CN102036072A

CN102036072A - Method and system for watermark detection

Info

Publication number: CN102036072A
Application number: CN2010105705585A
Authority: CN
Inventors: 赵健; 艾克哈德·科赫
Original assignee: Thomson Licensing SAS
Current assignee: Thomson Licensing SAS
Priority date: 2005-12-05
Filing date: 2005-12-05
Publication date: 2011-04-27

Abstract

The invention discloses a method and a system for watermark detection. The method for watermark detection comprises the following steps of: receiving a preprocessed watermark unit; receiving contents printed with a watermark; acquiring a plurality of coefficient values from the contents printed with the watermark; and acquiring a bit value of the watermark from the plurality of coefficient values, wherein the bit value is acquired by associating the current coefficient with the plurality of coefficient values encoded in the watermark unit to determine the optimized matching.

Description

The method and system that is used for watermark detection

The application is that international application no is that PCT/US2005/043895, international filing date are on December 5th, 2005, national applications number to be 200580052229.9, to be entitled as the dividing an application of application for a patent for invention of " for the encoded content pressing watermark ".

Technical field

The present invention relates to digital watermarking, more specifically, relate to embedding in encoded content, remove/replace and detect digital watermarking.

Background technology

In order to use content, encoded content needs " decoding " process.Some examples of encoded content comprise MPEG-1, MPEG-2, H264/AVC, WMA, MPEG4, JPEG2000, MP3, PDF, Windows Word, Postscript etc., and their encryption version.

For a kind of art methods of encoded content pressing watermark is to insert watermark signal in structure and syntax elements.US 6,687,384th, embed the example of data in the syntax elements of the coded bit stream such as MPEG-1 and MPEG-2.But this class watermark can't stand change of format or digital-to-analogue conversion and exist.

For the another kind of art methods of encoded content pressing watermark is watermarked by adding noise for the DCT coefficient that is used for the mpeg encoded content.The example of this method is described in " Digital Watermarking of MPEG-2 Coded Video in the Bit Stream Domain " by F.Hartung and B.Girod, Proc.IEEE ICASSP, and pp.2621-4, April 1997.The method of Hartung and B.Girod is not used cognition technology.

For another art methods of encoded content pressing watermark is directly to revise encoded content.US6,373,960 (hereinafter referred to as " Conover ") have described in the mpeg compressed video bit stream by revising the watermark drawing method of some DCT coefficient as follows, i.e. the mode that remains unchanged after revising of the entropy coding length of these coefficients.Do not specify concrete watermark embedding method among the Conover.Select the method for watermark locations to be limited in coefficient among the Conover with 0 company zero (run-zero).The coefficient of Conover is in high-frequency domain.The method of Conover neither comprises pretreatment stage, does not also exist in the later phases of describing to be used for any " substitution value " that actual watermark is inserted.Generate and use the notion of " watermark unit " in Conover, not disclosing Anywhere or instructed.

Be used to the art methods of encoded content pressing watermark not have the preprocessor of " substitution value " that produce some part that is used for encoded content.Be used to the art methods of encoded content pressing watermark not comprise the encoded content of further being covered up (mask), fuzzy (obfuscate), upsetting (scramble) or encryption (being referred to as " encryption ").

Summary of the invention

The present invention is by being the encoded content pressing watermark in that the select location of encoded content is watermarked.This is to finish by the original value in the encoded content is replaced with their substitution value.Each original value can have one or more substitution value and each substitution value comprises a watermark signal.In addition, can be by removing watermark with original value displacement substitution value or by replacing a substitution value with another substitution value.

By selecting one of a plurality of substitution value to replace the original value of relevant position in the encoded content, can realize following target:

-substitution value and original value are measure-alike, and any one displacement of a plurality of substitution value is created and introduced the perception illusion in the encoded content of the form that meets qualification and the digital content of not being on the permanent staff.

-substitution value can be improved the quality of encoded content.For example, original value may be may not be effective value also.In other words, if there is not the displacement of one of its a plurality of substitution value, encoded content may not be legal form.At original value is under the another kind of situation of effective value, if there is not the displacement of substitution value, original value may be introduced the degradation of encoded content.

-each substitution value comprises watermark signal.These watermark signals can have the information unit of one or more embeddings together or alone with other watermark signals of other positions in the encoded content.An information unit is made up of one or more bits.

Comparing with the data volume of encoded content must be very little for the data volume of-substitution value (being defined by the data size of each substitution value and the number of substitution value).

Be " local cypher " (localized encryption) directly for an important goal of encoded content pressing watermark.Local cypher allows expressly and the correspondence between the ciphertext.For example, expressly the unit comprises component c1, c2 ... cn.After local cypher, ciphertext is by c1 ', c2 ' ..., cn ' forms, and wherein ci ' is the ci (1＜i＜n) of encryption version.The simple examples of local cypher is that content is divided into a plurality of parts and encrypts each part respectively.For example, under extreme case, each coefficient or coefficient sets are encrypted respectively.One type the encryption that can realize local cypher is that selective encryption or part are encrypted.Selective cryptographic method " understanding " content syntactic structure (for example, the MPEG-2 structure) and some part of encrypted content optionally only, rather than content (audio or video) is considered as binary data stream (being also referred to as " local cipher ").Selective encryption occurs in after the compression.Can't permit in encrypted and have commercial value, though partial content can be visible.Some selective encryption schemes can keep format of content, bit rate and the size of not encrypted by selectivity.Other selective encryption schemes can increase bit rate or require dedicated decoders.

With the MPEG-2 content is example, and simple selectivity algorithm is only encrypted the I frame.The MPEG of selective encryption can be effective mpeg stream.Though there is not the knowledge of corresponding I frame, P among the MPEG-2 and B frame are with valueless at all, and the major part of MPEG video is because the relevant reason of interframe is still visible and is main from the unencryption I piece in P and the B frame.Other selective encryption schemes comprise the encryption of MPEG-2 stem and/or the encryption of DCT coefficient.The DCT coefficient is divided into " DC coefficient " and " AC coefficient ".The DC coefficient is the coefficient that has zero frequency in two dimensions, and the AC coefficient is the residual coefficient with nonzero frequency.The AU DC coefficient of all I pieces or the part value of AC coefficient can be encrypted.

A kind of system and method that is used for inserting at encoded content watermark has been described, comprise the received code content, receive at least one watermark unit, and, be embedded with watermark signal in the wherein said substitution value directly with a plurality of bits of substitution value displacement since a position.Also described the apparatus and method that are used to generate the digital watermarking unit, comprised and having selected that the value of this position encoded content is replaced replaced value, and calculated substitution value as upper/lower positions.The system and method that is used for the permutation encoding watermark in content has also been described in addition, comprise the received code content, receive at least one watermark unit, and with a plurality of bits that as upper/lower positions begin of following value displacement from encoded content, wherein said value is designated in described at least one watermark unit, and wherein said in addition position is designated in described at least one watermark unit.

The system and method that is used for inserting at encoded content watermark has also been described in addition, comprise the received code content, receive at least one watermark unit, receive bit sequence (wherein said bit sequence comprises the watermark load information), and carry out one of following operation: directly with a plurality of bits of substitution value displacement since a position, keep a plurality of bits of beginning from this position constant based on the bit value of the bit sequence of watermark load information, embedded watermark signal in the wherein said substitution value.Also describe the system and method that is used for watermark detection, comprised receiving at least one watermark unit, received the content that is suppressed with watermark, from the content that is suppressed with watermark, fetched a plurality of coefficient values, and the bit value of fetching watermark from described coefficient value.

Description of drawings

When reading in conjunction with the accompanying drawings following detailed description, the present invention will get the best understanding.Concise and to the point figure below of describing below these figure comprise:

Fig. 1 illustrates the block diagram that principle according to the present invention is used for the pretreated workflow of watermark embedding;

Fig. 2 has described watermark unit of the present invention (U);

Fig. 3 is the flow chart according to the method for detecting watermarks of principle of the present invention;

Fig. 4 is the block diagram of playback apparatus, and described playback apparatus receives and wherein embeds the encoded content that watermark unit is arranged.

Embodiment

For the encoded content pressing watermark can be divided into three independent steps:

● choice of location, this is to the selection as upper/lower positions, the substitution value displacement that the value of described position encoded content can involved watermark signal.

● substitution value is calculated, and this is to the determining of substitution value, so that substitution value has the identical bit number of value that will replace with this substitution value in encoded content, and this class is replaced the perception that can not cause content and changed.In addition, these substitution value comprise watermark signal.

● watermark embedding/insertion, this is the actual displacement of one of substitution value to the encoded content intermediate value.

The one or two step can be pretreated.As pretreated result, one group of watermark unit (WU) produces.WU comprises all information that are used for actual embedding/insertion watermark.Watermark pre-processor is used as encoded content as input and watermark keys, and the sequence of output watermark unit.For each WU generates under the situation of a substitution value, the watermark load can be changed into the additional input of preprocessor only.These watermark unit are as the metadata of encoded content, as the segregated channel multiplexing with content, cover close (steganographic) data or are integrated in the final content as storage in physical medium (CD, tape, hard disk etc.) or the individual files by network (TCP/IP, satellite etc.) transmission as what hide in syntax elements or content.

By selecting watermarked load information in the substitution value from WU.WU has at least one substitution value.When each WU has only a substitution value, there are two kinds of methods to embed/insert watermark information (load).First method is to embed single fixing watermark load by all original values among the WU are replaced with their substitution value.Second method allows to embed various watermark loads by switching between displacement that is used for WU or non-displacement.For example, the positive bit value of displacement indication, the negative bit value of non-displacement indication.For watermarked load " 00101001 ", the original value among first and second WU is not replaced, and the original value among the 3rd WU is replaced by its substitution value.For the WU that has more than a substitution value, each substitution value comprises the watermark signal of the different information units that may represent the watermark load.Select the substitution value that is used to replace based on the watermark load.For example, in order in WU, to embed a bit, only need two substitution value V1 and V2.For embedding ratio paricular value ' 0 ', selection V1 comes the original value in the permutation encoding content; For embedding ratio paricular value " 1 ", select V2.Utilize two value V1 and V2, can express ' 0 ' and ' 1 ' bit.If four substitution value (V1, V2, V3 and V4) are arranged, then can embed dibit information (promptly ' 00 ', ' 01 ', ' 10 ' and ' 11 ').Utilize more substitution value at same position place, can embed more bits, this allows very effective information to embed.

Usually in the 3rd step " watermark insertion ", receive the watermark load.The watermark load can be by the component stores or the calculating of watermark compact system outside.The watermark load information normally identifies the identifier of the date and time of recipient, playback apparatus (manufacturer, model and/or sequence number) or content playback uniquely.The difference of original value V and its substitution value can be stored among the WU to reduce the size of WU potentially.WU can further be compressed.

Importantly protect WU to avoid unwarranted visit or modification because by this information not only the watermark compact system be vulnerable to various attack, and the hacker inserts false watermark easily, perhaps change or remove existing watermark.If WU is stored and sends as covering close (watermark compacting) data, then can suppress key and control visit by watermark.If WU stores and transmits by any other passage (channel), then need WU is encrypted.

As previously mentioned, first and second steps can be used as pre-treatment step and carry out.Therefore, before the 3rd step, be not the encoded content pressing watermark.The 3rd step was carried out the watermark embedding by being worth with in the substitution value permutation encoding content of appointment among the WU some.These values can be encoded syntax elements, as package identification, stem, quantization table, the Huffman table that is used for entropy coding, encoded coefficient or encoded motion vector.WU specifies (in the substitution value) watermark signal that where watermark signal embeds and which is possible can be positioned at these positions by selecting suitable substitution value.

Fig. 1 is the block diagram that the pretreated workflow of the generation WU that is used for encoded content is shown.Fig. 2 has described the watermark unit according to principle of the present invention.Each WU is by vector (P, L, C, { V}, V1, V2 ..., Vn) portrayal, wherein P is the position of the original value V of the replaced value displacement of possibility in the future in the encoded content, L is the occupied bit number of original value V that starts from P in the encoded content, and C utilizes one group of global position of the coefficient (for example, DCT or small echo quantization parameter) of entropy coding method coding in (encoded content) L bit.The final stage that entropy coding such as the Huffman coding is applied to encode usually is to produce encoded content.V1, V2 ... Vn is that the effective substitution value of V and each of these values comprise watermark signal.(b represents that co) wherein for the unit of a video content, ch is the passage index to each coefficient positions for ch, f, and f is a frame index, and b is the piece index, and co is the coefficient index in the piece with c.If it keeps the form accordance and does not exist the perception to content to influence when substitution value displacement currency, then this substitution value is effective.In addition, V1, V2 ... Vn occupies the L bit as V in encoded content.{ V} indication original value is optional in embedding/insertion process.In watermark removal process, may need original value.

Original value V can comprise the coefficient of one or more coding forms.When WU generates, if input is the encoded content such as MPEG-2 or MPEG-4, then entropy coding (being variable length code (VLC)) at first needs " being untied " (undone) with the visit coefficient, finds appropriate location and substitution value subsequently and stores unprocessed form and these substitution value.Substitution value Vi can be corresponding to the coefficient identical with V, but Vi may be corresponding to the coefficient more more or less than V in some cases.

8 * 8 of the typical cases of the DCT coefficient of the quantification among the MPEG-2 are shown below.Most of high-order coefficients have been quantified as 0.

12 34 0 2 0 0 0 0

87 0 0 1 0 0 0 0

15 0 0 0 0 0 0 0

0 0 0 0 0 0 0 0

After " it " font scanning (zig-zag scanning), DCT coefficient sequence to be sent looks like like this:

12?34?87?16?0?0?2?0?0?0?0?0?0?1?0?0?0?0?0?0.....

First DCT coefficient (12) sends via independent Huffman table.After even zero-bit rank (Run-Level) resolved, residual coefficient with relevant company zero was:

34|87|16|0?0?2|0?0?0?0?0?0?1|0?0?0?0?0?0......

Utilize the DCT coefficient table of listing below zero of Moving Picture Experts Group-2 appointment, these coefficients are encoded into following 6 variable length codes (VLC) (representing with binary bits):

0000?0100?0000?0000?0010?0010|0000?0100?0000?0000?0101?0111|0000?00000111?110|

MPEG-2DCT coefficient table zero (part)

Variable length code (VLC)	Connect zero	The rank, position
			?10	The piece end
?...	...	...
			?0001?10s	1	2
?0001?11s	5	1
			?0001?01s	6	1
?0001?00s	7	1
			?...	...	...
?0000100s	2	2
			?...	...	...
?0000?01	Overflow (Escape)
			?...	...	...
0000?0000?0111?11s	0	16
			...	...	...

(note: the symbol on last bit ' s ' the indicating bit rank among top each VLC: ' 0 ' is just representing, ' 1 ' expression is negative)

The MPEG-2 coding on the rank, company zero-sum position of ESCAPE sign indicating number back

For the ease of reading, above-mentioned VLC (being also referred to as entropy coding) is cut apart with " | ".First and second DCT coefficients use fixed length code to encode to connecting rank, zero-sum position according to table " the MPEG-2 codings on the rank, company zero-sum position of ESCAPE sign indicating number back ", and come remaining DCT coefficient is encoded according to " MPEG-2DCT coefficient table zero ".Last VLC is special " piece end " sign indicating number, and remaining coefficient is 0 in the indicator collet.

Substitution value can be any part of above-mentioned coded-bit.For example, a substitution value can be by the part of VLC, one or more continuous VLC, and perhaps VLC adds that the part of next VLC forms.

In the following example of WU, substitution value is made up of a VLC, and it is the 5th VLC:

(P0+71，7，C，0001010，0001110，0001100，0001000)

Each element among this WU is explained as follows:

● P0 is the original position of this piece in encoded content." 63 " are with respect to the relative position of this piece starting point (bitwise), 7 indicating length L, and C is made up of following 7 coefficients:

C (f, ch, b, 7), c (f, ch, b, 8) ... c (f, ch, b, 13), wherein f is the present frame index, ch is when the prepass index, and b is the current block index of this piece in this example.

● substitution value is of a size of 7 bits, that is, and and L=7.

● 0001010 is original value (V), has three substitution value: 0001110 (V1), 0001100 (V2) and 0001000 (V3).These substitution value are encoded as following rank, company zero-sum position value (with reference to MPEG-2DCT coefficient table zero) respectively:

0001110: connect zero=5, rank=1, position

0001100: connect zero=1, rank=2, position

0001000: connect zero=7, rank=1, position

Connect the number sequence that zero (R) and rank, position (L) are followed L to expression with the individual zero beginning back of R.For example, connect rank, zero-position to (5,3) (promptly connecting zero=5 and rank=3, position) expression 0000 03.Therefore, above-mentioned substitution value was represented following coefficient before entropy coding:

000001 (connecting zero=5, rank=1, position)

02 (connecting zero=1, rank=2, position)

00000001 (connecting zero=7, rank=1, position)

Because the 5th VLC is last VLC that nonzero coefficient is encoded in this piece, so substitution value is not necessarily corresponding to the same factor of being encoded by V.Under the situation of V2 and V3, V2 only to 2 coefficient codings and V3 to 7 coefficient codings, rather than to 8 coefficient codings among the V1.

If the watermark inserter in the step 3 is selected first substitution value in the WU example, then this piece will become following quantization DCT coefficient:

12 34 0 2 0 0 0 0

87 0 0 0 0 0 0 0

16 0 1 0 0 0 0 0

0 0 0 0 0 0 0 0

Another example of encoded content is to utilize H.264 (MPEG4 class (Profile) 10) content encoded.MPEG-2 and H.264 between the main distinction of the entropy coding that exists have:

● H.264 both support CAVLC (context-adaptive variable-length encoding), also supported CABAC (context adaptive binary arithmetic coding).

● H.264 also can use 4 * 4 DCT piece, rather than among the MPEG-2 8 * 8.

● be different from the AVLC (adaptive variable length coding) among the MPEG-2, CAVLC does not encode by contraposition rank and company's zero and uses littler VLC.

● the VLC table can based on contextual information switch, rather than the table of the fixedly VLC among the MPEG-2.

H.264 in owing to carry out entropy coding based on context, therefore than MPEG-2, its encoded radio of more difficult prediction after coefficient is modified.But can force H.264 encoder reduces this class based on contextual coding during encoding.For example, encoder can use fixing NumTrail table (for example, NumTrail table 0), rather than based on the NumTrail table of some attribute of contiguous block.Another example is to use fixed length code xxxxyy to come to quantize the number (NumCoeff) of non-zero of DCT pieces and the number (T1) of tail one is encoded to being used for 4 * 4, wherein xxxx is used for the individual number encoder (0-16) to non-zero, and yy is used for the individual number encoder (0...3) to tail one.Can also only use 4 * 4 piece by the forced coding device.In the example below, some above-mentioned application of rules is in to 4 * 4 block encoding.

1.4 * 4 quantization DCT coefficients:

0 3 -1 0

0 -1 1 0

1 0 0 0

0 0 0 0

After the scanning of " it " font, the coefficient in this piece: 030 1-1-1,01 0....0

2.CAVLC coding is made up of following five steps:

A encodes to the number (NumCoeff) of non-zero and the number (T1) of tail one: (5,3).The maximum number of T1 is restricted to 3.Use 17 * 4Huffman table (as follows) with these two numerals (NumCoeff and T1) be encoded into bit " 0001011 " (NumCoef:0-16, T1:0-3).

NumTrail 17 * 4Huffman table

B is in reverse order to the encoding symbols of tail one (if any) (0 expression just, 1 expression is negative): 0,1,1.

C is in reverse order to remaining nonzero coefficient coding: 1,3.Nonzero coefficient utilizes a rank VLC 0 table and position rank VLC 1 table to be encoded as " 1 " and " 0010 " respectively.

D to from the off to the end the total remainder (TotalZeros) till some nonzero coefficients encode.Maximum total remainder is 16-NumCoeff.If NumCoeff is 16, then TotalZeros must be 0.If NumCoeff is 0, then do not need further coding, that is, and to the end-of-encode of this piece.For 15 kinds of remaining situations, every kind of situation uses the Huffman table to come total remainder encode (seeing the TotalZero table).In this case, TotalZeros is 3, utilizes the TotZeros table to be encoded as " 1110 " [NumCoeff=5].

The TotZeros table

E to from the off to the end positions of all till some nonzero coefficients zero encode.In this case, all positions of zero are encoded as " 001101 ".

This final coded-bit of 4 * 4 is comprised 25 bits: 000,101 11,011 1 10,010 11110 I 001101.An example of substitution value is whole 4 * 4 encoded coded-bit.Another example of substitution value can be that picture is at above-mentioned steps c) as described in to the bit of the encoding symbols of tail one.In this case, change the symbol that substitution value only changes tail one, therefore do not change total bit length of encoding block.

Another example of encoded content is the JPEG2000 picture.JPEG2000 uses digital wavelet transform (DWT), rather than dct transform.The DWT of pixel formula (pixeled) image is by calculating with vertical, low pass and high pass filter image pixel successive applications level, and the value of generation is called as " wavelet coefficient ".Small echo is to continue the only wave of oscillation in one or several cycle.When each iteration, the wavelet coefficient that only passes through low pass (low-pass only) filtration in last cycle is extracted, pass through low pass vertical filter and high pass vertical filter then, and the result of this process is transmitted by low pass level and high pass horizontal filter.The coefficient set that produces is grouped into four " subbands ", i.e. LL, LH, HL and HH subband.Each iteration is corresponding to a certain " layer " or " level " of coefficient.The ground floor of coefficient is corresponding to the highest resolution level of image, and last one deck is corresponding to the lowest resolution level.An example of substitution value that is used for the picture of JPEG2000 coding is the encoded LL subband at certain layer place (for example, lowest resolution).

The plain mode that produces WU is to use " trial-and-error method " by each WU is carried out following steps:

A) for each entropy coding (for example, the VLC among the MPEG-2), the candidate who selects a value to be worth as an alternative at random.

B) replace existing value with this candidate after, determine by verifying the form accordance whether this candidate is effective value.If not, then return step a).

C) exchange after this candidate and the existing value, check that perception changes.Can accept if perception changes, be the substitution value that is used for WU with this candidate record then.Otherwise, return step a)

Repeat above-mentioned steps until producing the WU that all need.

In one embodiment, can utilize the existing watermark compact system that works in transform domain to produce WU.At first utilize selected watermark compact system, generation has two copy (A and B that are suppressed with watermark of two different original coding contents that copy of load information, all be coding form): first copy comprises the watermark load of each bit for " 1 ", and another copy comprises the watermark load of each bit for " 0 ".Step 1 and 2 (see figure 1)s can be carried out as follows to select and to produce WU by scan A and B until satisfying following condition:

● (choice of location) P1a and P2a are original position and the final positions (entropy coding before their position be illustrated as C) of a plurality of continuous programming code coefficients in encoded content A, P1b and P2b are original position and the final position of same-code coefficient in encoded content B, and (P2b-P2a) He (P1b-P1a) must equate (and all equaling length L), and

● (substitution value calculating) supposes that at least one bit watermark load information will be embedded in (A and B the two) code coefficient.Write down in the following manner watermark unit (P, L, C, V, V1, V2): P=P1a, L=P2a-P1a, V are unavailable in this example, V1 is the value from position P1a to P2a among the A, V2 is the value from position P1b to P2b among the B.

● in order to select N watermark unit, repeat above-mentioned second step.

In alternate embodiment, a bit watermark load is encoded as in the intermediate frequency relation between two selected quantization parameters.By the scanning encoding content, can utilize following condition locate and definite watermark unit (P, L, C, V, V1, V2 ... Vn):

● V comprise between a plurality of continuous programming code coefficients and these coefficients and/or among have at least a relation (concerning that kind/classification is by pre-defined).For example, if C1, C2 are two continuous code coefficients, then exist and concern C1＞C2+T, wherein T is the threshold value that can be used to regulate relationship strength, relationship strength and then definite watermark robustness.

● according to the sensor model of watermark compact system, can revise a plurality of coefficients and establish another kind of relation and do not follow the perception of content to change.For example, C1 and C2 can be revised as C1 ' and C2 ' respectively to form opposite relation, i.e. C1 '＜C2 '.The new value record that will comprise C1 ' and C2 ' is substitution value Vi (1＜=i＜=n).Repeat this step until finding all substitution value.

● in order to select N watermark unit, repeat above step.Ideally, the top n watermark unit that comprises maximum coefficients is selected, so that can set up maximum relations potentially in each watermark unit.In the watermark compact system of being quoted, a kind of relation can be represented a bit watermark load, and the watermark unit that therefore has maximum relations can have the maximum capacity of the watermark load of the more bits of embedding.

In another example of watermark compact system, a bit watermark load is encoded as the relation between the property value of DCT/ wavelet coefficient or two groups of pixel values.The typical case of attribute comprises the energy in mean flow rate, average color histogram distribution or a certain frequency subband.

A principle of the present invention is the preliminary treatment encoded content and generates the substitution value that comprises the watermark throttle signal.The clearly part of definition that will be used for the permutation encoding content after these substitution value.Though can be applied to preliminary treatment (being step 1 and 2) and actual the insertion (step 3) can be suppressed algorithm at the new watermark of the special exploitation of this scheme with having the watermark compact system now.

A kind of simple and general method of detecting watermarks is the substitution value that current coefficient in the watermark content and the substitution value of WU are associated and mate most with current coefficient to find out.In order to carry out this class association, need to use corresponding entropy decoder that substitution value is decoded into coefficient value.Each substitution value of WU can comprise the watermark signal corresponding to the watermark load information.By in all WU, seeking the substitution value of coupling, can obtain the watermark load information.Watermark detection can be carried out in a plurality of ranks.In the first order, need P, L and the C element of WU.The number of supposing WU is N.For each WU, obtain coefficient value according to overall coefficient index C.If the input content is base band or encodes with the extended formatting that is different from the form that uses in watermark embeds, then at first content is transformed into watermark and embeds identical coded format.By observe between the coefficient and/or among, perhaps the relation between the property value of coefficient value or pixel value obtains original watermark load bit.Fig. 3 is the flow chart of watermark detection process of the present invention.Initialization index in 305.An acceptance/reception N watermark unit and in 315, receive suspicious content in 310.Test indices is to determine whether to detect/handled all watermarks in 320.If index is greater than watermark number N, then process is finished.If index is not more than N, then process advances to and obtains 325 of coefficient.In 330 according between the coefficient/among relation obtain bit value.Increase index and process advance to step 320 in 335.

If obtain very weak relation from WU, then, visit other elements V and the V1 of WU in order to provide extraneous information to determine relation, V2 ... Vn will be helpful.By with current coefficient with at V, V1, V2 ... the coefficient of encoding among the Vn is associated, and can determine better to concern.The highest degree of association value is indication V, V1, and V2 ... the degree of association of the most close coupling between one of Vn and the current coefficient.In other words, current coefficient and in Vi (the highest degree of association indication coupling among 1≤i≤n) between the coefficient of coding.For suspicious content, may need original contents to be used for further analysis with related with serious attack.

The another kind of mode of using WU to send the content of copyright protection is to use WU to recover the content of damaging before playback.In this case, be delivered to the encoded content of playback/rendering apparatus or apparent damage on form/structure (invalid form) or apparent damage in terms of content (content of degradation).For the content that form damages, decoder can't be decoded to it.Relate to content if damage, then this content can be decoded and broadcast but the broken parts of content may have various visual artifacts such as noise, warning and random pattern.

For utilize given watermark unit (P, L, V, V1, V2 ... Vn) produce the content that form damages, the L that the starts from P bit in the encoded content that can be simply will be sent with the random value displacement.In order to guarantee that further random value has damaged form really, can get rid of random value simply as the practical entropy coding.In order in the content of interior perhaps degradation, to introduce visual artifacts, select away from original value V and every other substitution value V1, V2 ... effective substitution value of Vn is come the currency in the permutation encoding content.A kind of exemplary application of the form that damages or the quality of degradation is an antagonism safety a kind of special equipment that endangered or that be used for content piracy or a kind equipment.Selection causes these substitution value of the content of the form that damages or degradation only to be used at the replacement of specific watermark load described watermark load marking equipment or device class.

If known WU then can be by becoming original value V to remove watermark content recovery in each WU.In order to rewrite the existing watermark in the encoded content, can be according to the load information V1 from each WU simply, V2 ... select appropriate value to come to comprise in the permutation encoding content value of existing watermark among the Vn.Removable watermark support many some for watermark use in of great use.Many examples for watermark are to embed modifier (forensic mark) in each step of content post-production.For fear of a plurality of watermarks the accumulation of the perception degradation that may introduce, some or all watermark before may wishing before embedding new watermark, to remove.Fig. 4 illustrates the example of the watermark inserter in the playback apparatus.The watermark inserter comes watermarked by the substitution value displacement that utilizes appointment in WU by some the value/coefficient in the encoded content of equipment reception.Be used for the copyright protection state of copy-protected content for switching, interchangeable watermark also of great use.For example, for content element, after once duplicating, may wish to change watermark information into " must not duplicate " from " duplicating once " with watermark information that indication " duplicates once ".There are a lot of methods can avoid any potential conflict of a plurality of watermarks (being the interface between a plurality of watermarks).A kind of method is to use unique subband for per generation watermark.Another kind method is the WU that selects to have the position that is different from previous watermark.

Should be appreciated that the present invention can implement in various forms of hardware, software, firmware, application specific processor or its combination.Preferably, the present invention implements as the combination of hardware and software.And the present invention preferably implements as the application program that is tangibly embodied on the program storage device.This application program can upload to the machine that comprises any suitable framework or be carried out by this machine.Preferably, this machine has such as one or more CPU (CPU), is depositing at random on the computer platform of the hardware phase memory (RAM) and (one or more) I/O (I/O) interface and implement.Computer platform also comprises operating system and micro-instruction code.Various processing described herein and function can be via the part of the micro-instruction code of operating system or the part of application program (or its combination).In addition, various other ancillary equipment can be connected to computer platform, such as additional data storage device and printing device.

Should also be understood that because some construction system assembly and the method step described in the accompanying drawing are preferably implemented in software, so the actual connection between the system component (or treatment step) depends on that the mode that the present invention is programmed may be different.The instruction of given this paper, those of ordinary skills will expect of the present invention these with similar execution mode or configuration.

Claims

1. method that is used for watermark detection, described method comprises:

Reception is through pretreated watermark unit;

Reception is suppressed with the content of watermark;

From the described content that is suppressed with watermark, obtain a plurality of coefficient values;

And the bit value of from described a plurality of coefficient values, obtaining described watermark, wherein said bit value is by current coefficient being associated with a plurality of coefficient values of encoding in described watermark unit to determine that optimum Match obtains.

2. method according to claim 1, wherein said bit value be according between described a plurality of coefficient values/among relation from described a plurality of coefficient values, obtain.

3. method according to claim 2, wherein said relation are that weak and current coefficient is associated to determine coupling from described a plurality of coefficient values of encoding described watermark unit.

4. method according to claim 1, wherein said is the output of outside pretreatment unit through pretreated watermark unit, and a plurality of coefficient value is based on, and described positional information in pretreated watermark unit obtains from the described content that is suppressed with watermark.

5. system that is used for watermark detection comprises:

Be used to receive device through pretreated watermark unit;

Be used to receive the device of the content that is suppressed with watermark;

Be used for obtaining the device of a plurality of coefficient values from the described content that is suppressed with watermark; And

Be used for obtaining from described a plurality of coefficient values the device of the bit value of described watermark, wherein said bit value is by current coefficient being associated with a plurality of coefficient values of encoding in described watermark unit to determine that optimum Match obtains.

6. system according to claim 5, wherein said bit value be according between described a plurality of coefficient values/among relation from described a plurality of coefficient values, obtain.

7. system according to claim 6, wherein said relation are that weak and current coefficient is associated to determine coupling from described a plurality of coefficient values of encoding described watermark unit.

8. system according to claim 5, wherein said is the output of outside pretreatment unit through pretreated watermark unit, and a plurality of coefficient value is based on, and described positional information in pretreated watermark unit obtains from the described content that is suppressed with watermark.