CN100403765C - Information processing apparatus and its control method, computer program and storage medium - Google Patents

Information processing apparatus and its control method, computer program and storage medium Download PDF

Info

Publication number
CN100403765C
CN100403765C CNB2006100844829A CN200610084482A CN100403765C CN 100403765 C CN100403765 C CN 100403765C CN B2006100844829 A CNB2006100844829 A CN B2006100844829A CN 200610084482 A CN200610084482 A CN 200610084482A CN 100403765 C CN100403765 C CN 100403765C
Authority
CN
China
Prior art keywords
information
digital watermarking
data
digital
embedded
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Expired - Fee Related
Application number
CNB2006100844829A
Other languages
Chinese (zh)
Other versions
CN1893530A (en
Inventor
林淳一
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.)
Canon Inc
Original Assignee
Canon Inc
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 Canon Inc filed Critical Canon Inc
Publication of CN1893530A publication Critical patent/CN1893530A/en
Application granted granted Critical
Publication of CN100403765C publication Critical patent/CN100403765C/en
Anticipated expiration legal-status Critical
Expired - Fee Related legal-status Critical Current

Links

Images

Abstract

According to this invention, whether or not information has been tampered with can be detected while authentication information is set to be inseparable from digital information, and original digital data can be restored as long as it is free from tampering. For this purpose, when authentication information is embedded into digital information input by an image input unit (201), a Hash value calculation unit (202) generates authentication information based on that digital information, an encryption unit (203) encrypts the authentication information using an encryption key, and a digital watermarking unit (204) embeds the encrypted information in the digital information as a digital watermark.

Description

Information processor and control method thereof
The application is that application number is 02105403.7, the applying date is on February 9th, 2002, denomination of invention is divided an application for the Chinese patent application of " information processor and control method thereof, computer program and storage medium ".
Technical field
The present invention relates to a kind of information processing method and device and storage medium that is used to identify with discriminating digit information.
Background technology
In recent years, computer and network develop rapidly, multiple digital information for example text data, view data, voice data etc. is utilized.These data messages can be because of for example aging and do not damage, and can be by for good and all, preserve well.And digital information can be easy to editor and processing.This way can be more favourable for the user of editor and processing digital information.
Yet for the insurance company that uses evidence photograph in accident is handled or the building constructor of the progress in record building place, numerical data is lower than the reliability of analogue data, and numerical data is more weak as the proof of evidence.
Therefore, check that any video input device and system of distorting and/or forging numerical data carried.
For example, well-known a kind of system of digital signature that uses is used to detect any system of distorting and/or forging.Below will briefly introduce digital signature.
By using digital signature, transmitter can send simultaneously data and with the corresponding characteristic of these data, and receiver is checked the reliability of data by the verification characteristics data.As described below, utilize hash function and public code art and the reliability of the digital signature check data that produces.
Make that Ks is a key, Kp is a public keys.So, transmitter compresses plain text data M to carry out a kind of arithmetical operation by utilizing hash function, and this arithmetical operation is used to calculate the output h of given length (for example 128 bytes).Transmitter is producing digital signature data s through such arithmetical operation: with key K s conversion h, just, D (Ks, h)=s.Afterwards, transmitter sends digital signature data s and plain text data M.
On the other hand, receiver produces a kind of like this arithmetical operation: with public keys Kp the digital signature data s that receives is carried out conversion, promptly, E (Kp, s)=E (Kp, D (Ks, h "))=h ", produce a kind of like this arithmetical operation simultaneously: by the plain text data M that uses the hash function identical to compress to receive with transmitter with calculating h '.If " be complementary, receiver is determined the data M that receives ' be reliable for h ' and h.
If plain text data M has been distorted between transmitter and receiver, E (Kp, s)=E (Kp, D (Ks, h "))=hh " be not complementary by the h ' that uses the plain text data M ' that receives with the same hash function compression of transmitter to obtain, then detect data and distorted.
If to digital signature data s distort and consistent to distorting of plain text data M, can't be checked through this distorting so.Yet in this case, plain text data M must calculate from h, but can not produce such operation owing to the one-way of hash function.As mentioned above, utilized the authentication data that the digital signature of public code art and hash function can be correct by use.
To explain hash function below, hash function is used for to produce digital signature etc. at a high speed.Hash function has a kind of like this function, and it can handle the plain text data M of random length, and exports the output h of a definite length.Note output variable h be called the hashed value of plain text data M ( Perhaps informative abstract, digital finger-print).Hash function needs one-way and anti-collision.One-way means if h determines, just can not draw the text data M that satisfies h=H (M) on calculating.Anti-collision means if text data M determines, just can not draw the plain text data M ' (M ≠ M ') that satisfies H (M)=H (M ') on calculating, and also can not calculate the plain text data M and the M ' that satisfy H (M)=H (M ') and M ≠ M '.
As hash function, MD-2, MD-4, MD-5, SHA-1, RIPEMD-128, RIPEMD-160 or the like are known, and these algorithms are general.
To introduce the public code art below.In the public code art, the encryption and decryption key is different, and encryption key is known, and decruption key is maintained secrecy.The public code art has following feature,
(a), do not need so to encrypt and transmit encryption key, and be easy to carry out the key transmission since the encryption and decryption key is different and encryption key is known;
(b) since each user's encryption key is disclosed, the user only needs his or her decruption key is maintained secrecy;
(c) Discrimination Functions that can realize the receiver use comes whether the authentication information sender is illegal, and whether information is distorted;
For example, if use common encryption key Kp to plain text data M encrypt by E (Kp, M) expression, use private decruption key Ks be decrypted by D (then the algorithm of public code art satisfies following two conditions for Ks, M) expression:
(1) if Kp is known, obtain easily E (Kp, M), if Ks is known, obtain easily D (Ks, M).
(2) if Ks the unknown, even Kp, E, (Kp M) can determine C=E, can not determine M. on algorithm
If when satisfying above-mentioned condition (1) and (2), condition (3) below satisfying, then secure communication can realize.
(3) can (Kp M) be defined as and be used for all text data M and D (Ks, E (Kp, M)))=M with E.That is to say, since Kp is known to the public, anyone can both obtain E, and (Kp, M), but the people who only knows private key Ks could pass through calculating D (Ks, E (Kp, M))) and draw M.If when satisfying above-mentioned condition (1) and (2), condition (4) below satisfying then can realize differentiating communication.
(4) can (Ks M) be defined as and be used for all text data M and E (Kp, D (Ks, M))=M with D.That is to say, the people who only knows private key s can access D (Ks, M), understand the people of key K s if the third party disguises oneself as him or she, by the key K s ' that uses a mistake calculate D (Ks ', M), then because E (Kp, D (Ks ', M)) ≠ M, be illegal so receiver can be determined the information that receives.Even when D (Ks, M) when being distorted, also can (Kp, D (Ks ', M)) ≠ M makes receiver can determine that the information that receives is illegal owing to E.
As the exemplary embodiments that can realize that secure communication and discriminating are communicated by letter, rsa cryptosystem is learned, the R cryptography, and W cryptography etc. is all known.
The encryption and decryption that modern widely used rsa cryptosystem is learned provide as follows:
Encrypt: encryption key (e, n) enciphering transformation C=M e(mod n)
Deciphering: decruption key (d, n) deciphering conversion M=C d(mod n)
N=pq (p and q are two diverse prime numbers here)
As mentioned above because rsa cryptosystem learns and all will carry out the mathematical operation of sum of powers remainder in encryption and decryption, so with the public code art for example DES etc. compare, need googol to learn operand, also just be not easy to reach high speed processing.
As mentioned above, in order to find whether data are distorted and/or forged the preceding, except numerical data, also need a digital signature.General, digital signature is affixed to the stem of numerical data and sends.Yet the digital signature of interpolation is easy to be eliminated along with the format conversion of numerical data.If digital signature has been eliminated, numerical data can not be differentiated.
The disclosed method of Japanese patent application publication No. 10-164549 can address the above problem.In Japan Patent 10-164549, digital information is divided into two fields, and digital signature results from first field of cutting apart, and described digital signature is embedded into second field cut apart as digital watermarking, thereby produces the digital information of signature.On the other hand, identification device is divided into first, second field with label information, produces one first digital signature from first field, and extracts one second mark that is embedded into as digital signature from second field.If first, second digital signature is complementary, can assert that digital information is not distorted and forged.
As mentioned above, for the discriminating digit data, the important point is can not cast aside digital information and set authentication information separately.In Japan Patent 10-164549 disclosed method, owing in order to differentiate that the original image signing messages is embedded into as digital watermarking, and digital watermarking can not eliminate, so can not obtain original image.Some are used and the user may be with embed digital watermark itself as " distorting ".
Summary of the invention
The present invention proposes having considered above-mentioned problem, the purpose of this invention is to provide a kind of information processor and control method therefor, computer program and storage medium, can when being set, authentication information determine whether digital information is distorted not breaking away from digital information, and, just can recover original digital information as long as digital information is not distorted.
According to the present invention, providing a kind of is used for by adding/deduct that from the element of forming numerical data a value is embedded into additional information the information embedding device of described element, this device comprises: checkout gear, be used to detect such element: after adding computing/subtract computing, its value has surmounted the scope that described element can be assert; By in conjunction with additional information with produce the generation device of effective embedding information by the information that described checkout gear detects; And flush mounting, be used to get rid of such element: when being embedded into numerical data, add computing/subtract after the computing through what embed to handle, its value has surmounted the scope that described element can be assert, and the effective embedding information that will be produced by described generation device is embedded into the element that drops within the scope that described element can assert as digital watermarking.
To achieve these goals, for example, according to the present invention, information processor that can additional authentication information comprises array structure down.
The information processor that just embeds authentication information in the digital information processing device comprises:
On the basis of the digital information that will embed authentication information, produce the device of authentication information; With
The figure notation device that the authentication information that is produced is embedded into recoverable digital information.
In addition, information processor that can authentication information comprises following structure.
That is, can differentiate that wherein embedding the information processor that has as the digital information of the authentication information of digital watermarking comprises:
From digital information, extract the extraction element of embedded authentication information as digital watermarking as first authentication information;
The digital watermarking cancellation element is used to eliminate authentication information and the temporary transient original digital information that recovers as digital watermarking that extracts from digital information;
Thereby produce the generating means of second authentication information according to the original digital information that temporarily recovers by digital watermarking device elimination digital watermarking; With
First authentication information and the second authentication information comparing unit.
To make further feature of the present invention and a bit more obvious by the following explanation of being done with reference to the accompanying drawings, identical reference marker is represented same or analogous parts in institute's drawings attached.
Description of drawings
Fig. 1 is the computer block diagram that is applied in the embodiment of the invention;
Fig. 2 is the block diagram of the signature apparatus of the embodiment of the invention;
Fig. 3 is the block diagram of identification device in the embodiment of the invention;
Fig. 4 is the more detailed block diagram of digital watermarking device shown in Figure 2;
Fig. 5 is the principle key-drawing of the method for piecing together;
Fig. 6 represents the reliability frequency distribution state diagram of d at interval about extracting digital watermarking;
Fig. 7 is the block diagram of digital watermarking unit in the embodiment of the invention;
Fig. 8 is the block diagram that is applied to the information processor in the another embodiment of the present invention;
Fig. 9 is the block diagram that the digital watermarking in the another embodiment of the present invention forms device;
Figure 10 is the block diagram of the image recovery device in the another embodiment of the present invention;
Figure 11 shows the example of the code sequence w in another embodiment of the present invention;
Figure 12 is the block diagram that the digital watermarking in further embodiment of this invention forms device;
Figure 13 is the block diagram of the image recovery device in further embodiment of this invention;
Figure 14 shows the main points of the process of embed digital watermark;
Figure 15 shows the main points of the process of recovering initial data;
The example that overflows when Figure 16 shows embed digital watermark;
Figure 17 shows and overflows the irreclaimable example of initial data when taking place;
Figure 18 is the flow chart of the order of digital watermarking formation in another embodiment of the present invention;
Figure 19 is the flow chart of initial data recovery order in the another embodiment of the present invention;
Figure 20 shows the example of the code sequence w in further embodiment of this invention;
Figure 21 shows the example of the pixel distribution state that is used to explain error correction procedure in further embodiment of this invention;
Specific embodiment
Below in conjunction with relevant accompanying drawing most preferred embodiment of the present invention is described.
The signature apparatus and the identification device that is used to differentiate described information that below explanation are used for information that authentication information is embedded.
<signature apparatus 〉
The signature apparatus of present embodiment is described below in conjunction with Fig. 2.
With reference to Fig. 2, code name 201 representative image input units; 202 represent the hashed value computing unit; 203 represent ciphering unit; 204 represent the digital watermarking unit; 205 representative image output units.
Picture signal is input to image input unit 201 with raster scan order.The storage medium of any image reading apparatus, memory image, the communicator etc. that comes from the image of long-range video disc by means of communications reception all can be used as the input source of image input unit 201.Introduce the picture signal of the monochromatic multivalue image of expression below.If picture signal represents to contain the image of coloured image of many color components etc., in each or brightness and the chromatic component in R, G, the B color component can be used as monochromatic component and handle.
Received image signal I is imported into hashed value computing unit 202 and digital watermarking forms unit 204.
Hashed value computing unit 202 calculates and exports the hashed value H of a received image signal I.Use hash function to calculate hashed value H.The present invention is not defined as particular Hash function, and for example used MD-2, MD-4, MD-5, SHA-1, PIPEMD-128, RIPEMD-160 etc. similarly, also be the algorithms of different of knowing to the public.The hashed value H that calculates is imported into ciphering unit 203.
Ciphering unit 203 utilizes private key S to give the hashed value that hashed value H encrypts and output is encrypted of input.Do not limit the invention to any special cryptographic algorithm that is used for encrypting, can use rsa cryptosystem art etc. as the public code art.Be input to digital watermarking through the hashed value Es (H) that encrypts and form unit 204.
Digital watermarking forms unit 204 the hashed value Es (H) that encrypts is embedded into the picture signal I of input as reversible digital watermarking, and exports this picture signal (reversible digital watermarking will be introduced in the back).The picture signal wI that is embedded with digital watermarking is outputed to image output unit 205.The object of output does not have particular restriction and this picture signal can output to above storage medium, the order wire etc.
<identification device 〉
To introduce the identification device of present embodiment by Fig. 3 below.
Referring to Fig. 3, label 301 presentation video input units; 302 expression digital watermarking extraction units; 303 expression decrypting device; 304 expression hashed value arithmetic elements; 305 expression comparing units.
One image input signal wI ' is imported into image input unit 301 with the order of raster scan.This input picture wI ' is output to digital watermarking extraction unit 302 again.Notice that wI ' preferably can equate with the output wI of feature unit shown in Figure 2, but owing to this received image signal may be distorted or forge, so it is represented with wI '.
Digital watermarking extraction unit 302 extracts the digital watermarking E ' s (H) that is embedded into from the picture signal wI ' of input, and eliminates reversible digital watermarking.Export digital watermarking E ' s (H) that is extracted and the picture signal I ' that has eliminated digital watermarking then.
If wI ' equals wI, will extract Es (H) as digital watermarking (being Es (H)=E ' s (H)).In addition, in this case, reversible digital watermarking will be eliminated fully, and output image data I (being I=I ').In addition, if wI ' is different with wI, the data that then are different from Es (H) will be used as digital watermarking output (being Es (H) ≠ E ' s (H)).In addition, in this case, do not eliminate reversible digital watermarking (being I ≠ I ') fully.
And if digital watermarking is embedded in the image of input, digital watermarking extraction unit 302 also has the function that detects.Utilize this function, if wherein the view data of the digital signature that embeds as digital signal is output as wI ', then the digital watermarking extraction unit is exported " no information ", and does not carry out the discriminating processing after a while.In this case, will utilize a for example display output information " view data can not be differentiated ".
Digital watermarking E ' the s (H) that extracts is output to decrypting device 303, and the view data I ' that has eliminated digital watermarking is output to hashed value arithmetic element 304.
Data E ' the s (H) of 303 pairs of inputs of decrypting device deciphering and with its output.About deciphering, will use and the corresponding public keys p of private key s that in signature apparatus, uses.Decrypted data H is imported into comparing unit 305.
Hashed value arithmetic element 304 will be calculated the hashed value H ' of view data I ' of input and with its output.As the hash function that is used for calculating hashed value, this hash function is equal to the hash function that uses in signature apparatus.The hashed value H ' that calculates will be output to comparing unit 305.
Whether the hashed value H of 305 pairs of inputs of comparing unit and H ' compare with check wI ' and are distorted or forged.If H and H ' are equal to each other, then wI=wI ' can determine that just the data of importing are not distorted or forged.If H and H ' do not wait each other, then wI ≠ wI ' that is to say and can determine that the data of importing are distorted or forged.In this case, will on display device, show the information that shows this point.
<digital watermarking forms the unit 〉
The digital watermarking that to introduce present embodiment below forms the details of unit 204.Concise and to the point introduce a process, the digital watermarking device embeds information so that recovered view data fully before embed digital watermark.
Fig. 4 shows the internal structure that digital watermarking forms unit 204.Introduce handling process with reference to Fig. 4 below.
View data I is imported into digital watermarking and forms unit 204, and an embedded location determining unit 401 definite embedded locations that will be embedded in the digital watermarking of view data I.This view data I is imported into additional information and embeds unit 402, and according to additional information Es (H) at the position embed digital watermark of determining by embedded location determining unit 401.
To achieve these goals, embedded location determining unit 401 outputs to additional information embedding unit 402 with the view data I of input and the data (having specified each regional coordinate and size) of indicating image position, and wherein said picture position is a position that will embed additional information Es (H).
Except image information I, additional information embeds unit 402 and also receives additional information Es (H) (some information).By using digital watermark technology, additional information Es (H) will be embedded at the embedded location of determined view data I.To introduce subsequently and utilize digital watermark technology to embed additional information Es (H).Additional information embeds that unit 402 outputs have embedded additional information Es (H) and through the data wI of digital watermark processing.
In the present embodiment, for the purpose of simplifying, the data that are input to the digital watermarking device are the gray level image data that every pixel (256 gray scales) has 8 bit gradation precision.Yet input image data may be color image data.When input color image, utilize the pixel value of a channel of coloured image or the brightness value of coloured image, then can carry out a similar embedding and handle.
When input audio signal, 2 dimension positional informations of image will be used as the 1 dimension information of time, can use identical carrying into execution a plan.When the input moving image data,,, used the present invention substantially so can handle each two dimensional image with identical scheme because it can be counted as a plurality of two dimensional images along time-axis direction.Therefore, scope of the present invention comprises that also digital watermarking is embedded into the situation of coloured image, audio frequency and live image.
To introduce the embedding and the basic principle of separating (extraction) of digital watermarking in the present embodiment below.
<piece together method 〉
Present embodiment uses the principle that is called the method for piecing together to embed additional information Inf (consistent with the Es (H) in the foregoing description).Piece together the NikkeiElctronics that method was published on February 24th, 1997, have disclosed by " the Data Hiding Technique that SupportsDigital Watermarking (Vol.1) " of propositions such as Walter Bender, Daniel Gruhl, Norishige Morimoto and Anthony Lu.At first introduction is pieced together the principle of method.
In the method for piecing together,, can embed additional information Es (H) by slip chart picture on statistics.
To introduce by Fig. 5 below and piece together principle.Among Fig. 5, A and B are two subclass selecting from an original image.Suppose that subclass A is by many subclass a iThe subclass of 501 expressions is formed, and the subclass that subclass B is represented by many subclass bi502 is formed.
If these two subclass do not overlap each other, the method for piecing together by present embodiment can embed additional information Inf so.
Suppose that among subclass A and the B each all comprises N element (A={a 1, a 2..., a N, B={b 1, b 2..., b N).Element a among subclass A and the B iAnd b iExpression has pixel or one group of such pixel of pixel value.
Next index d is defined as d=1/N ∑ (a i-b i), herein ∑ be i=1 to N and.
This has just expressed different desired value between two groups of pixel values.
When suitably select subclass A and B and definition index d from the image of general nature, if the value of N is enough big, we will obtain:
d≈0
D will be used as reliability at interval hereinafter.
On the other hand, as each embedding operation of forming additional information Es (H), for example, when embedding position information " 1 ", will produce such operation:
a′ i=a i+c
b′ i=b i-c
Such operation will add " c " at the pixel value of all elements of subclass A, and the pixel value of all elements of subclass B is all deducted " c ".Hereinafter " c " at present embodiment will be used as " insert depth ".
Resemble top situation, as the subclass A and the B that have selected embedding additional information Es (H) from an image, index d is:
d=1/N·∑(a i-b i)
=1/N·∑{(a i+c)-(b i-c)}
=1/N·∑{(a i-b i)+2c)}
≈2c
(∑ be i=1 to N and)
That is to say that d is rendered as such value, it and 0 differ a given interval.
For the position information (position information " 0 ") that embeds other, will carry out following operation:
a′ i=a i-c
b′ i=b i+c
Reliability interval d is so:
d=1/N·∑(a i-b i)
=1/N·∑{(a i-c)-(b i+c)}
=1/N·∑{(a i-b i)-2c)}
≈-2c
Just d is assumed to such value at interval, and it differs a given distance with 0 on negative direction.
When given a certain image,, just can determine whether embedded additional information in the image by the reliability distance d of computed image.
If that is: reliability distance d ≈ 0 does not then embed additional information Es (H); If reliability distance d be one predetermined on the occasion of or for than 0 much bigger value, then embed position information " 1 "; If reliability distance d is a predetermined negative value or is than 0 much smaller value, then embeds position information " 0 ".
In the present embodiment,, can in single image, embed a large amount of positions information segment by utilizing the previously mentioned method principle of piecing together.
In the present embodiment,, not only can adopt the combination of subclass A and B in the zones of different of single image, and can adopt many subclass A ' and B ', A " and B " ... combination embed the additional information Es (H) that includes many positions.Note subclass A and B, A ' and B ', A " and B " ... distribution can not be overlapping.
To check below from the method for the extracting data position information that has wherein embedded a large amount of information.
Label 601 is represented the distribution of the reliability distance d that calculates in the data of embed digital watermark never among Fig. 6.Distribution curve 601 shows the value of the reliability distance d of most possible appearance, because bigger to the distribution of the frequency of occurrences that should reliability distance d.
Distribution curve 602 and 603 is respectively the reliability distribution distance d that is obtained by the data computation that has embedded position information " 1 " and position information " 0 ".
Similarly, each distribution curve 602 and 603 shows the value of the reliability distance d of most possible appearance respectively, because bigger to the distribution of the frequency of occurrences that should reliability distance d.Note the corresponding reliability distance d of position information.
Distribution curve 601,602,603 among Fig. 6 all is the normal distribution state.To utilize central-limit theorem to introduce the compositing cause of this normal distribution curve below.
<central-limit theorem 〉
This theorem has disclosed: when from having average m cWith standard error δ cAll territories (not needing to be normal distribution) in randomly draw size for n cSample the time, have sample average s cDistribution be similar to n with increase cNormal distribution N (m c, s c/ √ n c) 2).
Usually, the standard error δ in all territories cBe unknown, yet, if sample number n cEnough big, and the several N in all territories cBe enough to than sample number n cBigger, in the way of routine, can use the standard error s of sample so cReplace δ c
In the present embodiment, each subclass A and B comprise N element (A={a 1, a 2..., a N, B={b 1, b 2..., b N), and these set are for having the pixel value of subclass A and B element, as shown in Figure 5.If set the enough big and pixel value a of N iAnd b iThere is not association, then reliability interval d (∑ (a i-b IThe desired value of)/N) becomes 0.As learning from central-limit theorem: this reliability spacing d has constituted the distribution of a normal state.
Therefore when determining the position information of embedding from reliability spacing d, to between 0 to reliability spacing 2c, introduce a suitable threshold value, and when the absolute value of reliability spacing is bigger than threshold value, determine that then information is embedded into, then on statistics, finished the reliable extraction of information.
For example, make that δ is the standard error of normal distribution 601.So, if do not embed additional information, then in shown-1.96 δ of Fig. 6 dash area arrived the scope of+1.96 δ (95% reliability scope), the reliability spacing had 95% possibility that occurs.
Therefore, the probability that occurs reliability spacing d beyond the threshold value scope reduces along with the increase of threshold value, then can extract high reliability information.
In addition, if set a big insert depth " c ", then normal distribution 602 and 603 will be separated from distribution 601, thereby a big critical value can be set.
If the number of elements N among subclass A and the B is big, then normal distribution 601,602 and 603 standard error δ diminish, and if insert depth c remain unchanged, then can determine higher reliability.
By the agency of piece together the basic thought of method.
In the present embodiment, digital watermarking forms unit 204 and digital watermarking extraction unit 30 just is to use the aforesaid method of piecing together.
To introduce conventional word watermark embedding, extraction and the removing method stated below.
<embedded location determining unit 〉
Because the method for piecing together embeds the additional information that comprises many positions, each information all needs subclass A and B.Therefore, in order to embed many position information, must determine A and B, A ' and B ', A " and B " ... the position.
The embedded location determining unit 401 of Fig. 4 determines to need to embed the embedded location of some position information.Determine method as a simple embedded location, may use the method for utilizing random number to determine the position.Many bits are with good stable embedding entire image, so that corresponding subset elements almost is equally distributed, and do not overlap each other between the subclass.
For example, briefly introduce use and have the method for onesize white noise mask with image.
The white noise mask is made up of the mask pixel of Two dimensional Distribution, and each mask all has a coefficient in 0 to 255 scope.Each coefficient of white noise mask between 0 to 255 all has been endowed the mask pixel of similar similar number.
For example, when original image has 2000 * 2000 pixels (=4000000 pixels),, then there is the individual mask pixel in 15625 (=4000000/256) to have " 0 " value because a white noise mask has disposed the mask pixel count identical with the original image pixels number.Such mask pixel is a random distribution, to form the white noise mask.
Therefore, in case embed 1 bit additional information, then work as the mask pixel that will have the odd number gray scale and be assigned to subclass A, and the mask pixel that will have the even number gray scale is when being assigned to subclass B, then subclass A has identical pixel count and not overlapping mutually with B, and additional information can be with good stable embedding entire image.
When embedding many (M) bar position information, pixel count is distributed to each information (1 to M) (for example, divided by 2M, remainder is distributed to subclass A or B uniformly with the scope of the white noise mask set) uniformly, then embeds many position information.
If each pixel all uses 8 bits to represent, then because number of greyscale levels is 256, thereby the bit number M of the maximum that can embed is 128.This maximum only satisfies a hashed value (64 or 128 bit), if but add other information (as copyright protection information), this value is with insufficient.Yet, when piece image is split into four parts, and, can embed 128 * 4=512 bit for relevant block configuration white noise mask.If the block number is greater than 4, the bit number of embedding also can increase.Yet, if the block number increases, because normal distribution is difficult to formation, so the extraction of the information of embedding may often be failed.That therefore, determines that the block number should be with original image is big or small corresponding.
Therefore embedded location determines that device only need produce cut-and-dried mask pattern.That attention is used in embedding, identical mask is to extract with the side preparation that separates digital watermarking and using.
<additional information embeds the unit 〉
Additional information embeds unit 402 and receives view data I, additional information E SAnd aforesaidly determine and the embedded locations consistent (H), with corresponding bit by embedded location determining unit 401.
According to the additional information E that forms input S(H) position information, the pixel value of the pixel of calculating and the corresponding subclass A of each binary number and B.
As in piecing together the method part, introducing, if position information is " 1 ".Then the pixel value from the pixel of subclass A adds " C ", and deducts " C " from the pixel value of the pixel of subclass B.If position information is 0, then the pixel value from the pixel of subclass A deducts " C ", and the pixel value of the pixel of subclass B is added " C ".By aforesaid processing, additional information embeds unit 402 and embeds additional information E S(H).
Can embed additional information by preceding method.Yet, if the method above carrying out satisfies c>a simultaneously iOr a i>255-c and c>b iOr b iThe pixel value a of the pixel of the embedding of>255-c i' and b i' become a i' (b i')<0 or a i' (b i')>255.Therefore digital watermarking elimination unit (back will be introduced) (is a in the pixel that these location of pixels can not recover original image iAnd b i).
Present embodiment is used to realize that the embedding of satisfying the pixel of such condition handles: c≤a i≤ 255-c and c≤b i≤ 255-c, and the pixel value that drops on outside this scope is not carried out any embedding.So, in the superincumbent narration, when embedding the M bit, " scope that the white noise mask presents is divided into 2M part ", but must get rid of those scopes in advance.Attention does not embed the locations of pixels information of processing and must export as spillover position information, and is input to digital watermarking extraction unit and digital watermarking elimination unit (back will be introduced).
<digital watermarking extraction element 〉
To introduce the digital watermarking extraction element 302 in the present embodiment below.Digital watermarking extraction element 302 has structure as shown in Figure 7.
As shown in Figure 7, digital watermarking extraction element 302 comprises embedded location determining unit 701, additional information extraction unit 702, statistical test unit 703, comparing unit 704 and additional information elimination unit 705.
The digital watermarking extraction element receives digital watermarking and embeds data wI '.Embedded location determining unit 701 is created in the positional information of this position embed digital watermark (with forming the identical pattern of using in the device 204 of white noise mask in digital watermarking).Input will from the basis of the positional information of its embed digital watermark, embed data for digital watermarking, additional information extraction unit 702 is carried out a predefined procedure, then calculates and is embedded into the interior additional information E ' of view data wI ' S(H) corresponding reliability spacing d.Statistical test unit 703 statistical tests and the additional information E ' that calculates by additional information extraction unit 702 S(H) precision of corresponding data.If determine additional information E ' S(H) enough accurate, then comparing unit 704 extracts additional information E S(H).If additional information is inaccurate, then output " no information ".If embedded information, additional information is eliminated unit 705 and is utilized view data wI ', the embedded position information that comes from embedded location determining unit 701 and the spillover position information of input to eliminate digital watermarking.
To introduce in detail below and be used for having the view data wI ' of digital watermarking to extract additional information E ' from embedding SThe operation of digital watermarking extraction unit (H).
<embedded location determining unit 〉
Embedded location determining unit 701 is determined to extract additional information E ' from it S(H) zone of view data wI '.Embedded location determining unit 701 is carried out and embedded location determining unit 401 identical operations.Identical embedded location is determined with 801 in unit 401 thus.The embedded position information of determining is output to additional information extraction unit 702 and additional information is eliminated unit 705.
<additional information extraction unit 〉
Additional information extraction unit 702 calculates and the corresponding reliability spacing of each bit d, and this bit comes from the embedded location that embedded location determining unit 701 is determined.In this case owing to do not embed additional information in the pixel, unit 702 do not use computed reliability apart from d in by the pixel of spillover position information representation of input.
<statistical test unit 〉
The analysis of statistical test unit 703 statistics and precision from the corresponding reliability distance d of each bit of additional information extraction unit 702 outputs.If embedded many position information, will obtain a plurality of reliability spacing d.If embed additional information E ' S(H), then reliability spacing d appear at Fig. 6 in center O on the position of 2c.
At this moment, along with the insert depth c that increases, reliability distance d appears on the farther position of Fig. 6 center O.Therefore,, and obtain the reliability spacing d bigger, determine that then the position information that embeds is " 1 " than c if threshold value is limited to position c; If it is little to obtain reliability distance d ratio-c, determine that then the position information that embeds is " 0 ".
Therefore,, along with the increase of insert depth " c ", in case embed additional information, the interval between the normal distribution 601,602 and 603 also increases thereupon, and guarantees the more high reliability of information extraction.Simultaneously, along with the increase of the number of elements N of subclass A and B, normal distribution 601,602 and 603 standard error descend thereupon.Therefore, even the number N of the element by increasing insert depth " c " and subclass A and B when threshold value is " c ", also can obtain the information extraction of high reliability.
Attention is not when embedding information, and reliability spacing d (more possible) appears at from-c in the close limit of c and by utilizing this factor can determine such situation.
Especially, when and the corresponding given number of some bits or more the reliability distance d of more number appear at scope-c between c the time, the information that do not embed are determined in the statistical test unit 703 of present embodiment, and show that a piece of news represents such result.
<comparing unit 〉
Comparing unit 704 among Fig. 7 is provided with the value of the corresponding reliability spacing of each the bar position information d that exports with process additional information extraction unit 702 and statistical test unit 703.
Since with some the corresponding reliability spacing of position information d that are input to comparing unit 704 be highly reliable, so only according to determining it is " 1 " or " 0 " with the sign of the corresponding reliability spacing of each information d.
Especially, if additional information E ' is formed in given being used to SThe reliability spacing d of position information (H) is bigger than " c ", then determines this information and is " 1 ", if reliability spacing d is littler than " c ", determines that then this information is " 0 ".
<additional information is eliminated the unit 〉
Introduce additional information below and eliminate the operation that unit 705 is carried out.Additional information is eliminated unit 705 and is received additional information E ' S(H) embedded location, view data wI ' and spillover position information, and the view data I ' of additional information has wherein been eliminated in output.
On the position identical with the position of additional information embedding unit 402 embeddings in the digital watermarking formation unit, when embedding, insert depth " c " is added to the subclass that is complementary with corresponding bit, eliminates additional information and recover original image with this by the symbol that changes insert depth " c ".
Especially forming additional information E ' SThe embedded location of predetermined position information (H), if position information is " 1 ", then executable operations:
a?′ i=a i-c
b′ i=b i+c
If position information is " 0 ", then executable operations:
a′ i=a i+c
b′ i=b i-c
Then before embedding, recover pixel value.In this case, utilized the spillover position information of input.Therefore, owing to do not embed additional information in the pixel on spillover position, this pixel need not be eliminated.
By above-mentioned computing, additional information is eliminated the unit 705 view data I ' of digital watermarking that eliminated digital watermarking and output elimination wherein from the data wI ' that has embedded digital watermarking.
Introduced the method for digital signature of using in the foregoing embodiments as authentication information.Yet the present invention is not limited to this special method, for example comprises, uses the method for MAC (Message Authentication Code) as authentication information.In addition, except digital signature and MAC, also may comprise at least one or more date and time information, positional information, temporal information, the unique information of device and signer's Proprietary Information.
In addition, in the above embodiments, the additional information of use can be converted to error correcting code.Can further improve the reliability of the additional information Inf of extraction in this way.
Attention can come most of composition element of implement device by software at an end that embeds information and an end of differentiating the information that embeds, and promptly the associated processing device shown in Fig. 2 and Fig. 3 can be realized by software.
In this case, the structure of device can be a kind of so general device: ordinary individual's computer or similar device, for example, can use structure shown in Figure 1.
Fig. 1 shows the general structure of the image processing apparatus that is applied to present embodiment.With reference to Fig. 1, main frame 101 is for example widely used PC.
In main frame 101, relevant portion described below is coupled together to exchange different data by bus 107.
With reference to Fig. 1, label 103 expression CPU, it can be controlled the operation of associated inner unit or carry out internally stored program.Label 104 expressions one can be stored the ROM of boot and BIOS.The RAM of the view data that the program when label 105 expressions one temporary transient storage is used for the CPU executable operations maybe will be handled.Label 106 expression one hard disks (HD), it can store the program and the view data that will be sent to RAM or similar units in advance, and can preserve the processed images data.When main frame uses as signature apparatus, hard disk 106 storage is corresponding with Fig. 2, called in RAM 105 when operation program.When main frame uses as identification device, hard disk 106 storage is corresponding with Fig. 3, be loaded into RAM 105 when operation program.
Label 108 expression one CD drive unit: it can be packed into and be stored in data as on a kind of external storage medium CD (CD-R), perhaps data is write on the CD-R.Label 109 expression one and CD drive unit 108 are similarly and can be from pack into data or data are write the FD drive unit of FD of FD.Label 110 expression one and CD drive unit 108 similar and can be from pack into data or data are write the DVD drive unit of DVD of DVD.Notice that such program is loaded into HD 106 when needed and is sent to RAM 105 when image editor or printed driver are stored on CD, FD, DVD or the analog.These storage mediums are used for storing original image, and embed characteristic information in the image of storage, perhaps differentiate the identification image that is stored on the existing storage medium.
Label 113 expression one is connected to keyboard 111 and mouse 112 to receive from the interface (I/F) of the instruction of its input.Label 114 expression one includes the display controller of image controller and video memory (not shown), in order to producing and to show relevant control, and is mapped to image on the video memory by handle and outputs to display device 115 and come display image.Label 115 expressions are used to be connected to the communication interface (for example, modulator-demodulator, Ethernet transducer or similar device) of the Internet.Can send or receive the additional information that characteristic information is arranged by communication interface 115.
Notice that the device that is used for input image data is not limited to above-mentioned device, also can replace scanner or analog.
Aforesaid, according to present embodiment, digital information and signing messages are with indivisible state setting, and signing messages is sent to identification device through similar format conversion.
[ Alternative types]
In the above-described embodiments, used the method for piecing together.To introduce other example of the application in the following narration.
As mentioned above, " digital watermarking " is known a kind of copyright protection technology.Digital watermarking is to be used for embedding copyright owner's name or buyer's ID with a kind of invisible form in DID, voice data, text data or similar data, and follows the tracks of the without previous notice that utilizes illegal copies to carry out and use.Because numeral may suffer different destruction, then it must have the intensity that opposing destroys.
Proposed in addition from figure to extract embed as the information of digital watermarking and recover the technology of original image fully from this image.
A kind of embedding grammar simply is expressed as:
I′ j,i=I j,i+c j×a i×x i (1)
Wherein j is the zone of expression additional information Inf and the positive number of bit position, and i is the positive number of remarked pixel position, I ' J, IFor having embedded the image of digital watermarking, if Inf j=" 1 ", then c jBe constant+1, if Inf j=" 0 ", then c jBe constant-1, a iBe weight coefficient, and x iFor dropping on the pseudo-random number sequence between-1 to+1.x iBe called and be used for the carrier signal of embed digital watermark.
When recovery operation, at first extract the digital watermarking that embeds.Utilize the digital watermarking of extracting, recover original image by following equation:
I″ j,i=I′ j,i-c j×a i×x i (2)
I wherein " J, iBe the view data that is resumed, I ' J, iFor being input to the view data of image recovery device, if the bit among the additional information Inf that is extracted by the digital watermarking extraction unit is " 1 ", then c jBe constant+1, if bit is " 0 ", then c jBe constant-1, and a iAnd x iWith middle represent identical of equation (1).
To introduce the formation of the digital watermarking that equation (1) provides and original image that equation (2) provides below in detail recovers.
Figure 14 shows the example that forms of the digital watermarking that is provided by equation (1).The various piece of each matrix notation image.It is 1 to be c that Figure 14 shows an information j=+1 situation.
At reset terminal, at first extract digital watermarking.Briefly introduce this extraction algorithm below.Detection pixel value and pseudo-random number sequence x in one 4 * 4 the pixel region of input picture I ' iCorrelation between (aforesaid) with identical in the random number sequence that embeds the side generation.If the image I of input ' and pseudo-random number sequence x iBetween the correlation height, the bit that then determine to embed is " 1 ".In addition, if the input image I ' and-x jCorrelation height between (result that the symbol of the element by changing corresponding pseudo-random number sequence obtains) determines that then the bit that embeds is " 0 ".If two correlations are all low, then determine not have embed digital watermark information.4 * 4 pixel regions in Figure 14, have been explained.Yet, if a bit only is embedded into a zone, can not expect to obtain high accuracy, as shown in figure 14, take such method to improve whole precision to many zones (for example n pixel region).If embedded m bit, then Repeated m embeds the process that bit is such time in the n pixel region.
After extracting digital watermarking by this way, carry out the process of eliminating this digital watermark information.
Figure 15 shows the process of eliminating the digital watermarking that embeds in the example shown in Figure 14.In this case, the position information of embedding is " 1 ".Shown in Figure 14 and 15, can eliminate the digital watermarking of embedding generally speaking fully and recover original image.
In addition, Figure 16 also shows the example that the digital watermarking that provided by equation (1) forms.In example shown in Figure 16, embedding has the view data of digital watermarking to comprise the pixel (those have the pixel of pixel value " 262 " and " 261 " after the formation of digital watermarking) of overflowing.In this case, taked the processing of rounding off shown in Figure 16.Digital watermarking may never be carried out in any image of handling that rounds off by correct extraction, but original image can not perfectly recover from the part of handling of having carried out rounding off.
In essence, by above-mentioned method, the numerical data of recovering digital data before digital watermarking forms that digital watermarking can be arranged from embedding, if but owing to the formation of digital watermarking has taken place to overflow, then can not recover the numerical data before digital watermarking forms completely, as mentioned above.Noting overflowing is a kind of like this phenomenon: if I I, jRepresent by the integer (8) 0 to 255, after processing through equation (1), I ' I, jHas the value that drops on outside this scope.Value less than 0 is rounded to 0 usually, and the value greater than 255 is rounded to 255 usually.If carried out such processing of rounding off, then can not eliminate digital watermarking completely by the digital watermarking cancellation element.
For this reason, then be difficult in the place of overflowing and recover the original image part fully.
Therefore present embodiment can be realized the perfect reproduction of initial numberical data.
[formation of digital watermarking]
The digital watermarking that to introduce below in the present embodiment forms device.Digital watermarking in the present embodiment forms device and embeds additional information so that can perfectly be reproduced in digital watermarking formation view data before.
Fig. 9 shows the internal structure that digital watermarking forms device.To utilize Fig. 9 to explain that digital watermarking forms the process chart of device below.
Image I is imported into digital watermarking and forms device.In order to simplify, image I is the gray scale multivalue image data with every pixel 8 bit accuracy.Yet the present invention is not limited to such particular data, and may import the gray scale multivalue image data with predetermined number of bits.In addition, when input one when including the coloured image of many elements, can select one perhaps multielement as the image or the image sets of input.The image I of this input is imported into overflows regional detecting unit 2201 and digital watermarking formation unit 2203.
At first introduction is overflowed regional detecting unit 2201.This overflows regional detecting unit 2201 and receives image I, if and form the digital watermarking of carrying out unit 2203 by next digital watermarking and form and taken place to overflow after handling, then detect all pixels of the image I be used to form input, extract all locations of pixels of overflowing.Export the coordinate information of these pixels simultaneously.
In the narration below, the positional information of the pixel of overflowing (0,1, or more in some cases) will refer to flooding information R.Flooding information R will be output to coding unit 2202 and digital watermarking forms unit 2203.The back will be introduced the operation of overflowing regional detecting unit in more detail.
To introduce coding unit 2202 below.Coding unit 2202 receives flooding information R and additional information Inf, and this flooding information R and additional information Inf are combined as a simple code sequence, and exports the code sequence w of this combination.For example, w has form shown in Figure 11.Notice that w is used for the code sequence as digital watermarking is embedded into image I by digital watermarking extraction unit 2203.For this reason, handle, can carry out compressed encoding w in order to realize that effective digital watermarking forms.In addition, if additional information Inf is a personal information, can encrypt w.Even for embedded as the additional information Inf of digital watermarking with will can be converted to error correcting code to w this additional information as the result of different destructions and after changing to another kind of information, also can correctly extract additional information Inf.Under any circumstance, handle, all include indicated location of pixels, and utilized pixel value at the original image of those positions by flooding information R from embedding.
To introduce digital watermarking below and form unit 2203.Digital watermarking forms unit 2203 and receives view data I, flooding information R and code sequence w, and w is embedded into image I as digital watermarking code sequence, and output embed image I that digital watermarking is arranged '.The processing that forms of digital watermarking is not carried out in attention by the indicated pixel of flooding information R.Therefore digital watermarking forms unit 2203 and can not produce overflowing that any formation owing to digital watermarking causes.The back will be introduced the operation that digital watermarking forms unit 2203 in more detail.
Aforesaid, will be embedded into image I as the additional information of digital watermarking, then produce to embed image I that digital watermarking is arranged '.
[image recovery device]
To introduce the image recovery device of present embodiment with Figure 10 below.The image recovery device is carried out such process, reverts to view data before digital watermarking forms in order to the image that embedding is had digital watermarking.
The image recovery device by above-mentioned method receive embed image I that digital watermarking is arranged '.Unit 2303 is eliminated in the image I of this input ' be imported into digital watermarking extraction unit 2301 and digital watermarking.
At first explain digital watermarking extraction unit 2301.Digital watermarking extraction unit 2301 receive embed image I that digital watermarking is arranged ', extract the code sequence w that embeds, and export the code sequence w of this extraction.Digital watermarking extraction unit 2301 use the image I of input ' all real pixel extract digital watermarking.That is to say, all pixels are carried out the digital watermarking extraction together with those by pixel flooding information R (not having embed digital watermark by the described position of flooding information R) expression and that get rid of outside digital watermarking embedding target.The back is with the more detailed operation of introducing digital watermarking extraction unit 2301.The code sequence w that extracts outputs to decoding unit 2302.
To introduce decoding unit 2302 below.Decoding unit 2302 receives the code sequence w that is extracted by unit 2301, it is divided into flooding information R and the additional information Inf that forms code sequence w, and exports this flooding information R of cutting apart and additional information Inf.2202 couples of w carry out compressed encoding when coding unit, and decoding unit 2302 is carried out the processing of decompress(ion) decoding.Simultaneously, after 2202 couples of w of coding unit encrypted, decoding unit 2302 was carried out the processing of a decompress(ion) decoding so that enciphered data is deciphered.In addition, when coding unit 2202 was converted to error correcting code to w, decoding unit 2302 was carried out error-correcting decodings and is handled so that correct a mistake.Flooding information R and additional information Inf are output to digital watermarking and eliminate unit 2303.
To introduce digital watermarking below and eliminate unit 2303.Digital watermarking eliminate unit 2303 receive embed image I that digital watermarking, flooding information R and additional information Inf are arranged ', utilize additional information Inf before the combine digital watermark from the image I that is embedded with digital watermarking ' recover original image, and export this image restored I.In this recovers to handle, got rid of pixel by flooding information R indication.This is because do not having embed digital watermark by aforesaid digital watermarking formation unit 2203 by the pixel that flooding information R indicates.The back will be introduced the operation that unit 2303 is eliminated in digital watermarking in more detail.
Aforesaid, from embedding the image I of digital watermarking ' the recover image I before forming digital watermarking is arranged.
[details of digital watermark processing]
To illustrate the details that forms digital watermarking below.
To analyze the situation of n bit information Inf embedded images I (promptly embedding the information w of the foregoing description) below.In this case, image I is split into n nonoverlapping area I j(j=1,2 ..., n).The forming process of digital watermarking is given as so:
I′ j,i=I j,i+c j×a i×x i (4)
Wherein j is the zone of expression additional information Inf and the positive number of bit position, and I is the positive number of remarked pixel position, I ' J, iFor embedding has the image of digital watermarking, if Inf j=" 1 ", then c jBe constant+1, if Inf j=" 0 " is c then jBe constant-1, a iBe weight coefficient, and x iFor dropping on the pseudo-random number sequence of scope-1 between+1.x iBe called the carrier signal that is used to realize embed digital watermark.
Make a MaxThe maximum that presents for weight coefficient (on the occasion of), and x MaxBe x iThe maximum that can reach (on the occasion of).So only ought satisfy condition:
a max×x max≤I≤255-a max×x max (4)
The given arithmetical operation of Shi Zhihang equation (3).
In this case, because a IAnd x iThe respective element value be to occur synchronously, and at the device that is used for embed digital watermark with known when extracting the device of digital watermarking, equation (4) can extend further to:
a i×x i≤I≤255-a i×x i (4′)
If do not satisfy the condition of equation (4) (or equation (4 ')), then location of pixels be recorded as flooding information R.In common natural image, the pixel count (just get rid of embed pixel count) that does not satisfy the given condition of equation (4) is than little many of the sum of all pixels of entire image.This shows that flooding information R has relatively little amount of information.Therefore, when by coding unit 2202 flooding information R being made up then when forming unit 2203 code sequence w is embedded as digital watermarking by digital watermarking as code sequence w, this information size is also ignored and is not remembered.As mentioned above, owing to satisfying under the given condition of equation (4) (or equation (4 ')), then can not overflow ground combine digital watermark forming process by equation (1) embed digital watermark.
[image recovers the details of processing]
To illustrate the details that image recovers processing below.
Code sequence w be embedded into as digital watermarking the image I of input '.In order to use equation (3) to extract digital watermarking, from the carrier signal x that is used for embed digital watermark be input to the image I of digital watermarking extraction unit ' calculate PFA, and no matter whether embedded digital watermarking, also no matter bit is " 0 " still is " 1 ", determines whether to have embedded digital watermarking on the basis of result of calculation.
When noticing that PFA represents in fact not have embed digital watermark, the probability of error checking digital watermarking.In order to calculate this probability, used the method that is called statistical test.Statistical test is known technology to those skilled in the art.Use the more detailed description of the example of statistical test during about the extraction digital watermarking, with reference to the I.C.I.I.P proceedings of publishing in September, 1996, " the A method for signature casting on digitalimages " of 215 to 218 pages of I.Pitas works.Utilize statistical test, for each bit calculates with test value q jThis test value q jFollow the standardized normal distribution of zero-mean and error=1, but when embedding in the numerical data had digital watermarking, it followed the normal distribution of Non-zero Mean and error=1.According to the assay value q that calculates jAnd the distance between 0, determine whether embedded digital watermarking.
In addition, the test value q that uses this to calculate jThen can calculate the information that is embedded into as digital watermarking.When with equation (3) embed digital watermark, if q jFor just, determine that then bit is " 1 ", if perhaps q jFor negative, then bit is " 0 ".
As the principle that is used to extract digital watermarking, with reference to the content of prior art.
Aforesaid digital watermarking is extracted and is handled, and is used to realize all pixels, also comprises those because it does not satisfy the given condition of equation (4) (or equation (4 ')) and forms the pixel that unit 2203 does not have embed digital watermark by digital watermarking.Because it is littler than the sum of all pixels of entire image not satisfy the pixel count of the given condition of equation (4) (or equation (4 ')), when extracting digital watermarking, even comprise those above-mentioned pixels, extraction digital watermarking that also can be correct.
The digital watermarking of Ti Quing is the code sequence w that is made up of flooding information R and additional information Inf by this way, and code sequence w is split into flooding information R and additional information Inf.Afterwards, the unit execution is eliminated in digital watermarking:
If?icR?then?I″=I′
else?I″=I′ j,i-c j×a i×x i (5)
I wherein " I, jBe image restored data, I ' I, jFor being input to the view data of image recovery device, if the bit among the additional information Inf that is extracted by the digital watermarking extraction unit is 1, C then jBe constant+1, if perhaps this bit is " 0 ", then C jBe constant-1, and a iAnd x iWith the same meaning in the equation (3).Just, if i is not included among the flooding information R, then carry out and the opposite processing of equation (3); If i is included among the flooding information R, then do not carry out any processing.
When correctly extracting code sequence w by aforesaid processing, " I is identical with original image for this image restored I.
[explanation of actual device]
To those skilled in the art, by carrying out the program of aforementioned processing, realize the device that is used for the device of embed digital watermark and is used to extract the digital watermarking of this embedding and recovers original image easily.
Fig. 8 shows the utility structure of this device, and the process (program) that is associated with the embedding of digital watermarking, and after introducing Fig. 8, introduces and recover to handle (program).
With reference to Fig. 8, main frame 2101 for example is a widely used PC.
Described below, the corresponding units in main frame 2101 inside couples together to exchange different data through bus 2107.
With reference to Fig. 8, label 2102 is represented a monitor (display device), and 2103 represent a CPU, and it can be controlled the operation of corresponding internal element or carry out internally stored program.The ROM of label 2104 expressions, one storage boot and BIOS.Label 2105 is represented when carrying out a processing by CPU, the RAM of the view data that temporary transient stored program maybe will be handled.Label 2106 is represented a hard disk, and it can store the program (OS and image processing program) and the view data that will be delivered to RAM or similar units in advance, and can preserve the processed images data.
Label 2108 expression one CD drive unit: it can be packed into and be stored in data as on a kind of external storage medium CD (CD-R), perhaps data is write on the CD-R.Label 2109 expression one and CD drive unit 2108 are similarly and can be from pack into data or data are write the FD drive unit of FD of FD.Label 2110 expression one and CD drive unit 2108 similar and can be from pack into data or data are write the DVD drive unit of DVD of DVD.Notice that such program is loaded into HD 106 when needed and is sent to RAM 105 when CD, FD, DVD or analog memory image edit routine or printed driver.It is one that link to each other with mouse 2112 with keyboard 2111 and can receive interface (I/F) from the instruction of its input that label 2113 is represented.Label 2114 representatives are used to be connected to for example communication interface of the Internet of network.
It may be the storage medium of similar HD, FD that embedding has the outlet terminal of the image of digital watermarking, or the file server on the network.Be applied to receive the medium that embeds the information that digital watermarking is arranged too.
When aforesaid device is used as the digital watermarking flush mounting, then according to the flow performing of Figure 18.Notice that the program with this flow process association described below is installed on the HD 2106.Simultaneously, the information that will embed is from for example keyboard input.Perhaps also can use the information that is stored in advance on HD2106, ROM 21047, RAM 2105 or the analog.
In step S101, the pictorial data that output will embed digital watermark.There is no particular limitation for input source, and pictorial data can be from network download, or use the device input of image scanner and so on.In step S102, all that detect the pictorial data come from input are overflowed locations of pixels, and are kept among the RAM 2105 as flooding information R.Shu Ru additional information Inf and the flooding information R coding (by reversible encoding) that is combined produces the information w that will embed when using in advance, and this information temporarily is kept among the RAM 2105.
Each pixel in the pictorial data of input is selected does the pixel that will study, and whether the pixel that verification will be studied is included among the flooding information R.If this locations of pixels is not included among the flooding information R, then from information w, extracts 1 bit, and carry out and the corresponding to digital watermarking forming process of this bit status.
In addition, if the location of pixels of being studied is included among the flooding information R, then do not carry out any digital watermarking forming process.
This process is repeated on all position information and pixel (step S106 and S107).In case finish the processing that is used for all position information and all pixel, then embedding has the pictorial data of digital watermarking to be output to RAM 2105 or HD 2106 (step S108).
If the embedding of all bits of information w is to finish before the embedding of all pixels is finished, may embeds bogus program information or information w and may be repeated to embed.
Its result, the pictorial information I ' after embedding is created on the RAM.This image I ' for example is stored on HD or the similar mediums at last as file.Afterwards, this image I ' may or be stored on for example removable medium through network delivery.
To explain the process that is used to extract digital watermarking and recovers original picture with Figure 19 below.
Embedding has the pictorial data I ' of digital watermarking to be input to RAM 2105 (step S1201).Aforesaid, input source is not subjected to specific limited.Afterwards, in step S1202, from embedded images I ', extract digital watermark information.The extraction principle of digital watermarking with introduce above identical.
In step S1203, the information of this extraction is decoded so that information w (the information Inf that=flooding information R+ embeds) is stored among the RAM 2105, has so just finished to extract and has handled.Notice that the information Inf that embeds may be presented on the display screen.
In step S1204, each pixel in the pictorial data is all selected as the pixel that will study, and whether the location of pixels that verification is studied is contained among the flooding information R.
If the pixel of being studied is not included among the flooding information R, then processing is eliminated in the combine digital watermark in step S1205.
In addition, if the pixel packets of being studied is contained among the flooding information R, then skips the digital watermarking that is used for this pixel and eliminate processing, and utilize the pixel value of the pictorial data of being imported.
Performed this processing is used for all pixels (step S1206).In case finish the processing that is used for all pixels, then, reappear original picture by eliminating digital watermarking.When needed, this result (original image data) is output to RAM 2105, HD 2106 or similarly installs, and with it demonstration.
In the above embodiments, overflow pixel count than relatively little many of the sum of all pixels of whole image.That is, the information capacity of flooding information R is less relatively, and with the metastable embedding information of digital watermarking R.For example, this situation is applicable to the natural image of similar photo and so on.
In addition, with the type of pictorial data, flooding information R can have big information capacity.For example, when digital watermarking embedded brightness value, when the full luminance of image or when high or low, it is big that flooding information R becomes.
When the capacity of flooding information R became very big, the information capacity that comprises the code sequence W of flooding information R also became very big, and this code sequence w can not be embedded into as digital watermarking.Therefore, the information capacity of flooding information R preferably reduces to minimum.Below introduction is used to reduce the example of process of the capacity of flooding information R.
Figure 12 shows the internal structure that digital watermarking forms device.Correcting unit 2501 and the coding unit 2503 that is used for process information have been to increase with the difference of device shown in Figure 9.Because other operation is saved it thus with shown in Figure 9 identical.
Pictorial data I is imported into correcting unit 2501, and the information capacity of a trimming process with minimizing flooding information R carried out in this unit, and the pictorial data I behind the output calibration " ' and represent to carry out the control information C that proofreaies and correct type.
The process of proofreading and correct is the process that is used to reduce the information capacity of flooding information R.For example, when the brightness of whole image when being high, then carry out the process of the given amount (d) that deducts from all pixel values.In this case, the information of expression " deducting given amount (d) from the pixel value of whole image " is output as control information C.
In addition, except flooding information R and additional information Inf, coding unit 2503 is also encoded to control information C.As shown in figure 20, the code sequence w behind the coding comprises control information C, flooding information R and additional information Inf.
By aforesaid process embed digital watermark.To introduce the image recovery device below.
Figure 13 shows the internal structure of the image recovery device in the present embodiment.Be to increase the decoding unit 2602 of reverse correcting unit 2604 and processing stored information with the difference of device shown in Figure 10.Because other operation identical with among Figure 13 and 10 is then with its omission.
Decoding unit 2602 receives the code sequence w that is extracted by digital watermarking extraction unit 2601, and the sequence w of this input is deciphered obtaining additional information Inf, flooding information R and control information C, and with they output.
Oppositely correcting unit 2604 receives by digital watermarking and eliminates the pictorial data I that the unit is therefrom eliminated digital watermarking " ' and by decoding unit 2602 to its control information C that deciphers; and this unit carries out the opposite process of carrying out with correcting unit 2503 of process, and the pictorial data I behind the output calibration.For example, if the information of expression " deducting given amount (d) from the pixel value of whole image " is transfused to as control information, then correcting unit 2604 is carried out such processing: the pixel value that a specified rate (d) is added to whole image.
By top process, can perfectly reproduce digital watermarking from the pictorial data I ' that is input to the image recovery device and form pictorial data I before.
To introduce definite method of control information C below for example.
Detection is with the Luminance Distribution of the image of embed digital watermark.For example, if the occurrence rate of brightness value big than in the low-light level district then produces and proofreaies and correct with the conversion luminance to the low-light level district.On the contrary, if the occurrence rate of low value big than at high luminance area then produces and proofreaies and correct with the conversion luminance to high luminance area.
In some cases, may carry out following processing.For easy understanding, will analyze the situation that the paper that uses image scanner (every pixel 8 bits) to read to be printed on text and this image that reads are used as visual I below.
General, as shown in figure 21, when reading the file (binary picture) that is printed on text and representing corresponding pixel with brightness value, the distribution of pixel concentrates on the highest and the minimum brightness value part, and does not almost have pixel between them.
In this case, if all pixel values add or deduct a specified value, then overflow in high or low brightness side.Therefore, carry out following the processing.
Because low intensity range is distributed between the brightness value 0 to a, then finds out consistent and satisfy the median luminance value of equation (4) (or equation (4 ')), and replace low brightness values with the brightness value of finding out with 0 occurrence rate.For example, if find out brightness value between 50 to 50+a, then 0 to a brightness value be converted to value between 50 to 50+a (0 is converted to 50,1 is converted to 51 ...).Also carried out similar processing for high intensity range.Be that high intensity range is transformed to lower brightness range.
As the result of said process, especially under the situation of the binary picture that is similar to text,, then can perfectly recover original picture if control information C shows the conversion value between brightness value 0 to a and the brightness region b to 255.In the superincumbent narration, image is represented with luminance component, yet the present invention is not limited to this, and image is the representation in components of available other similar colourity also.
Aforesaid, according to top embodiment, do not embed and handle and when being embedded into any pixel of not overflowing, can perfectly recover original image data as the information of digital watermarking when overflowing pixel (component of numerical data).
Note aforesaidly, the present invention can realize by on computers program of operation.Therefore, the present invention includes a computer program.Because the present invention can be by providing program to realize to computer, the device that then is used for program code is defeated by computer promptly is used for program code stored storage medium also within the scope of the present invention.
Of this sort program code stored storage medium for example, can use floppy disk, hard disk, CD, magneto optical disk, CD-ROM, tape, permanent memory card, ROM or analog.
The computer of the distinct device that the program code that can control alone and provide is complementary, in the time of not only can realizing the function of the foregoing description by such computer, and at synthetic operation program code and OS (operating system) or other software when realizing the function of this embodiment on computers, such program code all comprises within the scope of the invention.
Scope of the present invention in addition also comprises such a case: the program code that provides is stored in memory that is installed on the computer motherboard or the functional extender apparatus that links to each other with computer, the CPU that is installed on this function expansion board or the device carries out some or all of effective procedures on the basis of code instructions, and realizes present embodiment by these processes.
Aforesaid, according to the present invention, when authentication information and digital information is inseparable when being provided with, can detect information and whether be distorted, and as long as initial data do not distorted and it can be recovered.
Can embed digital watermark, and embed the initial numberical data that digital watermarking is arranged can be reproduced.
Because can produce many different embodiment, be not limited to the specific embodiment that is limited but be limited to the appended claims so be appreciated that the present invention without departing from the spirit and scope of the present invention.

Claims (12)

1. one kind is used for by add/deduct that a value is embedded into the information embedding device of described element with additional information from the element of forming numerical data, and this device comprises:
Checkout gear is used to detect such element: after adding computing/subtract computing, its value has surmounted the scope that described element can be assert;
By in conjunction with additional information with produce the generation device of effective embedding information by the information that described checkout gear detects; With
Flush mounting, be used to get rid of such element: when being embedded into numerical data, add computing/subtract after the computing through what embed to handle, its value has surmounted the scope that described element can be assert, and the effective embedding information that will be produced by described generation device is embedded into the element that drops within the scope that described element can assert as digital watermarking.
2. according to the device of claim 1, numerical data wherein is a view data, and
Described checkout gear is used to detect such location of pixels: this pixel value has surmounted the scope that described pixel value can be assert after adding computing/subtract computing.
3. according to the device of claim 1, further comprise at least a code device that carries out compressed encoding in the information that detects to additional information with by described checkout gear, wherein said generation device produces effective embedding information on the basis of the coding result of described code device.
4. according to the device of claim 1, further comprise at least a encryption device of encrypting in the information that detects to additional information with by described checkout gear, wherein said generation device produces effective embedding information on the basis of the encrypted result of described encryption device.
5. according to the device of claim 1, further comprise at least a code device that is converted to error correcting code in the information that detects with additional information with by described checkout gear, wherein said generation device produces effective embedding information on the transformation result basis of described code device.
6. according to the device of claim 1, further comprise and be used for the correcting digital data to reduce the means for correcting of numerical data number of elements, this element has the value that surmounts the scope that described element can assert after adding computing/subtract computing, the information of the calibration substance of wherein said generation device by embedding the described means for correcting of expression produces effective embedding information.
7. information recovery device is used for receiving by the information embedding device of claim 1 information is embedded wherein numerical data, and recovers initial numberical data, and this information recovery device comprises:
The digital watermarking extraction element is used for extracting the information of the numerical data that is embedded into input; With
The digital watermarking cancellation element is used for the information that the foundation expression embeds the element of processing, eliminates the information of embedding from carry out the element that embeds processing, and recovers initial numberical data.
8. according to the device of claim 7, further comprise the code translator that at least one additional information of being extracted by described digital watermarking extraction element is carried out decompress(ion) decoding, described information representation embeds the element of processing, and wherein said digital watermarking cancellation element is eliminated digital watermarking on the basis of the decode results of described code translator.
9. according to the device of claim 7, further comprise the code translator that at least one additional information of being extracted by described digital watermarking extraction element is decrypted decoding, described information representation embeds the element of processing, and wherein said digital watermarking cancellation element is eliminated digital watermarking on the basis of the decode results of described code translator.
10. according to the device of claim 7, further comprise the code translator that the error correcting code of at least one additional information of being extracted by described digital watermarking extraction element is deciphered, described information representation embeds the element of processing, and wherein said digital watermarking cancellation element is eliminated digital watermarking on the basis of the decode results of described code translator.
11. the method for a control information flush mounting, information embedding device wherein is by adding/deduct that from the element of forming numerical data a value is embedded into additional information the described element, and this method comprises:
Detect step, detect after adding computing/subtract computing, its value has surmounted the element of the scope that described element can assert;
By with additional information and in detecting step the detected combined generation step that produces effective embedding information of information; With
Embed step, be used for when being embedded into numerical data, to add/subtract those elements that surpassed the scope that described element can assert after the computing, eliminating is outside embed handling, and the effective embedding information that will produce in producing step is embedded into the element that drops within the scope that described element can assert as digital watermarking.
12. the method for a control information recovery device, this device receive by the information embedding device in the claim 1 information is embedded into wherein numerical data, and recovers initial numberical data, this method comprises:
The digital watermarking extraction step is used to extract the information of the numerical data that is embedded into input; With
The digital watermarking removal process is used for the information according to the element of pointing out to embed processing, eliminates the information that embeds the element of handling from having carried out embedding, and recovers initial numberical data.
CNB2006100844829A 2001-02-09 2002-02-09 Information processing apparatus and its control method, computer program and storage medium Expired - Fee Related CN100403765C (en)

Applications Claiming Priority (3)

Application Number Priority Date Filing Date Title
JP034057/2001 2001-02-09
JP2001034057A JP3854804B2 (en) 2001-02-09 2001-02-09 Information processing apparatus, control method therefor, computer program, and storage medium
JP040787/2001 2001-02-16

Related Parent Applications (1)

Application Number Title Priority Date Filing Date
CNB021054037A Division CN1305296C (en) 2001-02-09 2002-02-09 Information processing device and control method, computer program and storage medium

Publications (2)

Publication Number Publication Date
CN1893530A CN1893530A (en) 2007-01-10
CN100403765C true CN100403765C (en) 2008-07-16

Family

ID=18897721

Family Applications (1)

Application Number Title Priority Date Filing Date
CNB2006100844829A Expired - Fee Related CN100403765C (en) 2001-02-09 2002-02-09 Information processing apparatus and its control method, computer program and storage medium

Country Status (2)

Country Link
JP (1) JP3854804B2 (en)
CN (1) CN100403765C (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106156655A (en) * 2015-03-26 2016-11-23 中国科学院声学研究所 The compressing file of a kind of facing cloud storage and authentication method

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007528017A (en) * 2003-07-11 2007-10-04 コーニンクレッカ フィリップス エレクトロニクス エヌ ヴィ Digital watermark embedding and detection
WO2005106870A1 (en) * 2004-05-04 2005-11-10 Koninklijke Philips Electronics N.V. Access authorization across processing devices
JP2008312065A (en) * 2007-06-15 2008-12-25 Canon Inc Image processor and its method
JP5423512B2 (en) * 2010-03-18 2014-02-19 日本電気株式会社 Broadcast data stream protection system, method, and user terminal identification program
CN117597885A (en) * 2021-07-07 2024-02-23 索尼半导体解决方案公司 Hash generation device, hash determination device, and system

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH065936A (en) * 1992-06-16 1994-01-14 Seiko Epson Corp Superconducting transistor
JPH10164549A (en) * 1996-11-28 1998-06-19 Ibm Japan Ltd System for hiding in authentication information is image and image authentication system
US5809139A (en) * 1996-09-13 1998-09-15 Vivo Software, Inc. Watermarking method and apparatus for compressed digital video
CN1246239A (en) * 1997-09-02 2000-03-01 皇家菲利浦电子有限公司 Method and arrangement for detecting watermark

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPH065936A (en) * 1992-06-16 1994-01-14 Seiko Epson Corp Superconducting transistor
US5809139A (en) * 1996-09-13 1998-09-15 Vivo Software, Inc. Watermarking method and apparatus for compressed digital video
JPH10164549A (en) * 1996-11-28 1998-06-19 Ibm Japan Ltd System for hiding in authentication information is image and image authentication system
CN1246239A (en) * 1997-09-02 2000-03-01 皇家菲利浦电子有限公司 Method and arrangement for detecting watermark

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106156655A (en) * 2015-03-26 2016-11-23 中国科学院声学研究所 The compressing file of a kind of facing cloud storage and authentication method
CN106156655B (en) * 2015-03-26 2019-01-08 中国科学院声学研究所 A kind of compressing file and authentication method towards cloud storage

Also Published As

Publication number Publication date
CN1893530A (en) 2007-01-10
JP3854804B2 (en) 2006-12-06
JP2002237941A (en) 2002-08-23

Similar Documents

Publication Publication Date Title
KR100465950B1 (en) Information processing apparatus and its control method, storage medium storing computer program
US8612765B2 (en) Security based on subliminal and supraliminal channels for data objects
Zhang et al. Reference sharing mechanism for watermark self-embedding
Zhang et al. Reversibility improved data hiding in encrypted images
JP4625841B2 (en) System and method for digital content security
JP3804012B2 (en) Document image alteration determination method and system, and control program therefor
US6005936A (en) System for embedding authentication information into an image and an image alteration detecting system
EP0952728B1 (en) Method and apparatus for watermarking data sets
WO2005091547A2 (en) Watermark payload encryption methods and systems
CN111988144B (en) DNA one-time pad image encryption method based on multiple keys
CN113114869A (en) Ciphertext domain high-capacity image reversible data hiding method based on MSB prediction
CN100403765C (en) Information processing apparatus and its control method, computer program and storage medium
Şahin et al. Review of the Literature on the Steganography Concept
Duluta et al. Secure communication method based on encryption and steganography
Agarwal et al. Encrypted transfer of confidential information using steganography and identity verification using face data
Wazery et al. A hybrid technique based on RSA and data hiding for securing handwritten signature
JP2003092676A (en) Data processing method and apparatus, its program and storage medium
Munasinghe et al. Video steganography
Gao et al. Dual mode data hiding in fully encrypted images with pixel-shuffling for cloud applications
US20040034779A1 (en) Method and apparatus to authenticate digitally recorded information
JP5343659B2 (en) Image processing apparatus and image processing program
CN114119332A (en) Block chain-based digital watermark writing and reading method and device
Ghadi et al. Enhanced payload volume in the least significant bits image steganography using hash function
Coding A New Steganography Techniques using
Mulya et al. Least Significant Braille Method in Steganography Using Digital Image Media For Security Message

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C14 Grant of patent or utility model
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20080716

Termination date: 20160209

CF01 Termination of patent right due to non-payment of annual fee