CN106408495A - High PSNR fragile watermarking method based on chaos theory - Google Patents

Abstract

The invention provides a high PSNR fragile watermarking method based on a chaos theory. The method includes the following steps of S1 embedding a water mark into an original image; and S2 extracting the water mark from the image embedded with the water mark, changing the position, and recovering the image. The high PSNR fragile watermarking method overcomes all the defects in the original scheme, and meanwhile the storage space required for the watermark does not increase, so that high PSNR can be maintained. The chaotic mapping is used for guaranteeing the safety; a block division theory is used for improving the change area and creating the storage space of recovering the data; and the recovery result of the target image is improved by the edge matching method based on VQ. An experiment shows that the method is safe and effective, and the recovery capability is satisfactory.

Description

A kind of high PSNR vulnerable watermark method based on chaology

Technical field

The present invention relates to technical field of image processing, especially a kind of high PSNR fragile watermark side based on chaology Method.

Background technology

Digital picture is one of modal Digital Media in daily life, but it is faced with a series of safe prestige The side of body, such as carries out image procossing, distorts, distributes by instrument.Watermark is usually utilized to protect the peace of image in this aspect Full property [1-5] ([1] Naderahmadian Y, Hosseini-Khayat S (2014) Fast and robust (rest image being decomposed based on QR is fast for watermarking in still images based on QR decomposition Fast robust watermarking) .Multimedia Tools and Applications 72 (3)：2597-2618；[2] Tsai HH, Jhuang YJ, Lai YS (2012) An SVD-based image watermarking in wavelet domain using A kind of SVR and PSO (wavelet field image watermark of use SVR and PSO based on SVD) .Applied Soft Computing 12(8)：2442-2453；[3] Weng SW, Pan JS (2014) Reversible watermarking based on Multiple prediction modes and adaptive watermark embedding (based on multiforecasting model and The reversible water mark that self-adapting watermarking embeds) .Multimedia Tools and Applications 72 (3)：3063-3083； [4] Wu X, Sun W (2013) Robust copyright protection scheme for digital images Using overlapping DCT and SVD (using the digital picture robust copyright protection regime of overlapping DCT and SVD) .Applied Soft Computing 13(2)：1170-1182；[5] Huo YR, He HJ, Chen F (2014) A semi- A kind of fragile image watermarking algorithm with two-stage detection (two step detections half Fragile Watermarking) .Multimedia Tools and Applications 72 (1)：123-149).When image because Some specific purposes are tampered, and not through the permission of original owner, it is tight that those fakements may result in a series of comparisons The consequence of weight.Therefore, vulnerable watermark method is suggested.The purpose of fragile watermark is the verity of authentication image and complete Property, tampering detection and positioning, and recover to be tampered content [6,7] ([6] De Vleeschouwer C, Delaigle JF, Macq B(2002)Invisibility and application functionalities in perceptual Watermarking an overview (perception watermark invisibility and application function summary), Proc.IEEE 90 (1)：64- 77；[7] Petitcolas FA, Anderson RJ, Kuhn MJ (1999) Information hiding-a survey (letter Breath hides summary) .Proc.IEEE 87 (7)：1062-1078).It is, in general, that existing vulnerable watermark method is imperceptiblely Embed some excessive datas of being made up of authentication information and recovery information in multimedia bearer.Then, authentication information is extracted Out it is used for detecting and positioning potential tampered region, and recover information then for reconstructing the original contents of tampered region.In early days The purpose of vulnerable watermark method simply identify place [8-12] ([8] Wong PW, Memon N (2001) that image is tampered Secret and public key image watermarking schemes for image authentication and Ownership verification (for the secret public key image watermark mechanism of image authentication and proprietary rightss checking), IEEE Trans.Image Process.10(10)：1593-1601；[9]Suthaharan S(2004)Fragile image Watermarking using a gradient image for improved localization and security is (crisp Improve positioning and safety using gradient image in weak watermark) .Pattern Recognit.Lett.25 (16)：1893- 1903；[10] Yang H, Kot AC (2006) Binary image authentication with tampering Localization by embedding cryptographic signature and block identifier is (by embedding Enter private signature and block identifier realize the bianry image authentication mechanism with tampering location function) .IEEE Signal Process.Lett.13(12)：741-744；[11] Lu H, Shen R, Chung FL (2003) Fragile Watermarking scheme for image authentication (for the fragile watermark mechanism of image authentication), Electron.Lett.39(12)：898-900；[12] Zhang X, Wang S (2007) Statistical fragile Watermarking capable of locating individual tampered pixels (can position single distorting The fragile watermark based on statistics of pixel) IEEE Sighal Process.Lett.14 (10)：727-730), recent fragility Water mark method then emphasizes to carry out recovery [13-19] ([13] Zhang X, the Wang S of picture material on the basis of tampering location (2009) Fragile watermarking scheme using a hierarchical mechanism is (using layering Fragile watermark mechanism) .Signal Process.89 (4)：675-679；[14] Zhang X, Wang S (2008) Fragile Watermarking with error-free restoration capability (the fragile water with anti-error recovery capability Print) .IEEE Trans.Multimedia 10 (8)：1490-1499；[15] Qin C, Chang CC, Chen PY (2012) Self-embedding fragile watermarking with restoration capability based on (band based on adaptive bit dispatch mechanisms recovers the self-embedding of function to adaptive bit allocation mechanism Fragile watermark) .Signal Process.92 (4)：1137-1150；[16] He HJ, Zhang JS, Chen F (2009) Adjacent-block based statistical detection method for self-embedding Watermarking techniques (based on the statistical detection method facing block in self-embedding watermark technology) .Signal Process.89(8)：1557-1566；[17] Zhang X, Wang S, Qian Z, Feng G (2011) Reference Sharing mechanism for watermark self-embedding (for the reference sharing mechanism of self-embedding watermark), IEEE Trans.Image Process.20(2)：485-495；[18] Qian Z, Feng G, Zhang X, Wang S (2011) Image self-embedding with high-quality restoration capability (possesses high quality resume The image self-embedding method of function) .Digit.Signal Prog.21 (2)：278-286；[19] Qian Z, Feng G (2010) Inpainting assisted self recovery with decreased embedding data (embedding data by Step is borrowed in image mending technology in the case of reducing and realizes Self-recovery) .IEEE Signal Process.Lett.17 (11)：929-932).The content in one region of image is usually embedded in the middle of the different region of another one.Evil is being detected After meaning is distorted, the information extracting from the reservation region being not tampered with can be used to recover the mainly interior of tampered region Hold.In view of recover function all highly useful under many circumstances, such as traffic accident identification and obtain from image some weight Want information, it is very important for possessing trustworthy Restoration Mechanism under the premise of ensureing preferable picture quality.

But, because embedded extra recovery information needs bigger memory space, embedded mobile GIS is difficult to avoid that can be changed The content of original image, and reduce the quality of protected significance map picture.Obviously, detection and the possessing and image of recovery capability The holding of high PSNR is conflict it is difficult to compromise selects.Managing high PSNR provides tampering detection simultaneously and recovers much new Emerging field is all extremely important, including medical file, military report, ecommerce etc..Being badly in need of a kind of watermark mechanism of design can be true Protect enough high PSNR keeping host image to greatest extent on the premise of function.

In order to meet both demands simultaneously, we are in document [20] ([20] Rawat S, Raman B (2011) A chaotic system based fragile watermarking scheme for image tamper detection A kind of (fragile watermark mechanism being detected for distorted image based on chaos system) .AEU- Int.J.Electron.Commun.65(10)：Inspiration a kind of improved fragile watermark based on chaos of lower proposition 840-847) Mechanism.In document [20], it is widely used in digital image information safety [21-25] ([21] Liu H, Wang X (2012) Image encryption using DNA complementary rule and chaotic maps is (using DNA Yu collection method The then image encryption method with chaotic maps) .Applied Soft Computing, 12 (5)：1457-1466；[22]Lian SG, Chen X (2011) Traceable content protection based on chaos and neural Networks (the traceable content protecting mechanism based on chaos and neutral net), Applied Soft Computing 11 (7)：4293-4301；[23] ZhangY, Xiao D (2012) Double optical image encryption using Discrete Chirikov standard map and chaos-based fractional random transform (makes The doublet optical image encryption of the fraction stochastic transformation with discrete Qi Likefu Standard Map with based on chaos) .Optics and Lasers in Engineering 51：472-480；[24] Yuen CH, Wong KW (2011) A chaos-based joint Image compression and encryption scheme using DCT and SHA-1 is (using DCT's and SHA-1 Compression and encryption mechanism are combined based on the image of chaos) .Applied Soft Computing 11 (8)：5092-5098；[25] Wang Y, Wong KW, Liao X, et al. (2011) A new chaos-based fast image encryption Algorithm (a new Fast video encryption algorithm based on chaos), Applied soft computing 11 (1)： Chaos system 514-522) be used in watermark embed and extract in.Why high PSNR ensure that is because in watermark mechanism In implementation process, only one of which LSB plane needs to change.But, there are three weak points in this scheme：1st, not safe.When When front 7 bit planes are tampered, tampering detection mechanism can lose efficacy；2nd, detection zone is sufficiently complete.Only half tampered region Pixel energy is detected, also just because of this leads to follow-up resuming work cannot be carried out；3rd, recovery algorithms are not provided.Closely Come, document [26] ([26] Teng L, Wang X, Wang X (2013) Cryptanalysis and improvement of a Chaotic system based fragile watermarking scheme is (to the fragile watermark mechanism based on chaos system Cryptanalysis with improve) .AEU-Int.J.Electron.Commun.67：540-547) improve document to a certain extent [20] safety issue.But, the scheme after improvement is still not safe.Further, since three LSB planes are all embedding in watermark It is modified during entering, the PSNR of host image greatly reduces.Furthermore, the detection of tampered region remains incomplete, And author does not provide recovery algorithms yet.It is proposed that method in, original binary watermarking is to rely on plaintext host Image；For improving tampering detection region and the embedded thought vacateed memory space and take that block divides for recovering data； And employ VQ technical limit spacing block index, information is recovered with pie graph picture；Then, raw water print and index are combined and are formed Final watermark, and the method replaced by LSB is embedded in host image.Thus it is possible, on the one hand, the scheme proposing has safety concurrently Property, tampering detection and recovery capability, recovery capability；On the other hand, the image of embedded watermark is made to have higher PSNR.

Original fragile watermark mechanism [20] is made up of two processes：Watermark embeds and watermark extracting.

First, watermark telescopiny is as follows：

(1) line shuffle is entered by k image I to size for M × N of iteration Arnold Cat maps, form the image of scramble I_scr.

(2) the image I after scramble_scrIt is divided into 8 bit planes.

(3) logical mappings are used to generate the chaos sequence S for m × n for the length.Chaos image pattern is obtained after rearrangement S_cp.

(4) watermark W and S for m × n by size_cpCarry out xor operation, that is,Obtain two-value chaotic watermarking W_p.

(5) use W_pReplace I_scrLeast significant bit planes.

(6) I after the modification_scrObtaining the image after embedded watermark, wherein T is for upper iteration Arnold Cat maps T-k time The cycle of Cat maps.

The process of watermark extracting is as follows：

(1) the image I that (4) step processes embedded watermark is arrived, according to similar to (1) in watermark telescopiny, in-(4)_WTo obtain The watermark W extracting^ext.

(5) obtain and extract watermark W^extAbsolute difference with original watermark W.Iteration Arnold Cat maps T-k time are embedding to position Enter the tampered region of watermarking images.

In former scheme [20], because the watermark of the scramble being embedded in LSB plane is not rely on original plaintext image, Change first 7 bit planes being not embedded into watermark will be detected in the image of embedded watermark.As Fig. 1 (a)-(f) institute Show, when all bit planes are modified simultaneously, watermark mechanism is working properly.But when only front 7 bit planes are tampered, usurp Change detection will lose efficacy, such as shown in Fig. 1 (g)-(i).

Recently, the author in document [26] claims that this deficiency can be overcome, and method is to set up watermark and original image The first two MSB plane contacts, i.e. in two-value chaotic watermarking W_pOne is performed and the first two MSB plane of original image between Extra xor operation, with generate one new for embedded watermark.If so any change is related to the first two MSB and puts down Face, i.e. the highest significant position plane of image, it will be detected.

But, if while the attacker's modification front 5 MSB planes of image (because rear 3 LSB planes embedded in watermark) Keeping the XOR value of the first two bit plane, constant (for example, a MSB plane is set to random value, and the value of the 2nd MSB plane is with regard to basis Between depending on the result of XOR and the random value of a MSB plane), then modification just cannot be detected, such as Fig. 1 (j)-(l) Shown.This just illustrates that this improved plan remains unsafe.In addition, new watermark has dispersedly been embedded into three LSB An only LSB plane in plane rather than original, so the PSNR value of the image of embedded watermark is certain to reduce.

In document [20], the tampered region such as Fig. 1 (f) successfully detecting is shown.It is filled with black and white in region Pixel, this is because the tampered region of half, that is, black picture element part, is not detected among out.The reason lead to this problem be Each bit in binary watermarking during distorting be modified to another bit probability be 1/2, unmodified value nor It is detected by taking original watermark and extract the absolute difference of watermark.Additionally, former scheme does not provide tampered region Recovery algorithms, document [26] is same.

In order to recover the content of tampered region, the recovery data obtaining from original plaintext image should store and will embed Watermark in the middle of.In scheme proposed by the invention, using vector quantization (VQ) technology [27,28] ([27] Gray R (1984) Vector quantization (vector quantization) .IEEE ASSP Magazine 1 (2)：4-29；[28] Linde Y, (vector quantizer sets Buzo A, Gray R (1980) An algorithm for vector quantizer design Calculating method) .IEEE Trans.Commun.28 (1)：84-95) obtain and recover data.VQ is a kind of classical lossy compression method skill Art, coding and decoding are simply effective.Fig. 2 gives the flow process of VQ coding and decoding.

In coding stage, original image is first separated into nonoverlapping size identical block.It is big that the present invention sets piecemeal Little is 4 × 4.So, every piece comprises 16 pixels, can be expressed as the vector of 16 dimensions, the V in such as Fig. 2_{I, 1}.The main thought of VQ It is all of vector of mapping in a finite aggregate, i.e. code book CB={ C₁, C₂..., C_L, wherein C_iRepresent i-th in code book CB Individual code word, L represents the size of code book.Each code word C_iIt is image most representational in some image block vector set Block vector set V_{I, 1}, V_{I, 2}..., V_{I, di}To be obtained by cluster, wherein, public index i represents the classification sequence number of vector, di table Show the element number in the i-th class vector.Firstly generate code book.Want coded image, for each vectorial V in original image, Can be determined according to the Euclidean distance as shown in formula (1) from its nearest code word in code book.

Wherein v_jAnd c_{I, j}Represent vectorial V and C respectively_iIn j-th element, k be vector dimension.When nearest code word C_iAfter finding, index i is just used for coding vector V.After institute's directed quantity all encodes, original image just can final index of reference Represent, each index correspondence code word nearest with this image block vector distance.

In decoding stage, index value is used for finding out corresponding code word from code book, then goes to represent originally with this code word Vector.So, original image just can be reconstructed.Because in compression process, image block vector is replaced by nearest with it Code word, original image content can distortion to a certain extent.Obviously, the VQ compression quality of image and the close phase of quality of code book Close, and code book preferably can be designed by the method for document [28,29].

Content of the invention

In view of this, the present invention provides a kind of high PSNR vulnerable watermark method based on chaology.

A kind of high PSNR vulnerable watermark method based on chaology, it comprises the steps：

S1, embedded watermark in original image；

S2, the figure embedded in watermark extracts watermark, tampering location successively and carries out image recovery.

Of the present invention based in the high PSNR vulnerable watermark method of chaology, described step S1 includes：

S11, to original image piecemeal and by VQ obtain recover data；

S12, the scramble of original binary watermarking and piecemeal；

S13, merging recover data and original watermark, obtain the bit plane that a size is 512 × 512, then it are entered Line shuffle；

S14, the front 7 MSB planes acquisition detection plane according to original image, and execute bit with the result of step S13 Final watermark is obtained to the XOR of bit；

S15, complete the embedded of watermark with the LSB plane that original image is replaced in final watermark.

Of the present invention based in the high PSNR vulnerable watermark method of chaology, in described step S2, extract watermark Including：

S21A, the extraction LSB plane of water mark inlaying image and detection plane；And the LSB of detection plane and extraction is put down Face carry out by turn xor operation obtaining the bit plane of scramble；

S22A, the bit plane iteration block-based Cat maps T to scramble₂-K₂Secondary recovering bit plane, wherein T₂It is this step The rapid Cat maps cycle；

Watermark bit is isolated in S23A, the bit plane obtaining from step S22A；And the scattered watermark ratio that will obtain Spy is recombinated according to mode from left to right line by line, obtains the watermark of the scramble that size is 362 × 362；

S24A, to the watermark iteration of scramble the Cat maps T based on pixel₁-K₁Secondary recovering watermark, wherein T₁It is this step The Cat maps cycle, thus obtaining the watermark that extracts, its size is 362 × 362.

Of the present invention based in the high PSNR vulnerable watermark method of chaology, tampering location in described step S2 Including：

S21B, take extract watermark original watermark absolute difference, obtain and watermark size 362 × 362 identical difference Plane；

S22B, in difference plane the Cat maps K based on pixel for the iteration₁The secondary difference plane obtaining scramble；

S23B, difference plane is divided into 16384 pieces, every piece is made up of 8 bits；

S24B, distort judgement：If there being any one bit to be 1 in a block, that is, extract block and initial condition in watermark The block of India and China is inconsistent, then this block is just identified as distorting；All of piece all judges in this way, afterwards, each Block is extended to 4 × 4 size, and in block all bits value all same, whether its value be tampered determination by this block；Then, institute Some extension blocks are from left to right reset line by line, form the difference plane of extension, and size is 512 × 512；

S25B, extension difference plane on iteration block-based Cat maps K₂Secondary, obtain 512 × 512 tampered region.

Of the present invention based in the high PSNR vulnerable watermark method of chaology, in described step S2, image recovers Including：

S21C, all it is identified as distorting if there are being not detected by least 3 adjacent blocks around the block distorted, then this block Also it is identified as distorting block；The tampered region being thus improved, its size is 512 × 512；

S22C, on improved tampered region iteration block-based Cat maps T₂-K₂The secondary tampered region obtaining scramble；Right According to improved tampered region, can find possess following characteristics do not distort block：Do not distorted improved before Cat maps Region, and be moved to after Cat maps in improved tampered region；The recovery data being stored in these blocks be not by Distort, can be used for recovering corresponding block in improved tampered region；

S23C, by VQ recover distort block, obtain initial restoration result, size be 512 × 512；

If S24C distorts block full recovery after step S23C, just export the final image recovering, otherwise turn Mate recovery to side.

For fragile watermark mechanism, the high PSNR of host image is kept preferably to recover function itself with providing effect Be exactly conflict because the former needs embedding information in host image as few as possible, and the latter then need as many as possible Host image information is embedded in itself on ground.The present invention preferably solves this contradictory problems, is keeping host image Provide the very gratifying image of effect in the case of high PSNR and recover function.In watermark telescopiny, only use place The LSB plane of master image is entered row information and is embedded, hereby it is ensured that the high PSNR characteristic of host image.Then, the difficult point of problem is just It is how to lead to too small amount of information acquisition tampering detection, positioning and image recover function.The present invention adopts the thought of piecemeal, makes Obtain the recovery data of block image with VQ technology, by setting the size of code book and the size of piecemeal, recover data with Tampering detection and positioning data compression have arrived in each LSB piecemeal, and ensure that piecemeal has the following characteristics that：(1) possess and distort Detection function, false drop rate is low；(2) possesses tampering location function, tampered region is complete and edge sawtooth is inconspicuous；(3) it is provided that Sufficient recovery data requisite space；(4) the recovery data being stored possesses preferable recovery effects.Chaotic maps technology is at this Primarily to ensureing safety in invention, it is mainly reflected in following place：(1) recover data and former in S13 and S23A During beginning watermark data is reset after merging, produce a series of randoms number according to chaotic maps, by the ranking results of random number Determine putting in order of piecemeal Nepit；(2) obtain in detection planarization process in S14 and S21A, being produced by chaos system is needed To enter the random matrix of row operation with 7 MSB planes, 8 detecting positions different in thus obtaining every piece, so that LSB plane Effectively set up association with front 7 MSB planes；(3) realized based on piecemeal by Cat maps in S12, S13, S22B and S25B Scramble, in order to realizing inverting unrest according to the periodicity of Cat maps in S22A, S24A and S22C, and then can coordinate this In invention, embedded, extractions, tampering detection, positioning and the image recovery of watermark is smoothly implemented in block-based operation.Recover in image During, due to Cat maps it cannot be guaranteed that all recovery information distorting block is all not tampered with, that is, distort the corresponding original VQ of block Code word has not necessarily been saved in just by Cat maps and has not distorted among block, so the initial restoration result in S23C is not necessarily complete Whole.Finiteness based on code word in VQ code book and the spatial correlation of natural image pixel, the present invention is using the side of many wheels Join restoration methods and ensure that all tampered regions can be restored.In a word, the present invention is keeping the high PSNR of former scheme [20] On the basis of characteristic, it is used in combination the technology such as piecemeal, VQ, chaotic maps and side coupling, there is provided a safety is secure, and Possesses the new vulnerable watermark method recovering more by force function.Experiment shows our method safety effectively, and recovery capability is also non- Often satisfactory.

Brief description

Fig. 1 is the attack process displaying figure that previous work insecurity is described.Fig. 1 (a) is the displaying figure of original image；Figure 1 (b) is the displaying figure of original watermark；Fig. 1 (c) is the displaying figure of embedded watermarking images；Fig. 1 (d) is that all bit planes are all distorted Image shows figure；Fig. 1 (e) is that figure is shown in the watermark extracted from Fig. 1 (d)；Fig. 1 (f) detects from Fig. 1 (d) Figure is shown in tampered region；Fig. 1 (g) is the image shows figure that front 7 MSB bit planes are distorted；Fig. 1 (h) is to extract from Fig. 1 (g) Watermark show figure；Fig. 1 (i) is the tampered region displaying figure detecting from Fig. 1 (g)；Fig. 1 (j) is to be improved using follow-up The image shows figure that scheme [26] is distorted in the case that 2MSB plane XOR value is constant before holding at random；Fig. 1 (k) is to use [26] figure is shown in the watermark that the method in is extracted from Fig. 1 (j)；Fig. 1 (l) is that the method using [26] detects from Fig. 1 (j) Figure is shown in the tampered region come；

Fig. 2 is the schematic diagram of VQ coding and decoding；

Fig. 3 is the flow chart of watermark telescopiny；

Fig. 4 is the flow chart that watermark extracting, tampering location and image recover；

Fig. 5 is the complete procedure displaying figure that the present invention is implemented：Fig. 5 (a) is that the original image that size is 512 × 512 is shown Figure is the same with Fig. 1 (a)；Fig. 5 (b) is that the original watermark that size is 362 × 362 shows figure；Fig. 5 (c) is the image of embedded watermark Show figure (PSNR=51.15)；Fig. 5 (d) is the image shows figure distorted；Fig. 5 (e) is the watermark displaying figure extracting；Fig. 5 (f) It is the displaying figure of tampered region；Fig. 5 (g) is improved tampered region displaying figure；Fig. 5 (h) is the displaying figure initially recovering image (PSNR=34.58)；Fig. 5 (i) is the displaying figure (PSNR=44.61) finally recovering image；

Fig. 6 is the image recovery process displaying figure of the tampered region that reply is located at center：Fig. 6 (a) is 58% figure distorted As showing figure；Fig. 6 (b) is the watermark displaying figure extracting；Fig. 6 (c) is tampered region displaying figure；Fig. 6 (d) is improved to distort area Domain views figure；Fig. 6 (e) is the displaying figure (PSNR=9.91) initially recovering image；Fig. 6 (f) is the displaying finally recovering image Figure (PSNR=28.43)；

Fig. 7 is the image recovery process displaying figure of the tampered region that reply is related to entire image：Fig. 7 (a) is 48% to distort Image shows figure；Fig. 7 (b) is the watermark displaying figure extracting；Fig. 7 (c) is the displaying figure of tampered region；Fig. 7 (d) is improved Figure is shown in tampered region；Fig. 7 (e) is the displaying figure (PSNR=11.21) initially recovering image；Fig. 7 (f) is finally to recover image Displaying figure (PSNR=30.98)；

Fig. 8 is that the attack resistance process of the attack method that reply is applied to document [20] and [26] shows figure：Fig. 8 (a) is to usurp Change the image shows figure of front 7 MSB planes；Fig. 8 (b) is that figure is shown in the watermark extracted from Fig. 8 (a)；Fig. 8 (c) is from Fig. 8 The tampered region displaying figure detecting in (a)；Fig. 8 (d) is random in the case that the front 2 MSB plane XOR values of holding are constant The image shows figure distorted；Fig. 8 (e) is that figure is shown in the watermark extracted from Fig. 8 (d)；Fig. 8 (f) is to detect from Fig. 8 (d) Tampered region show figure；

Fig. 9 is that reply replicates the image recovery process displaying figure pasting attack：Fig. 9 (a) is the displaying figure of tampered image；Figure 9 (b) is the watermark displaying figure extracting；Fig. 9 (c) is the displaying figure of tampered region；Fig. 9 (d) is improved tampered region displaying figure； Fig. 9 (e) is the displaying figure (PSNR=42.57) initially recovering image；Fig. 9 (f) is the displaying figure (PSNR finally recovering image =47.24)；

Figure 10 is that reply word increases the image recovery process displaying figure attacked：Figure 10 (a) is the displaying figure of tampered image； Figure 10 (b) is the watermark displaying figure extracting；Figure 10 (c) is the displaying figure of tampered region；Figure 10 (d) is the exhibition of improved tampered region Diagram；Figure 10 (e) is the displaying figure (PSNR=20.47) initially recovering image；Figure 10 (f) is the displaying figure finally recovering image (PSNR=37.69)；

Figure 11 is the image recovery process displaying figure that reply content removes attack：Figure 11 (a) is the displaying figure of tampered image； Figure 11 (b) is the watermark displaying figure extracting；Figure 11 (c) is the displaying figure of tampered region；Figure 11 (d) is the exhibition of improved tampered region Diagram；Figure 11 (e) is the displaying figure (PSNR=40.24) initially recovering image；Figure 11 (f) is the displaying figure finally recovering image (PSNR=43.98)；

Figure 12 is the image recovery process displaying figure that attack is pieced in reply together：Figure 12 (a) be size be 512 × 512 embedded The Lena image shows figure of watermark；Figure 12 (b) is the Milkdrop image shows figure of the embedded watermark that size is 512 × 512；Figure 12 (c) is the Lena image shows figure distorted；Figure 12 (d) is that figure is shown in the watermark extracted from Figure 12 (c)；Figure 12 (e) is to usurp Change the displaying figure in region；Figure 12 (f) is improved tampered region displaying figure；Figure 12 (g) is the displaying figure initially recovering image (PSNR=24.51)；Figure 12 (h) is the displaying figure (PSNR=36.79) finally recovering image.

Specific embodiment

The principle of the present invention is as follows,

The present invention provides a kind of high PSNR vulnerable watermark method based on chaology, and it comprises the steps：

S1, embedded watermark in original image；

S2, the figure embedded in watermark extracts watermark, tampering location successively and carries out image recovery.

Alternatively, of the present invention based in the high PSNR vulnerable watermark method of chaology, described step S1 bag Include：

S11, to original image piecemeal and by VQ obtain recover data；

S12, the scramble of original binary watermarking and piecemeal；

S13, merging recover data and original watermark, obtain the bit plane that a size is 512 × 512, then it are entered Line shuffle；

S14, the front 7 MSB planes acquisition detection plane according to original image, and execute bit with the result of step S13 Final watermark is obtained to the XOR of bit；

S15, complete the embedded of watermark with the LSB plane that original image is replaced in final watermark.

Alternatively, of the present invention based in the high PSNR vulnerable watermark method of chaology, in described step S2 Extract watermark to include：

S21A, the extraction LSB plane of water mark inlaying image and detection plane；And the LSB of detection plane and extraction is put down Face carry out by turn xor operation obtaining the bit plane of scramble；

S22A, the bit plane iteration block-based Cat maps T to scramble₂-K₂Secondary recovering bit plane, wherein T₂It is this step The rapid Cat maps cycle；

Watermark bit is isolated in S23A, the bit plane obtaining from step S22A；And the scattered watermark ratio that will obtain Spy is recombinated according to mode from left to right line by line, obtains the watermark of the scramble that size is 362 × 362；

S24A, to the watermark iteration of scramble the Cat maps T based on pixel₁-K₁Secondary recovering watermark, wherein T₁It is this step The Cat maps cycle, thus obtaining the watermark that extracts, its size is 362 × 362.

Alternatively, of the present invention based in the high PSNR vulnerable watermark method of chaology, in described step S2 Tampering location includes：

S21B, take extract watermark original watermark absolute difference, obtain and watermark size 362 × 362 identical difference Plane；

S22B, in difference plane the Cat maps K based on pixel for the iteration₁The secondary difference plane obtaining scramble；

S23B, difference plane is divided into 16384 pieces, every piece is made up of 8 bits；

S24B, distort judgement：If there being any one bit to be 1 in a block, that is, extract block and initial condition in watermark The block of India and China is inconsistent, then this block is just identified as distorting；All of piece all judges in this way, afterwards, each Block is extended to 4 × 4 size, and in block all bits value all same, whether its value be tampered determination by this block；Then, institute Some extension blocks are from left to right reset line by line, form the difference plane of extension, and size is 512 × 512；

S25B, extension difference plane on iteration block-based Cat maps K₂Secondary, obtain 512 × 512 tampered region.

Alternatively, of the present invention based in the high PSNR vulnerable watermark method of chaology, in described step S2 Image recovers to include：

S21C, all it is identified as distorting if there are being not detected by least 3 adjacent blocks around the block distorted, then this block Also it is identified as distorting block；The tampered region being thus improved, its size is 512 × 512：

S22C, on improved tampered region iteration block-based Cat maps T₂-K₂The secondary tampered region obtaining scramble；Right According to improved tampered region, can find possess following characteristics do not distort block：Do not distorted improved before Cat maps Region, and be moved to after Cat maps in improved tampered region；The recovery data being stored in these blocks be not by Distort, can be used for recovering corresponding block in improved tampered region；

S23C, by VQ recover distort block, obtain initial restoration result, size be 512 × 512；

If S24C distorts block full recovery after step S23C, just export the final image recovering, otherwise turn Mate recovery to side.

Below it is described in detail：

1. the detailed description of specific embodiment

S1 embeds watermark in original image

The flow process that watermark embeds is as shown in Figure 3.

S11 obtains recovery data to original image piecemeal and by VQ

Original image is first subdivided into nonoverlapping piece, and tampered region will be detected based on block.As appointing in fruit block What bit is modified, then this block is it is assumed that be to be tampered.Because a bit is tampered as in addition in the case of random The probability of one value is 1/2, if only with a bit, to carry out tampering detection, false drop rate can be up to 1/2.Obviously, If have n bit for doing tampering detection in a block, false drop rate will be reduced to 1/2ⁿ, false drop rate is with the increase of n Exponentially reduces.In the present invention, original image only has last LSB plane for embedding watermark, for storing recovery number According to memory space can only be provided by reducing the method for original binary watermarking (as shown in Fig. 1 (b)), and former binary watermarking is then It is used to carry out tampering detection.That is, tampered region must by little bit in the middle of each piece of LSB plane Lai Detection.Therefore, it is necessary for carrying out tampering detection based on block, and it is also necessary for carrying out piecemeal to original image.If piecemeal is too Greatly, the jagged edges of tampered region can reduce detection quality；If piecemeal is too little, the storage recovering needed for data can cannot be provided Space, and false drop rate is as the raising for detecting the bit distorted very few of block memory storage.

In the present invention, piecemeal size is set as 4 × 4.16 bits are had in each block of LSB plane, they It is used for embedding original watermark and recover data.By our experiment, when code book is when being sized to 256, the final figure recovering As quality is gratifying, and the memory space recovering required for data is only 8 bits (2⁸=256).Other 8 ratios in block Spy is then used for embedding original watermark.False drop rate is 1/2⁸≈ 0.39%.The area of original watermark is the half of original image, but it regards Feel that effect is good enough.If the size of original image is N × N, then block number is (N/4)².In the present invention, do not lose one As property, N is set as 512, then block number is 16384.So, 16384 recovery data will be had after VQ is processed, each is extensive Complex data is 8 bits.

The scramble of the original binary watermarking of S12 and piecemeal

The area of original binary watermarking isOriginal watermark is logical first Cross the Cat maps K based on pixel for the iteration₁Secondary enter line shuffle, be then further subdivided into (N/4)²=16384 blocks.Block number and original graph Block number as dividing is identical.

Original watermark 362 × 362=131044 bit altogether, being available for the embedded bit number of original watermark in original image is 16384 × 8=131072.When carrying out piecemeal to original watermark, we from left to right select pixel line by line.For first 28 Block, one group of 7 bits；To block later, one group of 8 bits.8th bit of front 28 blocks is set to 0.So, initial condition Print has just been extended to 131044+28=131072 bit, for being embedded.The false drop rate of front 28 blocks is 1/2⁷≈ 0.78%.I.e. in 16384 blocks, the false drop rate of the block of 28/16384=0.17% improves 0.78%-0.39%=0.39%, Can ignore.

S13 merges recovery data and original watermark, obtains the bit plane that a size is N × N=512 × 512, then right It enters line shuffle.

Because original watermark is identical with the block number recovering data, it is new that these blocks one-to-one can merge formation 16384 blocks.For the block of each new formation, 16 bits therein are extensive by the original watermark information of 8 bits and 8 bits Complex data is constituted.16 bit rearrangement in every piece are formed 4 in bit planes (identical with original image size, that is, 512 × 512) × 4 block.After obtaining bit plane, by iteration block-based Cat maps K₂Secondary bitplanes enter line shuffle.

Rearrangement can according to a series of randoms number are ranked up realize.A series of randoms number can be given birth to by chaos system Become, its logical mappings is as follows：

q_n+1=α q_n(1-q_n), q_n∈ (0,1), (2)

When parameter alpha ∈ (3.5699456,4] when, system is chaos state.So initial parameter α and initial value q₀Can make For session key.In the present invention, if α=4.

S14 obtains detection plane according to front 7 MSB planes of original image, and executes bit pair with the result of step S13 The XOR of bit is obtaining final watermark.

For solving the safety issue in [20] and [26], it is proposed that new method is setting up watermark and original graph Contact as front 7 MSB planes.To each 4 × 4 block, all bits (16 × 7=112 bit) of front 7 MSB planes are first First it is extracted, a part of mutual XOR that then we randomly choose in 112 bits generates a detecting position.This operation Execution is used for for 8 times generating 8 different detecting positions in a block.Can be realized with formula (3).

S=(R × M) mod2, (3)

Wherein, M represents a column vector, is made up of 112 bits extracting from 7 MSB planes.R represents one greatly Little be 8 × 112 random matrix, its element value non-zero i.e. 1.S then represents the column vector being made up of 8 different detecting positions.Formula (2) the chaos system logical mappings described by can be used to generate the element in R.In the present invention, quantization threshold is set to 0.5, If the random number that i.e. formula (2) generates is more than 0.5, it is quantified as 1, is otherwise quantified as 0.Size is 512 × 512 detection Plane is initialized to full 0 first, and then detection bit in the 8 of each block is placed on the bit plane (knot of (3) with scramble On the corresponding position in the position of original watermark in really).Then, detection plane just obtains.Finally, the knot of detection plane and (3) Fruit carries out XOR by turn and obtains the final watermark that size is 512 × 512.

The S15 LSB plane that original image is replaced in final watermark completes the embedded of watermark.

S2 extracts watermark, tampering location in the figure embedded in watermark successively and carries out image recovery

The flow process that watermark extracting, tampering location and image recover is as shown in Figure 4.In figure has 3 row.Left-hand line is mainly water The process that print is extracted, will be described in S2A part, it is similar with the inverse process of watermark telescopiny in Fig. 3.Middle column Mainly follow-up tampering location process, will be described in S2B part, it is similar with watermark telescopiny in Fig. 3.Right Side row are mainly based upon the process that improved tampered region carries out image recovery, will be described with it in S2C part.

S2A watermark extracting

S21A extracts the LSB plane of water mark inlaying image and detection plane.Detection plane can be in water mark inlaying image Upper use obtains with step S14 identical method.Because not changing front 7 MSB of original image in watermark telescopiny Plane, so the detection plane obtaining on water mark inlaying image is identical with obtain on the original image, unless water mark inlaying Image is tampered.Next, to detection plane and extract LSB plane carry out xor operation by turn to obtain scramble position put down Face, corresponding with the bit plane of the scramble in Fig. 3.

The bit plane iteration block-based Cat maps (T to scramble for the S22A₂-K₂) secondary recovering bit plane, wherein T₂It is this The Cat maps cycle of step.

Watermark bit is isolated, the wherein position of watermark bit can lead in the bit plane that S23A obtains from step S22A Cross the method the same with step S13 to obtain, and the scattered watermark bit obtaining is carried out according to mode from left to right line by line Restructuring, obtains the watermark of the scramble that size is 362 × 362, similar to the inverse mistake of the original watermark piecemeal described in step S12 Journey.

S24A Cat maps (T based on pixel to the watermark iteration of scramble₁-K₁) secondary recovering watermark, wherein T₁It is this step In the Cat maps cycle, thus obtaining the watermark extracting, its size is 362 × 362.

S2B tampering location

S21B takes the absolute difference of the original watermark extracting watermark, obtains and watermark size (362 × 362) identical difference Plane.

S22B iteration Cat maps K based on pixel in difference plane₁The secondary difference plane obtaining scramble.

S23B is divided into 16384 pieces difference plane, and every piece is made up of 8 bits, similar to the piecemeal mistake described by step S12 Journey.

S24B distorts judgement：If there being any one bit to be 1 in a block, that is, extract block and original watermark in watermark In block inconsistent, then this block is just identified as distorting；All of piece all judges in this way, afterwards, each piece Be extended to 4 × 4 size, and in block all bits value all same, whether its value be tampered determination by this block；Then, own Extension blocks from left to right reset line by line, formed extension difference plane, size be 512 × 512.

S25B iteration block-based Cat maps K in the difference plane of extension₂Secondary, obtain 512 × 512 tampered region.

S2C image recovers

It is extensive that the individual not undetected tampered region (0.39%, illustrate in step s 11) of S21C may result in image Multiple Quality Down, so tampered region needs to improve in advance before image recovers.Tampered region in a small range is generally Continuous and smooth, therefore all it is identified as distorting if there are being not detected by least 3 adjacent blocks around the block distorted, then This block is also identified as distorting block；The tampered region being thus improved, its size is 512 × 512.

The S22C block-based Cat maps of iteration (T on improved tampered region₂-K₂) the secondary tampered region obtaining scramble；Right According to improved tampered region, can find possess following characteristics do not distort block：Do not distorted improved before Cat maps Region, and be moved to after Cat maps in improved tampered region；The recovery data being stored in these blocks be not by Distort, can be used for recovering corresponding block in improved tampered region.

S23C recovers to distort block by VQ, obtains initial restoration result, and size is 512 × 512.

If S24C distorts block full recovery after step S23C, just export the final image recovering, otherwise turn Mate recovery to side.

The principle that side coupling is recovered is the flatness based on natural image.On the adjacent side of two adjacent blocks between pixel definitely Difference sum is considered as the matching degree of this two blocks.The less explanation of value two is more mated.Obviously, based on statistics, correctly The matching degree of two adjacent blocks be higher than two incoherent piece of matching degree.One block has 4 adjacent blocks.If distorting block week It is with 1-4 adjacent block not to be tampered or recover, and the block of adjacent block coupling just can find from code book, and can use To recover to distort block.If block that surrounding is not not tampered with or recovery, then middle distort block in epicycle just not Can recover.

If all of block of distorting is recovered after the coupling recovery of epicycle side terminates, just export the final image recovering, Otherwise pass again to side coupling to recover, distort till block is all restored until all of.

2. experimental result

PSNR is the leading indicator weighing picture quality, and formula (4) is shown in its definition.

Wherein, MSE is original image I and by attack graph pictureMean square error, its definition is shown in formula (5).

Wherein, H and W is the height and width of image respectively.

(1) the full implementation process of the inventive method

Test image is 512 × 512 Lena.As shown in Figure 5.Fig. 5 (a) is original image, identical with Fig. 1 (a)；Fig. 5 B () is original watermark, size is 362 × 362, is the half of Fig. 1 (b)；Fig. 5 (c) is the image (PSNR=of embedded watermark 51.15).Because only that a LSB plane is changed in watermark telescopiny, so the PSNR value of the image of embedded watermark is relatively High；Fig. 5 (d) is the image distorted；Fig. 5 (e) is the watermark extracted, and central noise is to be caused by distorting of Fig. 5 (d)；Fig. 5 F () is tampered region, have 3 nd to distort block；Fig. 5 (g) is improved tampered region, nd in Fig. 5 (f) Distort what block was identified as distorting；Fig. 5 (h) is that some distort area based on the initial image (PSNR=34.58) recovering of Fig. 5 (g) Domain is not resumed；Fig. 5 (i) is the final image (PSNR=44.61) recovering, and all of tampered region is recovered.

Test further below the performance of proposed scheme to be described.The region that Fig. 6 (a) gives 58% is tampered Embedded watermarking images (Fig. 5 (c))；Fig. 6 (b) is the watermark extracted, and logo is still visible；Fig. 6 (c) is tampered region, has Three do not detect distort block；Fig. 6 (d) is improved tampered region, and the block of distorting not detected in Fig. 6 (c) is all recognized Determine not distort block；Fig. 6 (e) is initially to recover image (PSNR=9.91) based on Fig. 6 (d), and some tampered regions cannot be recovered； Fig. 6 (f) is the final image (PSNR=28.43) recovering, and all of tampered region is recovered.Fig. 7 gives another The tampered region of shape.Similar with Fig. 6, the tampered region in Fig. 7 (a) is 48%；Fig. 7 (b) is the watermark extracted；In Fig. 7 (c) Nd block number of distorting is 5, and 4 therein all are identified as distorting block in improved tampered region (Fig. 7 (d))；? The PSNR of the image (Fig. 7 (e)) just recovering is 11.21；The PSNR of the final image (Fig. 7 (f)) recovering is 30.98.

(2) safety

During solution security is tested, test image, original watermark, the image of embedded watermark are still distinguished For Fig. 5 (a), Fig. 5 (b) and Fig. 5 (c).

1. reply is applied to the attack method of document [20] and [26]

In document [20], if distorting generation in front 7 MSB planes, distorting and just cannot detect, such as Fig. 1 (g)- Shown in (i).In our scheme, this problem does not exist, sees Fig. 8 (a)-(c).Equally, the safety in document [26] Problem (Fig. 1 (j)-(l)) does not also exist in this programme, sees Fig. 8 (d)-(f).

In addition, implementing 4 kinds of attack methods in document [20] and [26]：Replicate and paste attack, text increase, content shifting Remove and piece attack together.We also provide the performance under our scheme is attacked at this 4 kinds below.

2. reply replicates to paste and attacks

As shown in Fig. 9 (a), Lena little perianth below replicates has pasted 7 times；Fig. 9 (b) is the watermark extracted；Distort area Domain can be successfully detected, shown in such as Fig. 9 (c) and (d)；Fig. 9 (e) is initially to recover image (PSNR=42.57), partly distorts Block fails to recover；Fig. 9 (f) is the final image (PSNR=47.24) recovering, and all of block of distorting is recovered.

3. reply text increases and attacks

As shown in Figure 10 (a), text " LENA " increases in image；Figure 10 (b) is the watermark extracted；Figure 10 (c) and D () is tampered region respectively and improves tampered region；Figure 10 (e) is initially to recover image (PSNR=20.47)；Figure 10 (f) is Final recovery image (PSNR=37.69).

4. reply content removes attack

As shown in Figure 11 (a), Lena medicated cap top is removed；Shown in the watermark such as Figure 10 (b) extracting；Figure 11 (c) and D () is tampered region respectively and improves tampered region；Figure 11 (e) is initially to recover image (PSNR=40.24)；Figure 11 (f) is Initial recovery image (PSNR=43.98).

5. attack is pieced in reply together

Figure 12 (a) is the Lena image of embedded watermark, and Figure 12 (b) is the Milkdrop image of embedded watermark.Keeping phase Position is combined in the case of constant with the different piece of two embedded watermarking images, just constitutes the collage picture of forgery, such as scheme Shown in 12 (c).Figure 12 (d) is the watermark extracted from Figure 12 (c)；It is some undetected that Figure 12 (e) is that tampered region includes Distort block；Figure 12 (f) is improved tampered region；Figure 12 (g) is initially to recover image, PSNR=24.51；Figure 12 (h) is Recover image, PSNR=36.79 eventually.

(3) image restorability

Due to failing to find suitable the contrasted scheme using 1 LSB plane, so we select Lee's and Lin Scheme [29] ([29] Lee TY, Lin SD (2008) Dual watermark for image tamper detection and Recovery (for the doubly time series model of distorted image, detection and recovery) .Pattern Recognit.41 (11)：3497- 3506), scheme [30] ([30] Yang CW, Shen JJ (2010) Recover the tampered of Yang and Shen Image based onVQ indexing (indexed based on VQ and recover tampered image) .Signal Process.90 (1)：331- 343) contrasted.All employ 3 LSB plane storage watermark datas in this two documents, and our scheme only makes With 1 LSB plane (ensure that embedded watermarking images have high PSNR value).Table 1 lists and carries out in different size and position The PSNR value of the recovery image after distorting, test image is 256 × 256 Lena.Obviously, our scheme is better than Lee and Lin Scheme, especially in the case of tampered region is less.For the scheme of Yang and Shen, when tampered region is in 49%- Our scheme to be slightly better than when 61%, but embed watermark data using 3 LSB planes and can store more recovery numbers According to being remarkably contributing to image and recover, also cannot ensure that embedded watermarking images have high PSNR value simultaneously.When tampered region exists 2.4%-41%, our scheme shows better performance.

Table 1. recovers the PSNR value of image

Test image：Lena(256×256)

3. conclusion

Propose an effective fragile watermark mechanism in the present invention, tampering detection, positioning and image can be carried out and recover. Scheme is subject to the inspiration of document [20], in order to ensure the high PSNR of image, only uses 1 LSB plane of original image to embed watermark number According to.Meanwhile, improve tampering detection region by way of piecemeal.By means of VQ technology, watermark devises image and recovers Function, and restoration result is further improved by side matching process.In addition, watermark is set up by the method based on chaotic maps Associating between original image, the safety issue in document [20] and its follow-up improvement project [26] is in our scheme All achieve a solution.Experiment shows, effectively and safety, and image restorability compared for using 3 LSB our scheme After plane carries out the embedded document [29] of watermark and [30], result is very satisfactory.

It is understood that for the person of ordinary skill of the art, can be done with technology according to the present invention design Go out other various corresponding change and deformation, and all these change the protection model that all should belong to the claims in the present invention with deformation Enclose.

Claims (5)

1. a kind of high PSNR vulnerable watermark method based on chaology is it is characterised in that it comprises the steps：

S1, embedded watermark in original image；

S2, the figure embedded in watermark extracts watermark, tampering location successively and carries out image recovery.

2. the high PSNR vulnerable watermark method based on chaology as claimed in claim 1 is it is characterised in that described step S1 Including：

S11, to original image piecemeal and by VQ obtain recover data；

S12, the scramble of original binary watermarking and piecemeal；

S13, merging recover data and original watermark, obtain the bit plane that a size is 512 × 512, then it are put Disorderly；

S14, the front 7 MSB planes acquisition detection plane according to original image, and the result execution bit contrast with step S13 Special XOR is obtaining final watermark；

S15, complete the embedded of watermark with the LSB plane that original image is replaced in final watermark.

3. the high PSNR vulnerable watermark method based on chaology as claimed in claim 2 is it is characterised in that described step S2 Middle extraction watermark includes：

S21A, the extraction LSB plane of water mark inlaying image and detection plane；And the LSB plane of detection plane and extraction is entered Row by turn xor operation obtaining the bit plane of scramble；

S22A, the bit plane iteration block-based Cat maps T to scramble₂-K₂Secondary recovering bit plane, wherein T₂It is this step The Cat maps cycle；

Watermark bit is isolated in S23A, the bit plane obtaining from step S22A；And the scattered watermark bit obtaining is pressed Recombinated according to mode from left to right line by line, obtained the watermark of the scramble that size is 362 × 362；

S24A, to the watermark iteration of scramble the Cat maps T based on pixel₁-K₁Secondary recovering watermark, wherein T₁It is the cat of this step In the mapping cycle, thus obtaining the watermark extracting, its size is 362 × 362.

4. the high PSNR vulnerable watermark method based on chaology as claimed in claim 3 is it is characterised in that described step S2 Middle tampering location includes：

S21B, take extract watermark original watermark absolute difference, obtain and watermark size 362 × 362 identical difference plane；

S22B, in difference plane the Cat maps K based on pixel for the iteration₁The secondary difference plane obtaining scramble；

S23B, difference plane is divided into 16384 pieces, every piece is made up of 8 bits；

S24B, distort judgement：If there being any one bit to be 1 in a block, that is, extract in block and the original watermark in watermark Block inconsistent, then this block is just identified as distorting；All of piece all judges in this way, afterwards, each piece of expansion Transform into 4 × 4 size, and in block all bits value all same, whether its value be tampered determination by this block；Then, all of Extension blocks are from left to right reset line by line, form the difference plane of extension, and size is 512 × 512；

S25B, extension difference plane on iteration block-based Cat maps K₂Secondary, obtain 512 × 512 tampered region.

5. the high PSNR vulnerable watermark method based on chaology as claimed in claim 4 is it is characterised in that described step S2 Middle image recovers to include：

S21C, be all identified as distorting if there are being not detected by least 3 adjacent blocks around the block distorted, then this block also by Regard as distorting block；The tampered region being thus improved, its size is 512 × 512；

S22C, on improved tampered region iteration block-based Cat maps T₂-K₂The secondary tampered region obtaining scramble；Comparison changes The tampered region entered, can find possess following characteristics do not distort block：I.e. not in improved tampered region before Cat maps, And be moved to after Cat maps in improved tampered region；The recovery data being stored in these blocks is to be not tampered with , can be used for recovering corresponding block in improved tampered region；

S23C, by VQ recover distort block, obtain initial restoration result, size be 512 × 512；

If S24C distorts block full recovery after step S23C, just export the final image recovering, otherwise go to side Coupling is recovered.