [12]).The RS method is that the pixel with image is divided into R, S and U (the non-use of Unusable) group, and pairing embeds bit 1 and 0 in R, S group.But the DE method has been used for reference the thought of inverse integer transform, bit to be embedded is attached in the difference of integer.In order to realize blind extraction, RS method and DE method have all used lossless compressiong compression side information to give information to be embedded with " vacating space ".The RS method is the state of compressed image group of pixels.And the DE method is the group of pixels position that compression is elected to be the embedding carrier, thereby obtains preferable performance, promptly bigger embedding capacity and higher PSNR (Y-PSNR), and wherein the method for Tian is the most representative.Be primarily aimed at anti-JPEG lossy compression method based on the method for patchwork theory and piecemeal-variance displacement method and design, the embedding capacity is very little, because of different with the purposes of the inventive method, does not discuss herein.The spatial domain main thought of histogram displacement method is to seek the peak and the zero point of image histogram, the histogram that (does not comprise peak and zero point) with peak and between zero point moves a unit to the direction monoblock at zero point, make histogram be sky on the peak next door, also promptly produce the breach (gap) on the histogram, realize the embedding of Bit data again by the pixel value of revising histogram peak correspondence.The specific practice of revising the pixel value of histogram peak correspondence is: according to from top to bottom, from left to right sequential scanning entire image, when running into the pixel value of histogram peak correspondence, check bit to be embedded, if be " 1 ", then the pixel value of peak correspondence reduces the pixel value of correspondence (zero point less than the situation of the pixel value of peak correspondence) or increases the pixel value of correspondence (zero point greater than the situation of the pixel value of peak correspondence) unit, if be " 0 ", the pixel value of peak correspondence remains unchanged.Receiving end extracts the data of hiding and recovers the concrete grammar of host image: according to same sequential scanning entire image, when running into the pixel value of histogram peak correspondence, extract bit " 0 ", when running into when lacking the pixel value of a unit or many units than the pixel value of histogram peak correspondence, extract bit " 1 ", and this pixel value is reverted to the pixel value of peak correspondence.The algorithm of Ni (document [11]) needs extra channel transmission side information, such as the pixel value of histogram peak correspondence and the pixel value of histogram correspondence at zero point.The algorithm of Leest (document [12]) has carried out simple discussion to the problem of side information, proposition is embedded into side information the idea in the pixel in the image upper right corner, but do not provide any practicable method and concrete experimental data, and before built-in edges information, embedded when hiding Info when the pixel in the image upper right corner, then this scheme is infeasible.Spatial domain histogram displacement method does not use any compress technique, realize simply, and it is good than other hidden methods to cover close image visual effect, but the embedding capacity is less, can't realize embedding when image histogram does not have zero point.
Summary of the invention
The deficiency that adopt lossless compress in order to overcome existing major technique, need extra independent channel transmission side information, the embedding capacity is little, the invention provides a kind of characteristics of utilizing the approximate Gaussian distributed of integer wavelet high-frequency sub-band coefficient of natural gray image, do not need lossless compressiong and extra channel transmission side information, the embedding capacity is the hiding reversible data technology greatly and flexibly.
Technical solution of the present invention is as follows: the inventive method is to utilize the characteristics of the approximate Gaussian distributed of integer wavelet high-frequency sub-band coefficient of natural image, the histogram of modification high-frequency sub-band and the lowest order plane of low frequency sub-band are carried out the embedding of data and are hidden to realize reversible information, and this method is divided into data hidden and two processes of Data Detection.
Described data hidden process steps is as follows:
1) revises histogram: be about to host image X MxNHistogrammic scope is modified to [G, 255-G] from [0,255], the initial default value of G is 0 or according to X MxNHistogram is set, and record to be modified pixel be P M, comprise initial value, coordinate and G, remember that revised image is X ' MxN
2) integer wavelet decomposes: promptly to X ' MxNCarry out integer wavelet and decompose, obtain a low frequency sub-band LL LWith high-frequency sub-band set C K, l, calculating accessible maximum embedding capacity under five kinds of main embedded models according to histogrammic peak of each high-frequency sub-band coefficient and time high point, decomposed class is got L=3;
3) embedded model is selected: the maximum embedding capacity under hiding data amount and the five kinds of main embedded models relatively, select a kind of embedded model, and choose some high-frequency sub-band that are used to embed data according to human visual system (HVS) characteristic, the embedded model that record is selected, the high-frequency sub-band number of choosing, under this embedded model the selected histogrammic peak of each high-frequency sub-band and/or the coefficient value of inferior high correspondence and/or the direction that the coefficient histogram moves, the information of these records is designated as side information A;
4) low frequency coefficient is selectively modified: select low frequency sub-band LL with key K LThe LSB of coefficient (LeastSignificant Bit) embeds low frequency sub-band LL with side information A in the mode of replacing selecteed LSB LObtain LL ' L, and the LSB that record is replaced is Ori_LSB;
5) high frequency coefficient is revised: with P M, Ori_LSB and encrypt with key K after hiding data M ' form bit stream, and revise the high-frequency sub-band coefficient according to selected embedded model, bit stream is embedded in the high-frequency sub-band of choosing, the high-frequency sub-band set that is embedded with data is designated as C ' K, l
6) integer wavelet reconstruct: with LL ' LAnd C ' K, lCarry out three grades of integer wavelet reconstruct;
7) pixel value is overflowed judgement: if the pixel value after the reconstruct exceeds the scope of [0,255], be judged as and overflow, then return to step 1), increase the value of G and repeat above-mentioned steps until there not being pixel value to overflow, the step-length acquiescence of G gets 5 or set up on their own, otherwise has obtained hiding the close image X of covering of information " MxN
Described data detection process step is as follows:
1) integer wavelet decomposes: promptly to the close image X that covers to be detected " MxNCarry out three grades of integer wavelets and decompose, obtain a low frequency sub-band LL ' LWith high-frequency sub-band set C ' K, l
2) selective extraction: obtain LL ' by key K LIn be embedded with the LSB position of data, extract LSB from these positions and obtain side information A, analyze high-frequency sub-band number that side information A can obtain embedded model, choose, the selected histogrammic peak of each high-frequency sub-band and/or the coefficient value of inferior high correspondence and the direction that the coefficient histogram moves under this embedded model;
3) extract and recover high frequency coefficient: according to embedded model from C ' K, lMiddle extraction bit stream also recovers C K, l, analyze bit stream and obtain P M, Ori_LSB and M ', decipher M ' with key K and obtain M;
4) low frequency coefficient recovers: replace LL ' with Ori_LSB LIn be embedded with the LSB position of data, recover LL L
5) integer wavelet reconstruct: with LL LAnd C K, lCarry out three grades of integer wavelet reconstruct, obtain picture signal X ' MxN
6) recover histogram: according to P MWith X ' MxNHistogram [G, 255-G] revert to [0,255], obtain undistorted host image X MxN
Described data hidden process steps 3) in selected embedded model determined embedding capacity, the hiding data of each high-frequency sub-band embedding, extract with recovery and this process in coefficient the histogram direction and the size that move, main embedded model has five kinds, respectively called after embedded model 1, embedded model 2, embedded model 3, embedded model 4 and embedded model 5.
The embedding capacity of described embedded model 1 equals the histogrammic peak value of high-frequency sub-band coefficient n 1, the method that embedded model 1 embeds, detects, extracts and recovers is as follows:
1) at transmitting terminal, keeping under the constant situation of histogram peak, with the histogram peak wherein is the whole mobile unit of histogram on the left side or the right on one side, the direction that histogram moves promptly moves still to move to the right to the left side can follow the principle that as far as possible obtains best visual effect, the coefficient value of promptly keeping as much as possible is constant, make on histogram peak next door and a breach (gap) occurs, concrete realization is at first to add up the coefficient histogram, obtains the coefficient value c of histogram peak correspondence 1st, calculate the histogrammic coefficient number on peak both sides and compare, the coefficient on the note peak left side adds up to N Left, the coefficient on note peak the right adds up to N RightIf, N Left＜N Right, then the histogram on peak left side integral body outwards moves a unit, and promptly all coefficient values of this part histogram correspondence reduce by units, if N Left〉=N Right, then the histogram integral body on peak the right outwards moves a unit, and promptly all coefficient values of this part histogram correspondence increase units, suppose that comparative result is N Left〉=N Rigt, then to whole high-frequency sub-band carry out from top to bottom, from left to right sequential scanning, all coefficient values are added 1 (comparative result is then got on the contrary and subtracted, and next coming in order are analogized) greater than the coefficient value of histogram peak correspondence;
2) once more high-frequency sub-band is scanned according to same order, when running into peak coefficient c 1stThe time, the data that inspection will embed, if to be embedded is bit " 1 ", then the peak coefficient value is added 1, it is fill up the gap, if to be embedded is bit " 0 ", then keep the peak coefficient value constant, up to all data bits being embedded the maximum embedding capacity that finishes or reach this high-frequency sub-band;
3) at receiving end, according to same order high-frequency sub-band is scanned, when running into the peak coefficient, extract bit " 0 ", when the coefficient that runs into than peak coefficient value big 1, extract bit " 1 ", thereby and this coefficient value subtracted 1 revert to the peak coefficient value, even breach reappears;
4) once more high-frequency sub-band is scanned according to same order, all coefficient values are subtracted 1 greater than the coefficient value of histogram peak correspondence, promptly fill breach, recover whole coefficient subband.
The embedding capacity of described embedded model 2 equals the histogrammic peak value of high-frequency sub-band coefficient n 1With inferior high point value n 2Sum, the method that embedded model 2 embeds, detects, extracts and recovers is as follows:
1) at transmitting terminal, keeping under the constant situation of histogram peak and time high integral body, with histogram peak and time high outside respectively whole unit that moves of histogram that both sides are the left side and the right, make at histogram peak and time this whole both sides of high point and occur two breach altogether, concrete realization is at first to add up the coefficient histogram, obtains the coefficient value c of histogram peak correspondence 1stWith inferior high corresponding coefficient value c 2ndAnd the size of two coefficient values relatively, then to whole high-frequency sub-band carry out from top to bottom, from left to right sequential scanning, if c 1st〉=c 2nd, then with all coefficient values greater than c 1stCoefficient value add 1, with all coefficient values less than c 2ndCoefficient value subtract 1 (comparative result is then got successively on the contrary and subtracted, adds, below similar);
2) once more high-frequency sub-band is scanned according to same order, when running into peak coefficient c1st, the data that inspection will embed, if to be embedded is bit " 1 ", then the peak coefficient value is added 1, if to be embedded is bit " 0 ", then keep the peak coefficient value constant, when running into time high dot factor c2nd, the data that inspection will embed, if to be embedded be bit " 1 ", then the peak coefficient value is subtracted 1, if to be embedded is bit " 0 ", then keep time high dot factor value constant, up to all data bits being embedded the maximum embedding capacity that finishes or reach this high-frequency sub-band;
3) at receiving end, according to same order high-frequency sub-band is scanned, when running into peak coefficient c1st or inferior high dot factor c2nd, extract bit " 0 ", when running into coefficient c1st+1 or c2nd-1, extract bit " 1 ", and coefficient value is reverted to c1st or c2nd;
4) once more high-frequency sub-band is scanned according to same order, all coefficient values are subtracted 1 greater than the coefficient value of c1st, all coefficient values are added 1 less than the coefficient value of c2nd, recover whole coefficient subband.
The embedding capacity of described embedded model 3 equals the histogrammic peak value of high-frequency sub-band coefficient n 1Twice, embedded model 3 embeds, detects, extracts with the method for recovery as follows:
1) at transmitting terminal, keeping under the constant situation of histogram peak, outwards move a unit and another side whole outwards move two units with the histogram on histogram peak both sides is wherein whole on one side, make on histogram peak next door and two breach that size is respectively a unit and two units occur, concrete realization is at first to add up the coefficient histogram, obtains the coefficient value c of histogram peak correspondence 1st, calculate the histogrammic coefficient number on peak both sides and compare, the coefficient on the note peak left side adds up to N Left, the coefficient on note peak the right adds up to N RightIf, N Left＜N Right, then the histogram on the peak left side is whole outwards moves two units, and the histogram integral body on peak the right outwards moves a unit, if N Left〉=N Right, then opposite, suppose that comparative result is N Left〉=N Rigt, then to whole high-frequency sub-band carry out from top to bottom, from left to right sequential scanning, with all coefficient values greater than c 1stCoefficient value add 2, with all coefficient values less than c 1stCoefficient value subtract 1 (comparative result is then got successively on the contrary and subtracted, adds, below similar);
2) once more high-frequency sub-band is scanned according to same order, when running into peak coefficient c 1stThe time, continuous two bits that inspection will embed, four kinds of combinations and four coefficient value c of continuous two bits 1st-1, c 1st, c 1st+ 1 and c 1st+ 2 is corresponding one by one, according to bit combination to be embedded with c 1stBe revised as c 1st-1, c 1st+ 1 or c 1st+ 2, or keep c 1stConstant, up to all data bits being embedded the maximum embedding capacity that finishes or reach this high-frequency sub-band;
3), according to same order high-frequency sub-band is scanned, when running into c at receiving end 1st-1, c 1st, c 1st+ 1 or c 1stExtracted its corresponding bit combination at+2 o'clock, and coefficient value is reverted to c 1st
4) once more high-frequency sub-band is scanned according to same order, with all coefficient values greater than c 1st Coefficient value subtract 2, with all coefficient values less than c 1stCoefficient value add 1, recover whole coefficient subband.
The embedding capacity of described embedded model 4 equals histogrammic high point value n of this high-frequency sub-band coefficient 2With twice peak value n 1Sum, essence are that time high point has adopted embedded model 1, peak to adopt embedded model 3.
The embedding capacity of stating embedded model 5 equals histogrammic high point value n of this high-frequency sub-band coefficient 2With peak value n 1The twice of sum, essence are that time high point and peak have all adopted embedded model 3.
The advantage that the present invention compared with prior art has is as follows:
1) has bigger embedding capacity and visual effect preferably.The reversible data concealing method of main flow is that the reversible data concealing method (as above-mentioned document [1]～[4], [6], [9] and [10]) that relies on compress technique more or less all takes lossless compressiong that the partial bit stream of host signal or host signal is after treatment carried out lossless compress, if the bitstream length before the compression is L, length is Ls after the lossless compress, and then embeddable bit length is no more than L-Ls.Because the general more complicated of texture of natural gray image, particularly lower bit-planes, characteristic with similar random noise, the reversible data concealing method of main flow mainly is to compress lower bit-planes, thereby the effect of lossless compress is undesirable, and the embedding capacity is very little, can increase the embedding capacity though compress higher bit-planes, but but can cause relatively poor visual effect, promptly cover close image and produce more serious distortion.For the texture complex image, the result of lossless compress is the increase rather than the minimizing of data length often, and promptly the embedding capacity is for negative.Document [4] has adopted more efficiently lossless compression method, and LSB (lowest bit position) expanded to general LSB (generalized-LSB), performance increases, and is to rely on method representative in the reversible data concealing method of compress technique, better performances.Document [6] carries out lossless compress with the state of image pixel group, performance that obtains and document [4] quite.Document [9] and [10] will be elected to be the group of pixels location drawing (position map) that embeds carrier and carry out lossless compress, be performance the best ways in document [1]～[12], and the most representative is the method for document [9] Tian.The another kind of reversible data concealing method of compress technique that is totally independent of is (as above-mentioned document [5], [11] and [12]) need not to adopt lossless compressiong, the method of document [5] adopts mould 256, but generation salt-pepper noise, visual effect is relatively poor, the method of document [11] and [12] all is a spatial domain histogram displacement method, main thought is to seek histogrammic peak of image pixel and zero point, the histogram that (does not comprise peak and zero point) with peak and between zero point moves a unit to the direction monoblock at zero point, make in the histogram of peak adjacency to empty, also promptly produce the breach on the histogram, realize the embedding of Bit data again by the pixel value of revising histogram peak correspondence, the embedding capacity equals the peak value.Spatial domain histogram displacement method does not use any compress technique, and it is good than other reversible data concealing methods to cover close image visual effect, but because the histogrammic peak value in image spatial domain is generally little, thereby the embedding capacity is little.The present invention utilizes the characteristics of the approximate Gaussian distributed of integer wavelet high-frequency sub-band coefficient of natural gray image, revise high-frequency sub-band coefficient histogram and realize the embedding of hiding data, because the high frequency coefficient histogram has peak point, and the value of peak point is more much bigger than the histogrammic peak value in image spatial domain, thereby can obtain bigger embedding capacity.Because the present invention just does slight modification to wavelet coefficient, therefore has visual effect preferably.
2. realization is simpler, has better practicability.The reversible data concealing method that relies on compress technique need increase the lossless compress coder/decoder in terminal, thereby system realizes complicated.The present invention does not need to adopt lossless compressiong, realizes fairly simple.Being totally independent of the reversible data concealing method of compress technique such as the scheme of document [11] needs extra independent channel transmission side information, such as the pixel value of pixel histogram peak correspondence and the pixel value of pixel histogram correspondence at zero point.The present invention does not need extra channel transmission side information, but side information is embedded in the low frequency sub-band of host image, has better practicability.
3. have wider applicability and better flexibility.The reversible data concealing method that relies on compress technique selects texture image simple, that have the bulk flat site could realize having the data hidden of using value as host signal, and the reversible data concealing method that is totally independent of compress technique can only select to have the image at histogram zero point as host signal.The present invention does not have special requirement to host image, so long as natural gray image gets final product, therefore has wider applicability.It is available that the present invention has multiple embedded model, the user can select any embedded model according to the size of hiding data amount and the needs of visual effect, even use and mix embedded model, have better flexibility, can be used on general image and sensitization picture such as hiding secret information, authentication information etc. in military picture, the medical picture.
4. has higher security.The present invention uses key that hiding data is encrypted, and use key to select the embedded location of side information, choosing of these two keys can be the same or different, and decides on the realization of concrete system, the use of double secret key has increased the difficulty that the assailant cracks, and has strengthened the security of system.
Description of drawings
Fig. 1 is the realization block diagram of the inventive method;
Fig. 2 is the realization block diagram of data detection process;
Fig. 3 is a high-frequency sub-band coefficient histogram;
Fig. 4 is five kinds of main embedded model synoptic diagram of this programme;
Fig. 5 is the performance comparison diagram of this programme (Liang) under the mixing embedded model with two kinds of main reversible data concealing methods (Celik and Tian);
Fig. 6 is the design sketch of 512 * 512 * 8 bmp image Lena to size for the present invention.
Embodiment
The present invention can be applicable to general image and sensitization picture such as reversible hiding secret information, authentication information etc. in military picture, the medical picture.The realization block diagram of the inventive method as shown in Figure 1, X ' wherein NxMExpression is the image of spatial domain histogram modification through pre-service, C K, l, LL LRepresent high-frequency sub-band and low frequency sub-band after integer wavelet decomposes respectively, C ' K, l, LL ' LExpression has embedded high-frequency sub-band and the low frequency sub-band after the data respectively.Fig. 2 is the realization block diagram of this programme Data Detection method, has comprised the extraction of hiding data and the recovery of host image.Fig. 3 is a high-frequency sub-band coefficient histogram, and wherein Fig. 3 (a) is the HH1 subband histogram of Lena image, and Fig. 3 (b) is the HL1 subband histogram of Lena image.Fig. 4 is five kinds of main embedded model synoptic diagram of this programme, if bit sequence to be embedded is " 101,101 00 10; 10 11 01 00 10; 10 11 01 00 10 ... ", wherein Fig. 4 (a) is the original histogram of coefficient, the coefficient histogram that Fig. 4 (b1) moves for 1 time process of embedded model, Fig. 4 (b2) is the coefficient histogram after 1 time data of embedded model embed, the coefficient histogram that Fig. 4 (c1) moves for 2 times processes of embedded model, Fig. 4 (c2) is the coefficient histogram after 2 times data of embedded model embed, the coefficient histogram that Fig. 4 (d1) moves for 3 times processes of embedded model, Fig. 4 (d2) is the coefficient histogram after 3 times data of embedded model embed, the coefficient histogram that Fig. 4 (e1) moves for 4 times processes of embedded model, Fig. 4 (e2) is the coefficient histogram after 4 times data of embedded model embed, the coefficient histogram that Fig. 4 (f1) moves for 5 times processes of embedded model, Fig. 4 (f2) are the coefficient histogram after 5 times data of embedded model embed.Fig. 5 is the performance comparison diagram of this programme (Liang) under the mixing embedded model with two kinds of main reversible data concealing methods (Celik and Tian), performance comprises that (unit is bpp to the embedding capacity, the bit number that every pixel embeds) and PSNR (unit is dB), experimental image is 512 * 512 * 8 bmp image Lena.Fig. 6 is the design sketch of 512 * 512 * 8 bmp image Lena to size for the present invention, Fig. 6 (a) is original/host image Lena, Fig. 6 (b) for embedded 36071 bits promptly the embedding capacity be the close image of covering of 0.1376bpp, PSNR is 43.20dB, Fig. 6 (c) for embedded 107216 bits promptly the embedding capacity be the close image of covering of 0.4090bpp, PSNR is 39.44dB.
Here be that 512 * 512 gray level image lena.bmp is an example with representative size, see shown in Fig. 6 (a), describe implementation method of the present invention.At transmitting terminal, at first be to revise histogram.The pixel histogram of elder generation's analysis image.The lena image is positioned at [0,255] pixel at edge seldom, therefore data hidden scheme of the present invention again is very little to the change of integer wavelet coefficient, gets acquiescence initial value G=0 for 1 time at embedded model and also can not produce and overflow, promptly the histogram of this image does not need to revise yet, and record is modified pixel P MIn include only the coefficient number 0 that is modified and the value 0 of G.Can not produce and overflow and get G=10 for 4 times at embedded model, perhaps under default value through judge for 2 times, circulation and reach G=10 (step-length that G increases is defaulted as 5), the lena image does not have pixel in [0,10] scope, a pixel (247 is arranged in [245,255] scope; 274,117), promptly pixel value is 247, and planimetric coordinates is (274,117), now changes pixel value 247 into 247-G=247-10=237, and record is modified pixel P MIn comprise pixel value and coordinate (247; 274,117) coefficient number 1 that, is modified and the value 10 of G.
Decompose use Via Lifting Scheme to carry out integer wavelet through the image after the above-mentioned processing then, calculating process is as follows:
s l ( 0 ) = x 2 l d l ( 0 ) = x 2 l + 1 - - - d l ( 1 ) = d l ( 0 ) + Int ( α ( s l ( 0 ) + s l + 1 ( 0 ) ) ) s l ( 1 ) = s l ( 0 ) + Int ( β ( d l ( 1 ) + d l - 1 ( 0 ) ) ) - - - d l 2 = d l ( 1 ) + Int ( γ ( s l ( 1 ) + s l + 1 ( 1 ) ) ) s l ( 2 ) = s l ( 1 ) + Int ( δ ( d l ( 2 ) + d l - 1 2 ) ) Formula group (1)
d l ( 3 ) = d l ( 2 ) + Int ( ( ζ - ζ 2 ) s l ( 2 ) ) s l ( 3 ) = s l ( 2 ) + Int ( ( - 1 / ζ ) d l ( 3 ) ) - - - d l ( 4 ) = d l ( 3 ) + Int ( ( ζ - 1 ) s l ( 3 ) ) s l ( 4 ) = s l ( 3 ) + d l ( 4 ) - - - s l = s l ( 4 ) d l = d l ( 4 ) Formula group (2)
Wherein Int () function is a bracket function, and the value of 5 parameters is respectively α=-1.586134342, β=-0.05298011854, γ=0.8829110762, δ=0.4435068522, ζ=1.149604398.This programme adopts three grades of integer wavelets to decompose, and obtains 9 high-frequency sub-band and 1 low frequency sub-band like this.
Then detect 9 histogrammic peaks of high-frequency sub-band coefficient and time high point, for the present invention program's performance is described, adopt pattern 1 and pattern 4 to carry out the embedding of maximum embedding capacity here respectively, this moment, all high-frequency sub-band all were used to embed data.The maximum embedded quantity of lena image is as shown in table 1 under five kinds of main embedded models.When adopting embedded model 1, the direction that 1 time each high-frequency sub-band histogram of logging mode moves and the coefficient value of peak value correspondence, as shown in table 2.Wherein 1 expression is moved to the right in table 2 " moving direction " hurdle, and 0 expression is moved to the left side, sees synoptic diagram Fig. 4 (b1).HH represents that the diagonal angle decomposes the subband of direction, and HL represents the subband of orthogonal decomposition direction, and LH represents the subband of horizontal decomposition, the progression that numeral 1,2,3 expressions on next door are decomposed.When adopting embedded model 4,4 times each high-frequency sub-band histogram peak values of logging mode and the corresponding respectively coefficient value of time high point value, as shown in table 3.Indivedual subbands have high point two times, can select the high point of one of them time according to the principle that obtains the best visual effect, and promptly the coefficient number of being changed in the telescopiny is minimum.The peak of indivedual subbands and time high corresponding coefficient value are not adjacent, and centre 1 to 2 unit of being separated by is used as an integral body to peak and inferior high point and histogram between the two in this case and treats.Pattern 4 does not need to write down moving direction, because relatively peak just can judge that with the corresponding respectively coefficient value of time high point which side histogram outwards moves 1 unit or 2 units, sees synoptic diagram Fig. 4 (e1).Here all select under two kinds of embedded models all detail subbands promptly 9 sons bring row into and embed.When pattern embeds for 1 time, pattern 1, the detail subbands number of choosing 9, the coefficient value of the histogrammic peak correspondence of each high-frequency sub-band coefficient, the direction that histogram moves are recorded as side information A; When pattern embedded for 4 times, the coefficient value corresponding respectively pattern 4, the high-frequency sub-band number of choosing 9, the histogrammic peak of each high-frequency sub-band coefficient and inferior high point was recorded as side information A.Side information A adds the length that a data head shows A.
Table 1
Embedded model Pattern 1 Pattern 2 Mode 3 Pattern 4 Pattern 5
Embedding capacity (bit) 36215 70996 72430 107211 141992
Table 2
Subband HH1 HL1 LH1 HH2 HL2 LH2 HH3 HL3 LH3
Moving direction 0 1 0 1 1 1 0 1 0
Peak coefficient of correspondence value -3 -7 -2 -3 -7 -2 -3 -5 -3
Table 3
Subband HH1 HL1 LH1 HH2 HL2 LH2 HH3 HL3 LH3
Inferior high some coefficient of correspondence value -2 -6 -3 -1 -6 -1 -4 -7 -2
Peak coefficient of correspondence value -3 -7 -2 -3 -7 -2 -3 -5 -3
Again then with side information A according to from top to bottom, from left to right order replaces the lowest bit position LSB of the select location of low frequency sub-band, up to A is embedded in the low frequency sub-band fully.A has only replaced the partial L SB on low frequency sub-band lowest bit plane under two kinds of embedded models.The position of the LSB that is replaced is chosen with key, specific practice is that key is generated and the equirotal pseudo-random bits matrix of low frequency sub-band as seed, low frequency coefficient on the plain corresponding relevant position of entry of a matrix, when element is " 1 ", this position is selected and is used for built-in edges information, otherwise is not selected.The LSB of the low frequency sub-band that record is replaced is Ori_LSB.
Then, earlier with P M, Ori_LSB and form bit stream with hiding data M ' three parts after the secret key encryption, and add the data head, obtain bit stream B, wherein the data head comprises the three parts length separately of the total length and the composition bit stream B of bit stream, according to embedded model 1 and 4 bit stream B is embedded in 9 detail subbands respectively again, the embedding order is the same with the high-frequency sub-band order of selecting to embed, all, promptly be followed successively by HH1, HL1, LH1, HH2, HL2, LH2, HH3, HL3 and LH3 according to the characteristic of human visual system (HVS).When embed the hiding data that just continues to embed surplus when data reach maximum embedding capacity in order high-frequency sub-band, up to all data are embedded in the high-frequency sub-band in next high-frequency sub-band.
At last, anti-process according to above-mentioned formula group (1) and formula group (2), unite the high-frequency sub-band and the low frequency sub-band that have embedded data and carry out the image that three grades of integer wavelet reconstruct obtain reconstruct, whether the image of inspection reconstruct pixel takes place overflows, overflow if having, then increase the G value, restart successively according to above-mentioned steps; If do not overflow, then obtain the close image of covering under pattern 1 and the pattern 4 respectively, its visual effect is seen Fig. 6 (b) and 6 (c) respectively.Table 4 is performances of embedded model 1 and pattern 4 the present invention program when being issued to maximum embedding capacity.
Table 4
Test pattern (512 * 512 * 8) Embedding capacity (bpp) PSNR(dB)
Pattern 1 Pattern 4 Pattern 1 Pattern 4
Lena 0.1376 0.4090 43.20 39.44
The realization block diagram that Data Detection of the present invention, extraction and host image recover as shown in Figure 2, the detailed description of concrete steps face " technical scheme of the invention " part of seing before.At receiving end, the recovery of the detection of hiding data, extraction and host image is the inverse process of above-mentioned telescopiny.At first, the close image of covering that receives is carried out three grades of integer wavelets decomposition, obtain 9 high-frequency sub-band and 1 low frequency sub-band, and according to from top to bottom, from left to right order extracts some locational LSB of low frequency sub-band and obtains side information A, the LSB position of extracting is obtained as the bit matrix that seed generates by key, analyze side information A, obtain information such as corresponding respectively coefficient value of embedded model, the high-frequency sub-band number that is embedded with hiding data and peak thereof and/or inferior high point and the histogrammic moving direction of coefficient.
From high-frequency sub-band, extract the data of hiding then successively,, recover former high-frequency sub-band coefficient value simultaneously according to embedded model in the time of extraction up to bit stream B is extracted fully.Analyze bit stream B and obtain P M, Ori_LSB and with totally three partial datas of the hiding data M ' after the secret key encryption.
Then according to from top to bottom, order from left to right is with the LSB that is extracted that Ori_LSB replaces low frequency sub-band, recovers low frequency sub-band.With key the hiding data after encrypting is decrypted, obtains former hiding data M.
Then the high-frequency sub-band and the low frequency sub-band that have recovered of associating carries out three grades of integer wavelet reconstruct again, obtains the spatial domain histogram without image restored.
At last according to P MThe pixel value that recovery was revised obtains undistorted host image.

Claims (9)

1, a kind of at natural gray image, based on the reversible data concealing method of integer wavelet statistic characteristics, it is characterized in that utilizing the characteristics of the approximate Gaussian distributed of integer wavelet high-frequency sub-band coefficient of natural image, the histogram of modification high-frequency sub-band and the lowest order plane of low frequency sub-band are carried out the embedding of data and are hidden to realize reversible information, and this method is divided into data hidden and two processes of Data Detection.
2, reversible data concealing method according to claim 1 is characterized in that described data hidden process steps is as follows:
1) revises histogram: be about to host image X MxNHistogrammic scope is modified to [G, 255-G] from [0,255], the initial default value of G is 0 or according to X MxNHistogram is set, and record to be modified pixel be P M, comprise initial value, coordinate and G, remember that revised image is X ' MxN
2) integer wavelet decomposes: promptly to X ' MxNCarry out integer wavelet and decompose, obtain a low frequency sub-band LL LWith high-frequency sub-band set C K, l, calculating accessible maximum embedding capacity under five kinds of main embedded models according to histogrammic peak of each high-frequency sub-band coefficient and time high point, decomposed class is got L=3;
3) embedded model is selected: the maximum embedding capacity under hiding data amount and the five kinds of main embedded models relatively, select a kind of embedded model, and choose some high-frequency sub-band that are used to embed data according to human visual system (HVS) characteristic, the embedded model that record is selected, the high-frequency sub-band number of choosing, under this embedded model the selected histogrammic peak of each high-frequency sub-band and/or the coefficient value of inferior high correspondence and/or the direction that the coefficient histogram moves, the information of these records is designated as side information A;
4) low frequency coefficient is selectively modified: select low frequency sub-band LL with key K LThe LSB of coefficient (LeastSignificant Bit) embeds low frequency sub-band LL with side information A in the mode of replacing selecteed LSB LObtain LL ' L, and the LSB that record is replaced is Ori_LSB;
5) high frequency coefficient is revised: with P M, Ori_LSB and encrypt with key K after hiding data M ' form bit stream, and revise the high-frequency sub-band coefficient according to selected embedded model, bit stream is embedded in the high-frequency sub-band of choosing, the high-frequency sub-band set that is embedded with data is designated as C ' K, l
6) integer wavelet reconstruct: with LL ' LAnd C ' K, lCarry out three grades of integer wavelet reconstruct;
7) pixel value overflows judgement: if the pixel value after the reconstruct exceeds the scope of [0,255], be judged as and overflow, then return to step 1), increase the value of G and repeat above-mentioned steps until there not being pixel value to overflow, the step-length acquiescence of G gets 5 or set up on their own, otherwise has obtained hiding the close image X of covering of information " MxN
3, reversible data concealing method according to claim 1 is characterized in that described data detection process step is as follows:
1) integer wavelet decomposes: promptly to the close image X that covers to be detected " MxNCarry out three grades of integer wavelets and decompose, obtain a low frequency sub-band LL ' LWith high-frequency sub-band set C ' K, l
2) selective extraction: obtain LL ' by key K LIn be embedded with the LSB position of data, extract LSB from these positions and obtain side information A, analyze high-frequency sub-band number that side information A can obtain embedded model, choose, the selected histogrammic peak of each high-frequency sub-band and/or the coefficient value of inferior high correspondence and the direction that the coefficient histogram moves under this embedded model;
3) extract and recover high frequency coefficient: according to embedded model from C ' K, lMiddle extraction bit stream also recovers C K, l, analyze bit stream and obtain P M, Ori_LSB and M ', decipher M ' with key K and obtain M;
4) low frequency coefficient recovers: replace LL ' with Ori_LSB LIn be embedded with the LSB position of data, recover LL L
5) integer wavelet reconstruct: with LL LAnd C K, lCarry out three grades of integer wavelet reconstruct, obtain picture signal X ' MxN
6) recover histogram: according to P MWith X ' MxNHistogram [G, 255-G] revert to [0,255], obtain undistorted host image X MxN
4, reversible data concealing method according to claim 2, it is characterized in that described data hidden process steps 3) in selected embedded model determined embedding capacity, the hiding data of each high-frequency sub-band embedding, extract with recovery and this process in coefficient the histogram direction and the size that move, main embedded model has five kinds, respectively called after embedded model 1, embedded model 2, embedded model 3, embedded model 4 and embedded model 5.
5, reversible data concealing method according to claim 4 is characterized in that the embedding capacity of described embedded model 1 equals the histogrammic peak value of high-frequency sub-band coefficient n 1, the method that embedded model 1 embeds, detects, extracts and recovers is as follows:
1) at transmitting terminal, keeping under the constant situation of histogram peak, with the histogram peak wherein is the whole mobile unit of histogram on the left side or the right on one side, the direction that histogram moves promptly moves still to move to the right to the left side can follow the principle that as far as possible obtains best visual effect, the coefficient value of promptly keeping as much as possible is constant, make on histogram peak next door and a breach (gap) occurs, concrete realization is at first to add up the coefficient histogram, obtains the coefficient value c of histogram peak correspondence 1st, calculate the histogrammic coefficient number on peak both sides and compare, the coefficient on the note peak left side adds up to N Left, the coefficient on note peak the right adds up to N RightIf, N Left＜N Right, then the histogram on peak left side integral body outwards moves a unit, and promptly all coefficient values of this part histogram correspondence reduce by units, if N Left〉=N Right, then the histogram integral body on peak the right outwards moves a unit, and promptly all coefficient values of this part histogram correspondence increase units, suppose that comparative result is N Left〉=N Rigt, then to whole high-frequency sub-band carry out from top to bottom, from left to right sequential scanning, all coefficient values are added 1 (comparative result is then got on the contrary and subtracted, and next coming in order are analogized) greater than the coefficient value of histogram peak correspondence;
2) once more high-frequency sub-band is scanned according to same order, when running into peak coefficient c 1stThe time, the data that inspection will embed, if to be embedded is bit " 1 ", then the peak coefficient value is added 1, it is fill up the gap, if to be embedded is bit " 0 ", then keep the peak coefficient value constant, up to all data bits being embedded the maximum embedding capacity that finishes or reach this high-frequency sub-band;
3) at receiving end, according to same order high-frequency sub-band is scanned, when running into the peak coefficient, extract bit " 0 ", when the coefficient that runs into than peak coefficient value big 1, extract bit " 1 ", thereby and this coefficient value subtracted 1 revert to the peak coefficient value, even breach reappears;
4) once more high-frequency sub-band is scanned according to same order, all coefficient values are subtracted 1 greater than the coefficient value of histogram peak correspondence, promptly fill breach, recover whole coefficient subband.
6, reversible data concealing method according to claim 4 is characterized in that the embedding capacity of described embedded model 2 equals the histogrammic peak value of high-frequency sub-band coefficient n 1With inferior high point value n 2Sum, the method that embedded model 2 embeds, detects, extracts and recovers is as follows:
1) at transmitting terminal, keeping under the constant situation of histogram peak and time high integral body, with histogram peak and time high outside respectively whole unit that moves of histogram that both sides are the left side and the right, make at histogram peak and time this whole both sides of high point and occur two breach altogether, concrete realization is at first to add up the coefficient histogram, obtains the coefficient value c of histogram peak correspondence 1stWith inferior high corresponding coefficient value c 2ndAnd the size of two coefficient values relatively, then to whole high-frequency sub-band carry out from top to bottom, from left to right sequential scanning, if c 1st〉=c 2nd, then with all coefficient values greater than c 1stCoefficient value add 1, with all coefficient values less than c 2ndCoefficient value subtract 1 (comparative result is then got successively on the contrary and subtracted, adds, below similar);
2) once more high-frequency sub-band is scanned according to same order, when running into peak coefficient c1st, the data that inspection will embed, if to be embedded is bit " 1 ", then the peak coefficient value is added 1, if to be embedded is bit " 0 ", then keep the peak coefficient value constant, when running into time high dot factor c2nd, the data that inspection will embed, if to be embedded be bit " 1 ", then the peak coefficient value is subtracted 1, if to be embedded is bit " 0 ", then keep time high dot factor value constant, up to all data bits being embedded the maximum embedding capacity that finishes or reach this high-frequency sub-band;
3) at receiving end, according to same order high-frequency sub-band is scanned, when running into peak coefficient c1st or inferior high dot factor c2nd, extract bit " 0 ", when running into coefficient c1st+1 or c2nd-1, extract bit " 1 ", and coefficient value is reverted to c1st or c2nd;
4) once more high-frequency sub-band is scanned according to same order, all coefficient values are subtracted 1 greater than the coefficient value of c1st, all coefficient values are added 1 less than the coefficient value of c2nd, recover whole coefficient subband.
7, reversible data concealing method according to claim 4 is characterized in that the embedding capacity of described embedded model 3 equals the histogrammic peak value of high-frequency sub-band coefficient n 1Twice, embedded model 3 embeds, detects, extracts with the method for recovery as follows:
1) at transmitting terminal, keeping under the constant situation of histogram peak, outwards move a unit and another side whole outwards move two units with the histogram on histogram peak both sides is wherein whole on one side, make on histogram peak next door and two breach that size is respectively a unit and two units occur, concrete realization is at first to add up the coefficient histogram, obtains the coefficient value c of histogram peak correspondence 1st, calculate the histogrammic coefficient number on peak both sides and compare, the coefficient on the note peak left side adds up to N Left, the coefficient on note peak the right adds up to N RightIf, N Left＜N Right, then the histogram on the peak left side is whole outwards moves two units, and the histogram integral body on peak the right outwards moves a unit, if N Left〉=N Right, then opposite, suppose that comparative result is N Left〉=N Rigt, then to whole high-frequency sub-band carry out from top to bottom, from left to right sequential scanning, with all coefficient values greater than c 1stCoefficient value add 2, with all coefficient values less than c 1stCoefficient value subtract 1 (comparative result is then got successively on the contrary and subtracted, adds, below similar);
2) once more high-frequency sub-band is scanned according to same order, when running into peak coefficient c 1stThe time, continuous two bits that inspection will embed, four kinds of combinations and four coefficient value c of continuous two bits 1st-1, c 1st, c 1st+ 1 and c 1st+ 2 is corresponding one by one, according to bit combination to be embedded with c 1stBe revised as c 1st-1, c 1st+ 1 or c 1st+ 2, or keep c 1stConstant, up to all data bits being embedded the maximum embedding capacity that finishes or reach this high-frequency sub-band;
3), according to same order high-frequency sub-band is scanned, when running into c at receiving end 1st-1, c 1st, c 1st+ 1 or c 1stExtracted its corresponding bit combination at+2 o'clock, and coefficient value is reverted to c 1st
4) once more high-frequency sub-band is scanned according to same order, with all coefficient values greater than c 1stCoefficient value subtract 2, with all coefficient values less than c 1stCoefficient value add 1, recover whole coefficient subband.
8, reversible data concealing method according to claim 4 is characterized in that the embedding capacity of described embedded model 4 equals histogrammic high point value n of this high-frequency sub-band coefficient 2With twice peak value n 1Sum, essence are that time high point has adopted embedded model 1, peak to adopt embedded model 3.
9, reversible data concealing method according to claim 4 is characterized in that the embedding capacity of described embedded model 5 equals histogrammic high point value n of this high-frequency sub-band coefficient 2With peak value n 1The twice of sum, essence are that time high point and peak have all adopted embedded model 3.
CNA2006100363867A 2006-07-07 2006-07-07 Reversible data concealing method based on integer wavelet statistic characteristics Pending CN1885340A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CNA2006100363867A CN1885340A (en) 2006-07-07 2006-07-07 Reversible data concealing method based on integer wavelet statistic characteristics

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CNA2006100363867A CN1885340A (en) 2006-07-07 2006-07-07 Reversible data concealing method based on integer wavelet statistic characteristics

Publications (1)

Publication Number Publication Date
CN1885340A true CN1885340A (en) 2006-12-27

Family

ID=37583480

Family Applications (1)

Application Number Title Priority Date Filing Date
CNA2006100363867A Pending CN1885340A (en) 2006-07-07 2006-07-07 Reversible data concealing method based on integer wavelet statistic characteristics

Country Status (1)

Country Link
CN (1) CN1885340A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102136129A (en) * 2011-04-22 2011-07-27 中北大学 Method for hiding information based on curve cluster in bit plane of image
CN110047029A (en) * 2019-04-22 2019-07-23 广东工业大学 A kind of combination multilayer difference extension has the reversible information hidden method and device of contrast enhancing
CN111045069A (en) * 2019-12-11 2020-04-21 山东省科学院海洋仪器仪表研究所 Data correction method for seawater radionuclide detection

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102136129A (en) * 2011-04-22 2011-07-27 中北大学 Method for hiding information based on curve cluster in bit plane of image
CN102136129B (en) * 2011-04-22 2012-11-14 中北大学 Method for hiding information based on curve cluster in bit plane of image
CN110047029A (en) * 2019-04-22 2019-07-23 广东工业大学 A kind of combination multilayer difference extension has the reversible information hidden method and device of contrast enhancing
CN110047029B (en) * 2019-04-22 2023-02-10 广东工业大学 Reversible information hiding method and device with contrast enhancement by combining multi-layer difference expansion
CN111045069A (en) * 2019-12-11 2020-04-21 山东省科学院海洋仪器仪表研究所 Data correction method for seawater radionuclide detection

Similar Documents

Publication Publication Date Title
Shi et al. Lossless data hiding: Fundamentals, algorithms, and applications
Weng et al. Reversible watermarking based on invariability and adjustment on pixel pairs
Wang et al. Wavelet tree quantization for copyright protection watermarking
Yuan et al. An integer wavelet based multiple logo-watermarking scheme
Spaulding et al. BPCS steganography using EZW lossy compressed images
Keyvanpour et al. Robust dynamic block-based image watermarking in DWT domain
Divecha et al. Implementation and performance analysis of DCT-DWT-SVD based watermarking algorithms for color images
Chen et al. A simple and efficient watermark technique based on JPEG2000 codec
Dalal et al. A robust and imperceptible steganography technique for SD and HD videos
Temi et al. A robust image watermarking using multiresolution analysis of wavelet
Ahuja et al. All aspects of digital video watermarking under an umbrella
Ibrahim et al. Video multiple watermarking technique based on image interlacing using DWT
Malallah et al. Reversible color video watermarking scheme based on hybrid of integer-to-integer wavelet transform and Arnold transform
CN1885340A (en) Reversible data concealing method based on integer wavelet statistic characteristics
Raja et al. High capacity lossless secure image steganography using wavelets
Narasimhulu et al. A new SVD based hybrid color image watermarking for copyright protection using contourlet transform
Maity et al. An image watermarking scheme using HVS characteristics and spread transform
Mohammed et al. A new robust binary image embedding algorithm in discrete wavelet domain
Arivazhagan et al. Colour Image Steganography Using Median Maintenance
Agilandeeswari et al. A novel block based video in video watermarking algorithm using discrete wavelet transform and singular value decomposition
Shahid et al. A Non-Blind Watermarking Scheme for Gray Scale Images in Discrete Wavelet Transform Domain using Two Subbands
Negrat et al. Variable length encoding in multiple frequency domain steganography
Fan et al. ROI-based watermarking scheme for JPEG 2000
Kapadia et al. Secured Reversible matrix embedding based on dual image using Integer wavelet and Arnold Transform
Lee et al. Reversible data embedding for tamper-proof watermarks

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication