CN103975605B - Watermark extracting based on tentative watermark - Google Patents
Watermark extracting based on tentative watermark Download PDFInfo
- Publication number
- CN103975605B CN103975605B CN201280059704.5A CN201280059704A CN103975605B CN 103975605 B CN103975605 B CN 103975605B CN 201280059704 A CN201280059704 A CN 201280059704A CN 103975605 B CN103975605 B CN 103975605B
- Authority
- CN
- China
- Prior art keywords
- watermark
- distortion
- predistortion
- content
- tentative
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 claims abstract description 77
- 238000000605 extraction Methods 0.000 claims abstract description 50
- 238000001514 detection method Methods 0.000 claims description 221
- 230000001360 synchronised effect Effects 0.000 claims description 146
- 230000011218 segmentation Effects 0.000 claims description 119
- 238000012360 testing method Methods 0.000 claims description 36
- 239000013589 supplement Substances 0.000 claims description 31
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 30
- 238000003780 insertion Methods 0.000 claims description 24
- 230000037431 insertion Effects 0.000 claims description 24
- 239000012634 fragment Substances 0.000 claims description 12
- 238000011156 evaluation Methods 0.000 claims description 11
- 238000007689 inspection Methods 0.000 claims description 8
- 230000005540 biological transmission Effects 0.000 claims description 2
- 238000004590 computer program Methods 0.000 abstract description 10
- 238000012545 processing Methods 0.000 description 30
- 238000013213 extrapolation Methods 0.000 description 25
- 238000003860 storage Methods 0.000 description 17
- 238000004891 communication Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 11
- 239000000284 extract Substances 0.000 description 11
- 238000005516 engineering process Methods 0.000 description 8
- 230000006835 compression Effects 0.000 description 7
- 238000007906 compression Methods 0.000 description 7
- 230000008569 process Effects 0.000 description 7
- 238000002474 experimental method Methods 0.000 description 6
- 238000005469 granulation Methods 0.000 description 6
- 230000003179 granulation Effects 0.000 description 6
- 238000012986 modification Methods 0.000 description 6
- 230000004048 modification Effects 0.000 description 6
- 241000196324 Embryophyta Species 0.000 description 5
- 230000006399 behavior Effects 0.000 description 5
- 230000008859 change Effects 0.000 description 5
- 238000006243 chemical reaction Methods 0.000 description 5
- 238000013461 design Methods 0.000 description 5
- 230000036961 partial effect Effects 0.000 description 5
- 230000009471 action Effects 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 4
- 101150080085 SEG1 gene Proteins 0.000 description 3
- 101100421134 Schizosaccharomyces pombe (strain 972 / ATCC 24843) sle1 gene Proteins 0.000 description 3
- 229910002056 binary alloy Inorganic materials 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000012937 correction Methods 0.000 description 3
- 230000008447 perception Effects 0.000 description 3
- 230000003252 repetitive effect Effects 0.000 description 3
- 238000012795 verification Methods 0.000 description 3
- 101100202858 Saccharomyces cerevisiae (strain ATCC 204508 / S288c) SEG2 gene Proteins 0.000 description 2
- 206010044565 Tremor Diseases 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000005520 cutting process Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000007639 printing Methods 0.000 description 2
- 230000009467 reduction Effects 0.000 description 2
- 230000002441 reversible effect Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 235000017060 Arachis glabrata Nutrition 0.000 description 1
- 241001553178 Arachis glabrata Species 0.000 description 1
- 235000010777 Arachis hypogaea Nutrition 0.000 description 1
- 235000018262 Arachis monticola Nutrition 0.000 description 1
- 206010011224 Cough Diseases 0.000 description 1
- XUIMIQQOPSSXEZ-UHFFFAOYSA-N Silicon Chemical compound [Si] XUIMIQQOPSSXEZ-UHFFFAOYSA-N 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 210000001367 artery Anatomy 0.000 description 1
- 238000003556 assay Methods 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 230000002457 bidirectional effect Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 238000012790 confirmation Methods 0.000 description 1
- 230000006837 decompression Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 230000036039 immunity Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000001727 in vivo Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000012804 iterative process Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 235000020232 peanut Nutrition 0.000 description 1
- 230000000149 penetrating effect Effects 0.000 description 1
- 238000004080 punching Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000002829 reductive effect Effects 0.000 description 1
- 230000003362 replicative effect Effects 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000009666 routine test Methods 0.000 description 1
- 238000009738 saturating Methods 0.000 description 1
- 229910052710 silicon Inorganic materials 0.000 description 1
- 239000010703 silicon Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
- 210000003462 vein Anatomy 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N21/00—Selective content distribution, e.g. interactive television or video on demand [VOD]
- H04N21/80—Generation or processing of content or additional data by content creator independently of the distribution process; Content per se
- H04N21/83—Generation or processing of protective or descriptive data associated with content; Content structuring
- H04N21/835—Generation of protective data, e.g. certificates
- H04N21/8358—Generation of protective data, e.g. certificates involving watermark
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T1/00—General purpose image data processing
- G06T1/0021—Image watermarking
- G06T1/005—Robust watermarking, e.g. average attack or collusion attack resistant
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L19/00—Speech or audio signals analysis-synthesis techniques for redundancy reduction, e.g. in vocoders; Coding or decoding of speech or audio signals, using source filter models or psychoacoustic analysis
- G10L19/018—Audio watermarking, i.e. embedding inaudible data in the audio signal
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/32—Circuits or arrangements for control or supervision between transmitter and receiver or between image input and image output device, e.g. between a still-image camera and its memory or between a still-image camera and a printer device
- H04N1/32101—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title
- H04N1/32144—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title embedded in the image data, i.e. enclosed or integrated in the image, e.g. watermark, super-imposed logo or stamp
- H04N1/32352—Controlling detectability or arrangements to facilitate detection or retrieval of the embedded information, e.g. using markers
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2201/00—General purpose image data processing
- G06T2201/005—Image watermarking
- G06T2201/0065—Extraction of an embedded watermark; Reliable detection
Landscapes
- Engineering & Computer Science (AREA)
- Signal Processing (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Theoretical Computer Science (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Human Computer Interaction (AREA)
- Acoustics & Sound (AREA)
- General Physics & Mathematics (AREA)
- Computer Security & Cryptography (AREA)
- Editing Of Facsimile Originals (AREA)
- Image Processing (AREA)
Abstract
Method, apparatus and computer program product promote the extraction of embedded watermark in the case where being distorted there are content.The sync section of embedded watermarked frame is detected using predistortion synchronization template.The predistortion synchronization template of the sync section of embedded watermarked frame described in best match generates the estimation to being present in one or more distortions in content.The rest part of watermarked frame can be assessed based on result of the comparison.
Description
Related application
Present application is advocated to come from No. 13/288,937, No. 13/288,945, No. 13/288,952 and the 13/th
The priority of 288, No. 958 U.S. patent application cases, the patent application case are filed an application on November 3rd, 2011.Before
The full content of mentioned patent application case is incorporated by reference a part for the disclosure for present application.
Technical field
Present application generally relates to field of content management.More particularly, disclosed embodiment is related to coming from media
The insertion and extraction of the watermark of content.
Background technology
This section is intended to provide the background or context of the disclosed embodiment stated in claims.Herein
Illustrate concept that is may include to carry out but being not necessarily the concept for previously having dreamed up or having carried out.Therefore, unless separately having herein
Instruction, otherwise the content described in this section in present application explanation and claims be not the prior art, and not
Think that the prior art can be become due to comprising this section.
Watermark be essentially be embedded into content host can not perceptual signal.Content host can be appointing in the following terms
One：Audio, static image, video can be stored on physical medium or be transmitted or broadcast to appointing for another point from a point
One other contents.Watermark is designed to carrying auxiliary information, and does not substantially influence the fidelity of content host or do not interfere master
The normal use of machine content.For this reason, covert communications are executed using watermark sometimes, is hidden wherein focusing on through hiding
In the presence of signal.In addition, other extensive uses of watermark comprising：Prevent the use of the unauthorized of copyrighted multimedia content
(for example, replicate, play and propagate), proof of ownership, verification, tampering detection, content integrity are examined, broadcast monitoring, transaction
Tracking, audience measurement, triggering secondary activity (such as interactive with software program or hardware component) are transmitted and are believed about the auxiliary of content
Breath (such as title text, full name and artist name or on how to buy the instruction of content) etc..Use above list is not
Plan tool is exhaustive, because many other current and future systems can benefit from the mainly total channel transmitting with auxiliary information.
Design watermaking system need to reach the transparency (not sentience) of embedded watermark, the robustness of embedded watermark (that is,
The intentionally and unintentionally ability of distorted signals is born in watermark) with the security requirement of system (that is, embedded watermark can escape detection, delete
The degree removed and/or manipulated by unauthorized party) between appropriate balance.It has to carry out this balance play while limitation is less than usual
For the processing behaviour for the specific grade that the actual software and/or hardware embodiments of water mark embedding device and/or watermark embedder are forced
The average number and/or maximum number that work (that is, processing load) and memory use.
Invention content
This section is intended the summary provided to particular exemplary embodiment and is not intended to limit disclosed in present application
The range of embodiment.
The one side of disclosed embodiment is related to a kind of method, it includes：Reception includes sync section and packet portion
Watermarked frame, the watermarked frame are obtained from the content for being embedded with watermark；By the received sync section and one or more pre- mistakes
True synchronization template is compared；And received watermarked frame described in an at least outcome evaluation based on the comparison.
In an example embodiments, generates one or more described predistortions at least partially through following operation and synchronize
Template：Generate multiple navigation contents；The multiple navigation content is set to be embedded with the watermark including sync section；By one or more
Distortion makes the multiple embedded navigation content distortion；And it obtains candidate predistortion from the multiple distortion navigation content and synchronizes mould
Plate.In a kind of exemplary version, generates one or more described synchronization templates and be further contained in by emitting media hair
The multiple embedded navigation content that a subgroup is received after penetrating, wherein one or more distortions are introduced into institute by the transmitting media
In the multiple embedded navigation content for stating subgroup.In addition, one or more synchronization templates described in generating also include：It is connect described
Each of the sync section of the multiple navigation content of receipts subgroup and the candidate predistortion synchronization template are compared,
And the sync section matched one of the selection in predefined tolerance with the multiple navigation content of the received subgroup
Or multiple candidate predistortion synchronization templates.
It is described relatively to generate each of one or more described predistortion synchronization templates in an example embodiments
Error count, the error count indicates the received sync section of the watermarked frame and one or more described predistortions
The number of mismatch symbol between each of synchronization template.It is described relatively to generate institute in another example embodiments
The correlation of each of one or more predistortion synchronization templates is stated, the correlation indicates the synchronization of the watermarked frame
Part to what extent matches each of one or more described predistortion synchronization templates.
According to another embodiment, method referred to above further includes：After the comparison, best match is identified
The predistortion synchronization template of the received sync section of the watermarked frame, and selection mould synchronous with the identified predistortion
One or more associated type of distortion of plate and amount distortion indicate the distortion being present in the content.Implement at one
In example, method referred to above further comprises distortion and/or distortion combination for each type with order group of presorting
Knit the candidate predistortion synchronization template.In this embodiment, the order of presorting makes with expression real content distortion
The first candidate predistortion synchronization template ranking compared with high likelihood is higher than with the property of may be less likely to for indicating real content distortion
Second candidate predistortion synchronization template.
In another embodiment, one or more described predistortion synchronization templates correspond to the specific search in distorted space
Granularity.In a kind of version of above example, select the number of one or more predistortion synchronization templates to reduce
One or both of the following terms：Computing resource needed for false watermark detection probability and/or assessment the received watermarked frame.
According to another embodiment, the assessment generates the existing instruction to one of the following terms：Tentative watermark,
It indicates the candidate watermark for having more than the associated false watermark detection probability for wanting false watermark detection probability；Or conclusive water
Print indicates there is the watermark for being less than or equal to the associated false watermark detection probability for wanting false watermark detection probability after testing.
In another embodiment, the assessment identification is present in one or more type of distortion and amount distortion in the content.
In another embodiment, the content includes at least first and second watermark message, and first watermark message
With the payload less than second watermark message.In this embodiment, assessing the received watermarked frame causes institute
It states the first watermark message and is detected as tentative watermark, and carried using the content distortion estimation obtained for the tentative watermark
Take second watermark message.
The another aspect of disclosed embodiment is related to a kind of device comprising receiver, the receiver are configured to connect
Packet receiving includes the watermarked frame of sync section and packet portion, and the watermarked frame is obtained from the content for being embedded with watermark.Described device is also
Including：Comparator is configured to the received sync section being compared with one or more predistortion synchronization templates；And
Evaluator is configured to received watermarked frame described in an at least outcome evaluation for device based on the comparison.
Another embodiment is related to a kind of device, it includes：Water mark embedding device is configured so that multiple navigation contents are embedded in
There is the watermark including sync section；Distortion processing component is configured to make the multiple insertion lead by one or more distortions
Frequency content is distorted；And watermark embedder, it is configured to generate candidate predistortion synchronization mould from the multiple distortion navigation content
Plate.In one embodiment, this device further comprises：Receiver is configured to by emitting being followed by for media transmission
The multiple embedded navigation content of a subgroup is received, wherein one or more distortions are introduced into the subgroup by the transmitting media
In the multiple embedded navigation content；And template matches component, it is configured to lead described the multiple of received subgroup
Each of the sync section of frequency content and the candidate predistortion synchronization template are compared, and are selected in predefined tolerance
It is interior same with one or more matched candidate predistortions of the sync section of the multiple navigation content of the received subgroup
Walk template.
The another aspect of disclosed embodiment is related to a kind of device, it includes processor and memory, the memory packet
Include processor executable code.The processor executable code configures described device when being executed by the processor to receive
Watermarked frame including sync section and packet portion, the watermarked frame are obtained from the content for being embedded with watermark；It is received described
Sync section is compared with one or more predistortion synchronization templates；And institute described in an at least outcome evaluation based on the comparison
Receive watermarked frame.
The another aspect of disclosed embodiment is related to a kind of computer being embodied in non-transitory computer-readable media
Program product comprising for receiving the program code of the watermarked frame for including sync section and packet portion, the watermarked frame from
The content for being embedded with watermark obtains.The computer program product also include for by the received sync section with it is one or more
Program code that a predistortion synchronization template is compared and for being connect described in an at least outcome evaluation based on the comparison
Receive the program code of watermarked frame.
The another aspect of disclosed embodiment is related to a kind of method, and it includes multiple tentative from embedded content host extraction
Watermark, wherein it is more than the associated false watermark detection probability for wanting false watermark detection probability that each tentative watermark, which indicates to have,
Candidate watermark.The method also includes：Using in the extracted tentative watermark both at least obtain be present in it is described embedding
Enter the associated estimated distortion information of one or more distortions in content host；And it based on the estimated distortion information, obtains
Obtain one or more predistortion watermark templates.The method is further included using one or more described predistortion watermark templates again
Assess at least one of described extracted tentative watermark.
In one embodiment, described reappraise causes to detect and be associated in the extracted tentative watermark
The false watermark detection probability of each compares the watermark with improved false watermark detection probability.In another embodiment, described
Reappraising leads to detect the associated false watermark detection probability with less than or equal to the wanted false watermark detection probability
Conclusive watermark.In another embodiment, the expression of each of described predistortion watermark template can pollute described embedded main
The specific type and specific quantity of the distortion of at least one type of machine content.According to another embodiment, described reappraise causes
To being present in the improved estimation of the distortion in the embedded content host.
In one embodiment, method referred to above, which is further contained in, obtains one or more predistortion watermark templates
Before, determine whether the estimated distortion information corresponds to over the amount distortion of certain distortion threshold value.In this embodiment, only
The estimated distortion information correspond to no more than the certain distortion threshold value amount distortion in the case of obtain described one or
Multiple predistortion watermark templates.
In another embodiment, using two or more synchronization portions of described at least two extracted tentative watermarks
Separately win to obtain the estimated distortion information.In this embodiment, by by described in the extracted tentative watermark at least
Spacing between the two is compared with the nominal spacing for being associated with the watermark for not being distorted content and obtains the estimated distortion
Information.In another embodiment, the extraction of each of described tentative watermark includes detecting the tentative watermark
The triggering tentative watermark packet portion extraction sync section.
The another aspect of disclosed embodiment is related to a kind of device, it includes extractor, the extractor be configured to from
Embedded content host extracts multiple tentative watermarks, wherein it is more than to want false watermark detection general that each tentative watermark, which indicates to have,
The candidate watermark of the associated false watermark detection probability of rate.This device includes also distortion estimator, and the distortion estimator is through matching
Set with use in the extracted tentative watermark both at least obtain with one be present in the embedded content host or
It is multiple to be distorted associated estimated distortion information and obtain one or more predistortion watermark templates.This device, which further includes, to be commented
Estimate device, the evaluator is configured to reappraise using one or more described predistortion watermark templates described extracted tentative
At least one of watermark.
The another aspect of disclosed embodiment is related to a kind of device comprising processor and memory, the memory packet
Containing processor executable code.The processor executable code is configured when being executed by the processor states device with from insertion
Content host extracts multiple tentative watermarks, wherein it is more than to want false watermark detection probability that each tentative watermark, which indicates to have,
The candidate watermark of associated vacation watermark detection probability.The processor executable code is also configured when being executed by the processor
Described device is to use at least the two in the extracted tentative watermark to obtain and be present in the embedded content host
The associated estimated distortion information of one or more distortions；And based on the estimated distortion information, it is pre- to obtain one or more
It is distorted watermark template.The processor executable code further configures described device when being executed by the processor to use
One or more described predistortion watermark templates reappraise at least one of described extracted tentative watermark.
The another aspect of disclosed embodiment is related to a kind of computer being embodied in non-transitory computer-readable media
Program product, it includes the program codes for extracting multiple tentative watermarks from embedded content host, wherein each tentative
Watermark indicates the candidate watermark for having more than the associated false watermark detection probability for wanting false watermark detection probability.The computer
Program product also includes described embedded main for being obtained and being present in using at least the two in the extracted tentative watermark
One or more in machine content are distorted the program code of associated estimated distortion information, and are used for based on the estimated mistake
The program code of one or more predistortion watermark templates of true information acquisition.The computer program product is further included for making
The program generation of at least one of described extracted tentative watermark is reappraised with one or more described predistortion watermark templates
Code.
The another aspect of disclosed embodiment is related to a kind of method, and it includes extract tentative water from embedded content host
Print, wherein the tentative watermark indicates the time for having more than the associated false watermark detection probability for wanting false watermark detection probability
Select watermark.The method also includes：Instruction is obtained using the extracted tentative watermark to be present in the embedded content host
Certain distortion or distortion combination coarse value estimated distortion information；And selection is corresponding to one near the coarse value
Or one or more supplement predistortion watermark templates of multiple distortion values.The method is further included using described selected one or more
A supplement predistortion watermark template reappraises the extracted tentative watermark.
In one embodiment, obtaining the estimated distortion information includes：Correspond to specific mistake with coarse-grained selection
Multiple rough predistortion watermark templates of the range of true or distortion combination distortion value；By the extracted tentative watermark and institute
Each of multiple rough predistortion watermark templates are stated to be compared；And it is extracted tentative described in best match by identifying
First rough predistortion watermark template of watermark and obtain the estimated distortion information.In a kind of version of this embodiment
In, the multiple rough predistortion watermark template is randomly choosed in the larger set of predistortion watermark template.Implement herein
In another version of example, selection corresponds to one or more types for being likely to be present in the distortion in the embedded content host
The multiple rough predistortion watermark template.
According to another embodiment, one or more described supplement predistortion watermark templates of selection are searched for realizing with fine granulation
Distorted space near the coarse value.In another embodiment, described reappraise causes to being present in the embedded master
The improved estimation of distortion in machine content.In a kind of version, using the improved estimation come main from the insertion
The additional watermark of machine contents extraction.In one embodiment, described reappraise leads to detect and be associated with the extracted examination
The false watermark detection probability of the property tested watermark compares the watermark with improved false watermark detection probability.In another embodiment, institute
Stating to reappraise causes to detect with less than or equal to the associated false watermark detection probability for wanting false watermark detection probability
Conclusive watermark.
In another embodiment, it is described reappraise including：It will one or more described supplement predistortion watermark templates and institute
Extracted tentative watermark is stated to be compared；Generation and each more associated error count, wherein the error count table
Show that the corresponding supplement predistortion watermark template to what extent matches the extracted tentative watermark；And selection corresponds to
The supplement predistortion watermark template of minimum produced error count.In another embodiment, with repetitive mode one time one
One or more described supplement predistortion watermark templates are selected aly.
The another aspect of disclosed embodiment is related to a kind of device, it includes extractor, the extractor be configured to from
Embedded content host extracts tentative watermark, wherein it is more than to want false watermark detection probability that the tentative watermark, which indicates to have,
The candidate watermark of associated vacation watermark detection probability.This device further includes distortion estimator, and the distortion estimator is through matching
It sets to use the extracted tentative watermark to obtain certain distortion or distortion that instruction is present in the embedded content host
The estimated distortion information of the coarse value of combination, and select to correspond to one of one or more distortion values near the coarse value
Or multiple supplement predistortion watermark templates.This device also comprises evaluator, and the evaluator is configured to using described selected
It selects one or more supplement predistortion watermark templates and reappraises the extracted tentative watermark.
According to the another aspect of disclosed embodiment, a kind of device is provided, it includes processor and memory, the storage
Device includes processor executable code.The processor executable code configured when being executed by the processor described device with
Tentative watermark is extracted from embedded content host, wherein it is more than to want false watermark detection probability that the tentative watermark, which indicates to have,
Associated false watermark detection probability candidate watermark.The processor executable code is also matched when being executed by the processor
Described device is set to use the extracted tentative watermark to obtain the specific mistake that instruction is present in the embedded content host
The estimated distortion information of true or distortion combination coarse value, and select to correspond to one or more mistakes near the coarse value
One or more supplement predistortion watermark templates of true value.The processor executable code by the processor when being executed into one
Step configuration described device is to use one or more described selected supplement predistortion watermark templates to reappraise the extracted examination
The property tested watermark.
The another aspect of disclosed embodiment is related to a kind of computer being embodied in non-transitory computer-readable media
Program product, it includes the program codes for extracting tentative watermark from embedded content host, wherein the tentative watermark
Indicate the candidate watermark for having more than the associated false watermark detection probability for wanting false watermark detection probability.The computer program
Product also includes specific in the embedded content host for being present in using the extracted tentative watermark acquisition instruction
The program code of the estimated distortion information of the coarse value of distortion or distortion combination, and for selecting to correspond in the coarse value
The program code of one or more supplement predistortion watermark templates of one or more neighbouring distortion values.The computer program product
It further includes for reappraising the extracted experiment using one or more described selected supplement predistortion watermark templates
The program code of property watermark.
The another aspect of disclosed embodiment is related to a kind of method, and it includes extract tentative water from embedded content host
Print, wherein the tentative watermark indicates the time for having more than the associated false watermark detection probability for wanting false watermark detection probability
Select watermark.The method further includes be positioned in the embedded content host relative to described extracted tentative by obtaining
The symbol of potential watermarked frame at the predefined position of watermark and form one or more extrapolated watermarks, and determine with the warp
Detect whether extrapolated watermark when tentative watermark is evaluated jointly meets wanted false watermark detection probability.In one embodiment
In, the symbol location of the potential watermarked frame in the embedded content host relative to the extracted tentative water
At the time predefined of print and/or spatial position.
In another embodiment, determine whether the extrapolated watermark meets wanted false watermark detection probability and include：It determines
The number of mismark in one or more of described extrapolated watermark is based at least partially on the described through true of mismark
The one or more of the fixed number mesh into the extrapolated watermark assigns weight, and determine with the watermark tentative after testing
Whether the weighted extrapolated watermark meets the wanted false watermark detection probability when combination.In this embodiment, can pass through
One or more of the following terms determines the number of the mismark in each extrapolated watermark：The extrapolated watermark symbol
Comparison number with the symbol of the extracted tentative watermark executes the error-correcting code decoding of the extrapolated watermark symbol,
And/or the comparison of the extrapolated watermark and one or more predistortion watermark templates.
In one embodiment, forming each of described extrapolated watermark includes：It obtains in one or more segmentations
The number of mismark is based at least partially on the number of the mismark in each of one or more described segmentations
Assign weight, combination associated described assigned with one or more described segmentations to each of one or more described segmentations
Weight is to generate one or more weighted extrapolated watermark sections, and determination is combining when institute with the watermark tentative after testing
State whether weighted extrapolated watermark section meets the wanted false watermark detection probability.In this embodiment, described one is obtained
Or the number of the mismark in each of multiple segmentations includes：It will be in each of one or more described segmentations
The symbol of the potential watermarked frame be compared with one or more predistortion watermark templates；And generate it is described one or more
The counting of each of segmentation, it is described to count the symbol for indicating the potential watermarked frame in one or more described segmentations
The number of mismatch symbol between one or more described predistortion watermark templates.In addition, in a kind of variation of this embodiment
In form, the number of the segmentation in the extrapolated watermark is determined based on one or more of the following terms：The extracted examination
The range of the property tested watermark, the amount distortion being present in the embedded content host and/or it is present in the embedded content host
Type of distortion.In addition, in a kind of version of this embodiment, it is based at least partially in length and is equal to particular fragments
Non-watermarked segmentation expection weight assignment described in particular fragments the weight.According to another embodiment, at least partly
The range of each segmentation is determined based on one or more of the following terms：It is expected that being present in the distortion in the embedded content host
Amount and the expected type of distortion being present in the embedded content host.In another embodiment, for jointly across than described
One or more described segmentations of the small range of tentative watermark are executed about combining when institute with the watermark tentative after testing
State the determination whether weighted extrapolated watermark section meets the wanted false watermark detection probability.
According to another embodiment, the false watermark detection probability corresponds to the detection of conclusive watermark.In another reality
It applies in example, method referred to above further includes distortion or the distortion group reported to being present in the embedded content host
The estimation of the amount and type of conjunction.
The another aspect of disclosed embodiment is related to a kind of device, it includes extractor, the extractor be configured to from
Embedded content host extracts tentative watermark, wherein it is more than to want false watermark detection probability that the tentative watermark, which indicates to have,
The candidate watermark of associated vacation watermark detection probability.Described device also includes：Watermark extrapolator is configured to be determined
Potential watermarked frame in the embedded content host at the predefined position relative to the extracted tentative watermark
Symbol and form one or more extrapolated watermarks；And evaluator, be configured to determine with the water tentative after testing
Whether the extrapolated watermark meets wanted false watermark detection probability when printing common evaluation.
The another aspect of disclosed embodiment is related to a kind of device, it includes processor and memory, the memory packet
Containing processor executable code.The processor executable code configures described device when being executed by the processor with from embedding
Enter the tentative watermark of host contents extraction, wherein it is more than the phase for wanting false watermark detection probability that the tentative watermark, which indicates to have,
The candidate watermark of the false watermark detection probability of association.The processor executable code also configures institute when being executed by the processor
Device is stated to be positioned in the embedded content host relative to the predefined of the extracted tentative watermark by obtaining
The symbol of potential watermarked frame at position and form one or more extrapolated watermarks, and determine with the water tentative after testing
Whether the extrapolated watermark meets wanted false watermark detection probability when printing common evaluation.
The another aspect of disclosed embodiment is related to a kind of computer being embodied in non-transitory computer-readable media
Program product, it includes the program codes for extracting tentative watermark from embedded content host, wherein the tentative watermark
Indicate the candidate watermark for having more than the associated false watermark detection probability for wanting false watermark detection probability.The computer program
Product is also comprised mean for obtaining and is positioned in the embedded content host relative to the extracted tentative watermark
The symbol of potential watermarked frame at predefined position and the program code for forming one or more extrapolated watermarks, and exist for determining
Whether the extrapolated watermark meets the journey of wanted false watermark detection probability when being evaluated jointly with the watermark tentative after testing
Sequence code.
The another aspect of disclosed embodiment is related to a kind of method comprising：From the content for being embedded with one or more watermarks
Potential watermarked frame is extracted, wherein each embedded watermark includes forming multiple symbols of watermarked frame；The watermarked frame is divided into more
A segmentation, wherein each of the multiple segmentation includes the two in more watermark symbols；Weight is assigned to each segmentation；
The combination assigned weight associated with two or more segmentations is weighted partially or completely to generate one or more
Watermarked frame；And determine whether one or more described weighted complete or partial watermarked frames meet wanted false watermark detection probability.
In one embodiment, the range of each segmentation is selected according to the type for the mistake being present in the content.
In another embodiment, it is based at least partially on the expection weight assignment institute of the non-watermarked segmentation in length equal to particular fragments
State the weight of particular fragments.
The another aspect of disclosed embodiment is related to a kind of device comprising extractor, the extractor be configured to from
It is embedded with the potential watermarked frame of contents extraction of one or more watermarks, wherein each embedded watermark includes forming multiple symbols of watermarked frame
Number.Described device also includes evaluator, and the evaluator is configured to the watermarked frame being divided into multiple segmentations, wherein described
Each of multiple segmentations include the two in more watermark symbols；Weight is assigned to each segmentation；Combination and two or two
A above associated assigned weight of segmentation is to generate one or more weighted partially or completely watermarked frames；And determine institute
State whether one or more weighted complete or partial watermarked frames meet wanted false watermark detection probability.
The another aspect of disclosed embodiment is related to a kind of device, it includes processor and memory, the memory packet
Containing processor executable code.The processor executable code configures described device when being executed by the processor with from embedding
Enter to have the potential watermarked frame of the contents extraction of one or more watermarks, wherein each embedded watermark includes forming multiple symbols of watermarked frame
Number；The watermarked frame is divided into multiple segmentations, wherein each of the multiple segmentation includes in more watermark symbols
The two；Weight is assigned to each segmentation；The combination assigned weight associated with two or more segmentations is to generate
One or more weighted partially or completely watermarked frames；And determine whether one or more described weighted complete or partial watermarked frames are full
Foot wants false watermark detection probability.
The another aspect of disclosed embodiment is related to a kind of computer being embodied in non-transitory computer-readable media
Program product, it includes：For the program code from the potential watermarked frame of contents extraction for being embedded with one or more watermarks, wherein often
One insertion watermark includes forming multiple symbols of watermarked frame；Program code for the watermarked frame to be divided into multiple segmentations,
Each of wherein the multiple segmentation includes the two in more watermark symbols；Journey for assigning from weight to each segmentation
Sequence code；It is weighted to generate one or more for combining the assigned weight associated with two or more segmentations
The partially or completely program code of watermarked frame；And for determining whether one or more described weighted complete or partial watermarked frames are full
Foot wants the program code of false watermark detection probability.
Description of the drawings
Fig. 1 is the block diagram for the watermaking system for being suitable for disclosed embodiment.
Fig. 2 is the block diagram for the watermark embedder device for being suitable for disclosed embodiment.
Fig. 3 A displayings have the exemplary watermarked frame of synchronous (SYNCH) and the parts PACKET.
Fig. 3 B are illustrated in the success of the parts SYNCH of watermarked frame and the exemplary PACKET after failed compilation converges
It compiles and attempts.
Fig. 3 C illustrate the PACKET compilation operations using predistortion SYNCH templates according to one exemplary embodiment.
The predistortion SYNCH templates of Fig. 4 displayings according to one exemplary embodiment.
Fig. 5 illustrates the one group of operation executed according to one exemplary embodiment.
Fig. 6 is the specific group of some or all of operations in the operation that can be used for executing Fig. 5 according to one exemplary embodiment
The block diagram of part.
Fig. 7 illustrate according to one exemplary embodiment can be executed with generate one group of predistortion watermark template one group of behaviour
Make.
Fig. 8 is the block diagram of the specific components of some or all of operations in the operation for can be used for executing Fig. 7.
Fig. 9 illustrates the one group of operation that can be executed the extraction to promote embedded watermark according to one exemplary embodiment.
Figure 10 illustrates can be executed to be embedded in one group of watermark from distortion contents extraction according to one exemplary embodiment
Operation.
Figure 11 A displayings have the exemplary watermarked frame of synchronous (SYNCH) and the parts PACKET.
The success and the exemplary PACKET after failed compilation that Figure 11 B are illustrated in the parts SYNCH of watermarked frame
Compilation is attempted.
Figure 11 C illustrate the PACKET compilation operations after detecting tentative watermark according to one exemplary embodiment.
Figure 11 D illustrate being segmented after detecting tentative watermark and using watermark according to one exemplary embodiment
PACKET compilation operations.
Figure 12 illustrate according to one exemplary embodiment by separate watermark search with sparse granular search and fine granulation
Exemplary plot of the false watermark detection probability that search is compared to watermark weight threshold.
Figure 13 illustrates can be executed to extrapolate one group of watermark on the basis of by segmentation according to one exemplary embodiment
Operation 1300.
Figure 14 is some or all of behaviour in the operation that can be used for executing Figure 13 and/or Figure 18 according to one exemplary embodiment
The block diagram of the specific components of work.
The curve graph of the expection weight of the non-watermarked segmentation of Figure 15 displayings according to one exemplary embodiment.
Figure 16 illustrates the simplification figure for the device that can implement various disclosed embodiments in it.
Figure 17 illustrates SYNCH compilations and predistortion SYNCH template operations according to one exemplary embodiment.
Figure 18 illustrates the one group of operation that can be executed with watermark of extrapolating according to one exemplary embodiment.
Specific implementation mode
In the following description, following detailed description, state details and explanation in order to provide to disclosed reality
Apply the thorough understanding of example.However, it will be apparent to those skilled in the art that, it can be in other implementations away from these details and explanation
The present invention is put into practice in example.
In addition, in this description, meaning to serve as example, example or diagram illustrating using word " demonstration ".
Any embodiment or design here depicted as " demonstration " are not necessarily to be construed as preferably or more advantageous than other embodiments or design.
But, it is intended to concept is presented in specific ways using word exemplary.
It can operate with various watermark embedding techniques by following and embed watermark data into content host：For example, exist
In time domain or frequency domain manipulate host signal least significant bit, using spread-spectrum, phase, amplitude or frequency modulating technology by
Separate carrier signals are inserted into watermark, and are inserted into water using the interdependent carrier signal of host (such as feature modulation and informed embedded technology)
Print.Most embedded technologies using content host psycho-visual or psychologic acoustics (or both) analysis is inserted into watermark to determine
Optimal location and amplitude.This analyzes the degree that the generally recognized host signal can be hidden or shelter the insertion watermark perceived such as people.
Emit usually using various storages and/or the embedded content host of launch channel storage and/or by embedded content host
To another location.These channels by intrinsic noise and distortion characterize, such as due to pollution optical media on data cut and
Mistake caused by fingerprint, the noise in the radio broadcasting of audio-visual content, via internet or from the more of media server
Packet discarding in the flowing of media content etc..The additional damage that can influence the fidelity of embedded watermark may be due to usually to more
The various signal processing operations (such as lossy compression, scaling, rotation, analog/digital conversion etc.) that media content executes cause.More
In number digital watermarking application, embedded watermark allows for maintaining its integrality under this noise like and distortion condition.
The safety of embedded watermark is the another aspect of watermaking system.In specific application, (such as proof of ownership, source are tested
Card, tracing pirate, the access control to copyrighted content) in, embedded watermark, which must resist, to be intended to the presence of detection watermark, translates
Solution is by the data of those watermark carryings, modification or is inserted into illegal value (forgery) and/or removes the intentional manipulation of embedded watermark.
It is watermark payload capacity to design the another consideration in watermaking system.This is required depending on the specific of watermaking system
Using.Typical case is between from needing to detect the presence (that is, it needs to single state watermark) of only watermark to tens per second of needs
In the range of auxiliary information.In latter, embedded position can be used to come carrying identification and timing information, for example, sequence number and
Timestamp, metadata (metadata), such as title, artist name, purchase information etc..
Another factor in design actual watermark system is embedded and/or extraction unit calculating cost.This factor is to must
The consumer electronics implemented or software must be required to utilize by limited silicon area or calculating becomes more and more important.This factor
It can be closely related with the given application of watermaking system.For example, the watermark for being used for court's tracking of pirate channel (such as will not
Watermark those of is embedded in each copy of content spread via internet with code) it needs to be simply embedded into device and tolerable is multiple
Miscellaneous and high-cost court extractor.On the other hand, for example, be designed to prevent more matchmakers in consumer electronics
The copy control system tolerable exquisiteness embedded device but the simple and highly efficient extractor of needs of the access of the unauthorized held in vivo.
The another key factor designed in watermaking system is false watermark detection probability.False watermark can be from unlabelled content
It generates, or may cause due to detecting the watermark value for the watermark value for being different from initially being embedded in when detecting watermark.Want false
Watermark detection grade may also depend upon the set application of watermaking system and change.For example, in copy control application, vacation detection
Probability must be very low (for example, about 10-12), because executing limitation row due to the false watermark detection of the content to legal purchase
Dynamic (for example, stopping the playback of content) certainly will make user dejected and have passive shadow to device manufacturing person and/or content provider
It rings.On the other hand, the number that the following song has been broadcasted is tracked for (for example) to generate the broadcast of royalty or pop roll
Supervision application, the relatively high false verification and measurement ratio of tolerable is (for example, about 10-6), because the presence of several false detections can be to counting most
Result has little effect on afterwards.
Another key factor for influencing the overall performance of watermaking system is the particular technology for watermark insertion and extraction
Selection.In view of specific application on hand, it is non-appointing often with challenge that optimal compromise is made between requirement referred to above
Business.
Disclosed embodiment significantly improves the insertion and extraction of watermark, while balancing various system requirements, including watermark is saturating
Lightness, robustness, false positive detection, safety, payload capacity and processing load.In some embodiments, watermark extracting
Based on tentative watermark detection as the intermediate steps towards conclusive watermark extracting.
Following conversion can be used to promote the understanding with the relevant concept of disclosed embodiment.However, it should be understood that can be used
Other labels and the watermaking system for indicating the embodiment to indicate to be suitable for present application.Watermaking system can by tuple O, W, K,
EK, DK, CTDescription, wherein O is one group of whole origin host content (that is, in one or more hosts without embedded watermark
Hold), W is one group of whole watermark, and K is one group of whole stego-key, EKFor embedded operation, DKFor extraction operation and CTTo compare fortune
It calculates.Embedded operation EKAnd extraction operation DKIt can be indicated respectively by equation (1) and (2).
EK：O×W×K→Ow (1)。
DK：Ow×K→W′ (2)。
In equation (1) and (2), OwIt is from described group of insertion content O for described group of insertion content and W 'wThe institute of extraction
State a group candidate watermark.The comparison operation C indicated by equation (3)TExtracted watermark W ' and described group of whole watermark W are compared
Compared with.
CT：<W ', W>→ { 0,1 } (3).
It should be noted that in some embodiments, whole watermarks can not be directed to but execute institute for a subgroup of whole watermarks
It states and compares.For example, this subgroup may include thering is the watermark of specific range of values, the watermark for being embedded in certain types of content
Deng.After the successful match between obtaining W ' and W, the comparison operation generates the first value (for example, " 1 ") at once, and not at
After work(matching, comparison operation generates second value (for example, " 0 ") at once.
The simplified block diagram of the watermaking system 100 of Fig. 1 displayings according to one exemplary embodiment.Origin host content 102, coAnd water
It is the input to water mark embedding device 104 to print 110W.Watermark 110 is by using embedded stego-key 106, K and origin host content 102
Watermarking arrangements module 108 generate.Produced watermark 110 is embedded into content host 102 to generate embedded content host 112,
cw.Pass through embedded operation EKThe generation of insertion content host 112 can be indicated by following equation (4).
cW=Ek(co, W) and (4).
Embedded content host 112 emits/stores the storage of media 114 and/or transmitting using one or more.Transmitting/storage matchmaker
Body 114 can be one-way or bi-directional transmitting media.In the embodiment using two-way launch channel, the information from watermark embedder
The reflowable embedded components to watermaking system 100.This feeds back the hair that may include, but are not limited to identification, verification and/or authorization message
It penetrates, the instruction of the success or failure about watermark extracting operation operates the report etc. of relevant various statistics with extractor.
Referring back to Fig. 1, by transmitting ,/content that storage media 114 emit may include due to inherently emitting/stores media
114 cause and/or the distortion of result for the intentional attack to embedded content host 112.This modified embedded content host
116, c 'wFor to the input of watermark embedder 118, for watermark embedder 118 using extraction stego-key 120, K and (possibly) are former
Beginning content host 102 generates one or more candidate watermarks 122, W '.By extracting operation DKThe production of the candidate watermark 122 of execution
Life can be indicated by following equation (5).
W '=DK(c′w, co) (5)。
Note that many watermark detectors do not need origin host content 102 to extract embedded watermark, and therefore, at those
In situation (commonly referred to as blind Detecting), in equation (5), operation D is extractedKTo the dependence of origin host content 102 can under
Drop.It is further noted that in some embodiments, embedded stego-key 106 is identical as extraction stego-key 120.However, at it
In its embodiment, embedded stego-key 106 is different from extraction stego-key 120.
In general candidate watermark 122 is different from needle due to possible content manipulation or due to imperfect embedded program step
Watermark 110 to embedded selection.By comparator 124 to candidate watermark 122 and one or more possible watermarks 126, WiIt is compared.
The comparator generates the matched watermark for indicating whether to generate between one of candidate watermark 122 and possible watermark 126 and examines
Survey device output 128.For example, watermark detector output 128 can be generated according to equation (3).Watermark detector output 128 can
Further include the instruction of the value about watermark after testing, state and/or pattern.It should be noted that the comparator 124 of Fig. 1 only illustrates
Illustrate for evaluating the candidate watermark W ' whether really upper exemplary mechanisms for indicating the true watermark in content.For evaluating watermark
In the presence of and/or value other exemplary techniques including but not limited to using error detection and/or error correction code.For example, can incite somebody to action
Candidate watermark W ' decodings through error-correcting code (ECC) coding are to determine whether to restore faultless watermark packet and/or acquisition
Pollute the number of the mistake of specific watermark packet.
In view of wanted false watermark extracting probability PfGenerate watermark detector output 128.False watermark detection can two kinds of differences
Mode occurs.First, for any watermark extracting system, be present in non-watermarked content mistakenly detect the small of watermark but
Limited probability (being also known as false positive detection sometimes).For example, Secure Digital Music Initiative proposal (SDMI) and DVD audio requests
Regulation is less than 10-12The false positive watermark detection probability of/15 seconds assay intervals.This is it is meant that flat between false positive watermark detection
The equal time is necessary for continuously monitoring 4,760,000 years non-watermarked content.The false detection of Second Type is checking the content for having watermark
And occur when detecting the watermark value different from initially embedded watermark value.The false detection (commonly referred to as error detection) of this type can
Generate undesirable result.For example, error detection, which can potentially result in, detects " not allowing to play back " watermark rather than " does not allow
Copy " watermark, and therefore, the playback of the content of legal purchase is caused to stop.For the watermaking system appropriately designed, false detection rate
The order of magnitude identical with false positive detection rate as described above should be about.
In some systems, the output of the comparison operation indicated by equation (3) is determined using threshold value T.That is, comparator
Output can be indicated by following equation (6).
In equation (6), m is comparison result (and/or correlation) so that as m increases false watermark detection probability PfIt is single
Ground is adjusted to reduce.
Comparator operation and the calculating of false watermark detection probability depend on selected watermarking algorithm.For example, many bases
Use the calculation of correlation between W ' and W as comparison measuring m in the watermarking algorithm of spread spectrum.In particular, estimate candidate water
Related coefficient m between print W ' and the pseudorandom pattern W used in embedded watermark.Next, if estimated related coefficient m
More than specific threshold T, then reporting successfully watermark extracting.In the situation of unmarked content, extracted false watermark style with
The related coefficient of operation roughlys abide by normal distribution between long pseudorandom pattern, this means that well-known error letter can be used
Number operation vacation watermark detection probability Pf.As related coefficient m increases, false watermark detection probability PfMonotonously reduce.Therefore, threshold value T
With PfWith one-to-one mapping.
Alternatively, can also by candidate watermark extracting be bit string, and then can be used error detection or with potential watermark
Position pattern matching direct pattern operation bit-errors number e.Then, bi-distribution operation vacation watermark detection probability can be used
Pf.For example, if watermark is less than or equal to e for the number of N bit lengths and bit-errors, equation (7) can be used to calculate
False detection probability.
If watermark is formed by N=64 and detector only finds e=5 bit-errors, false watermark detection probability PfIt is small
In 10-12.This PfCondition is typically enough to carry out conclusive watermark extracting and (for example) copyright implementation completely.According to equation
(7) the false watermark detection probability P f obtained is as the number m (that is, N-e) of matched position increases and monotonously reduces.
Comparison operation as described above only generates two results.In the first result, extracts and report watermark (for example)
To implement copyright rule.This result sometimes referred to as detects strong watermark.In the second result, watermark is not found and is not started other
Action.In addition to previously mentioned conclusive watermark detection and no-watermark detect, the specified inspection of comparison program step can be also established
Measure the third result of " tentative " watermark.For example, third result may specify that wherein comparing generation is directed to strong watermark detection
Value more than or equal to threshold value T but less than another threshold value t is (for example, have the extraction situation of the number e) of vicious position.Compare fortune
Three outputs calculated can be indicated by following equation (8).
Consider that wherein watermark is made of bit string and comparator output counts the situation being made of the number of bit-errors again.
In above example, up to 5 mistakes are found in 64 bit length strings can be sufficient for conclusive watermark extracting.However, if detection
Device finds 20 bit-errors, then false watermark detection probability is less than 0.2%, this makes watermark exist may but to be not enough to determine and prove it
Use outside extractor.Therefore, any error count more than 5 (that is, T=6) but less than 21 (that is, t=21) can be used
By tentative watermark detection sender.If error count is greater than or equal to 21 mistakes, false watermark detection probability is more than
0.2%, and therefore, watermark can be ignored.
Fig. 2 displayings are suitable for the simplified block diagram of the watermark embedder device 200 of disclosed embodiment.Watermark embedder device
200 can be self-contained unit or may be incorporated into as larger content spaces disposal plant (such as content playback apparatus, content relay device, content
Recording device, set-top box etc.) component.It is received at receiver 204 and is passed to modified embedded content 202.It will be received in Fig. 2
Device 204 is portrayed as the independent assembly in watermark embedder device 200.However, it should be understood that receiver 204 can be at larger media
It sets the component in the different sub-systems of device or may be incorporated into as a part for another component in watermark embedder device 200.By
It can be in various formats in modified embedded content 202 and may include several audios, video, multimedia or data-signal, therefore connect
Modified embedded content 202 is suitably adjusted to be recognized by other components of watermark embedder device 200 proper by receipts device 204
When form can be necessary.The adjusting may include signal processing operations, such as demodulation, decompression, release of an interleave, decryption, solution
Disturb, resample, A/D conversion, reformat, filtering etc..
Stego-key selection component 208 is then from the stego-key for being used to extract embedded watermark by watermark extracting engine 206
At least one stego-key of Resource selection.In some embodiments, stego-key is stored in stego-key storage assembly 210
In.Watermark extracting result 216 is generated by watermark extracting engine 206.Watermark extracting result 216 may indicate that the presence of recovered watermark
And/or the instruction of value and the reliability to recovered watermark and other information.It can be with predetermined time interval (for example, to specific
User or device) report watermark extracting result 216.In some embodiments, distortion estimation component 212 is generated to may be present in
The estimation of the type and amount of one or more distortions in reception content.Estimated distortion information can be there are content distortion the case where
Under be used to promote the detection of embedded watermark by watermark extracting engine 206.Estimated distortion information can also be transmitted to assessment device assembly
214, assessment device assembly 214 is promoted the detection of embedded watermark by assessing by the watermarked frame that watermark extracting engine 206 is handled.One
In a little embodiments, assessment device assembly 214 can perform the operation of such as the following terms：Refer to extracted potential watermark or watermark segmentation
Send weight；Combine assigned weight；It assesses extracted watermark or whether to meet wanted false watermark detection general through accumulating watermark segmentation
Rate etc..In some embodiments, watermark extracting result 216 can be defeated by assessment device assembly 214 rather than watermark extracting engine 206
Go out.Distortion estimation component 212, assessment device assembly 214 and watermark extracting engine 206 can with contain predistortion watermark template and other
The storage device 218 (for example, database) of information communicates.Predistortion watermark template will be discussed in detail in following sections.Storage
Device 218 can be the physical memory cell identical or different with stego-key storage assembly 210.It should be noted that promote to basis
The understanding of concept, Fig. 2 illustrate the independent assembly in watermark embedder device 200.However, it should be understood that in some embodiment party
It, can be by the two in described component or both combination of the above together in case.For example, in some embodiments, may be used
Distortion estimation component 212, assessment one or more of device assembly 214 and storage device 218 are incorporated as watermark extracting engine 206
A part.
It in some applications, can be if the instruction of watermark extracting result 216 detects at least one conclusive watermark
The 200 other coercion actions of outside execution of watermark embedder device (such as stop content playback, stop content record, display alarm
Message etc.).If the instruction of watermark extracting result 216 detects at least one tentative watermark, extracted tentative watermark
Use can be limited to the various adjustment operated to detector with (for example) starting in the watermark embedder device 200.Such as compared with
Mentioned by morning, can be based on the tentative watermark of threshold test, the threshold value depend on tentative watermark use and handle load with
Being traded off between robustness gain.
Many intentional attacks to digital watermarking or even regular content processing operation cause to change watermark location, size,
The watermark distortion of orientation and other parameters.For example, picture rotation can be led to, in time domain by recording the film being shown on screen
Or cutting, picture stretching, angle modification etc. in spatial domain.Similarly, audio frequency watermark can be subjected to time shift (for example, due to sound
Frequency cut or concatenated with other audios cause), time-scaling, pitch-shift, tremble and tremble.Can by watermark extracting it
Preceding detection content is distorted and makes content distortion reverse phase and compensates most distortions in the distortion.Alternatively, in some embodiments,
It can be by adjusting watermark search program step so that extracted candidate's watermark and predistortion watermark template are related and complete or partial
Alleviate the influence of such distortion in ground.
Distortion detection and compensation can be that manually or automatically, and it can be with or without the original contents at detector
In the case of realize.Clearly, most desirable method is to provide automatic distortion detection without the use of original contents to simplify water
Print detection program step.In general, this can be realized by introducing synchronous (or registration) watermark.
The exemplary watermarked frame of length L of Fig. 3 A displayings comprising sync section (SYNCH) and packet portion (PACKET).Scheming
In the exemplary pattern of 3A, to put it more simply, describing a dimensional view of watermarked frame (for example, corresponding to audio content).However, should manage
Solution, similar principles are applicable to the watermark with two or more dimensions (such as in being embedded in image, video sequence etc.
Watermark).The synchronization of watermarked frame and packet portion can include respectively multiple symbols (that is, binary system or non-binary symbols).One
In a little embodiments, the sync section of watermarked frame can be the special style or sequence being embedded in content host.In this way, watermarked frame
Sync section is sometimes referred to as synchronization watermarking, synchronous pattern, sync mark, synchronizing sequence etc..One in the function of sync section
Person is the watermarked frame helped in positioning host content.For example, watermark embedder can search for content (for example, process content
Sample) to find the known synchronous pattern of the beginning of label watermarked frame.After successfully being detected sync section, watermark embedder can
Then it attempts to restore the watermark packet symbol after sync section at once.In some embodiments, the detection of synchronous pattern is with thick
For slightly granularity by the beginning sender of watermarked frame, this can further be triggered carries out extra process to increase watermark detection with fine granulation
Reliability.In some systems, watermarked frame can contain be useful for (for example) with different grain size grade establish watermarked frame position and/
Or by two or more sync sections of the beginning of watermarked frame, end and/or other section senders.
In exemplary extraction operation workflow, after the successful compilation of SYNCH and detection, collect PACKET.So
And when there is the distortion of significant quantity in the content, detection SYNCH can be difficult or impossible to.Fig. 3 B shows are in successfully detection mark
Collect PACKET#1 after the parts SYNCH of the beginning of note PACKET#1.However, every in PACKET#2 and #3 due to leading over
The distortions of the parts SYNCH of one and the PACKET#2 and #3 that do not collect.The extraction of watermark for further illustration considers wherein
Audio content is embedded with watermark and is then linearly compressed (example in time due to intentionally or accidentally content processing operation
Such as, moved up by associated tone) application.Depending on specific watermark embedded technology, the parts SYNCH through embedded watermarked frame can
The time-scaling of specific quantity can be tolerated before its detection failure.Some situations of linear session scaling are suffered from content
In, if can estimate the amount of time-scaling, can by attempt extract SYNCH patterns before in the opposite direction to content
It zooms in and out and improves the immunity to time-scaling.By this method, make the reducing compressed reverse phase of audio time (that is, by compensating) and
Watermark detection becomes to damage for large range of compression feasible.
In another embodiment, content of the alternative modified through time-scaling (or damaging in other ways), generates one or more
It a predistortion synchronization template (for example, compressing synchronization template through the time) and uses it for examining in the case where not changing content itself
It surveys through embedded synchronous pattern.As mentioned earlier, the position of watermarked frame is established using SYNCH patterns.SYNCH templates are through shape
At the predefined SYNCH patterns of the detection to promote the SYNCH patterns being embedded in content.According to some embodiments, SYNCH inspections
Survey program step include processing received content with obtain candidate SYNCH patterns and then by candidate SYNCH patterns with may include
One or more SYNCH templates of multiple predistortion SYNCH templates are compared.Analysis and empirical result instruction generate one group and lose in advance
True SYNCH templates provide and zoom in and out the watermark detection reliability as much but under relatively low processing cost to content.
In order to promote the understanding to disclosed embodiment, consider that wherein SYNCH patterns include embedding using specific embedded mobile GIS
The example for entering 50 sequences in the sequence section of audio content is beneficial.It is further assumed that SYNCH patterns is each
Position was across 1 second of host audio content.For the audio content with per second 44,100 samples, SYNCH patterns each
It is embedded on 44,100 samples of content.It detects a kind of method of the SYNCH patterns from received content and collects institute for (1)
The 44 of reception content, 100 samples, (2) handle the sample to extract candidate place value, and (3) repeat step (1) and (2) 49 times
To form 50 candidate's SYNCH patterns, and candidate SYNCH patterns and one or more SYNCH templates are compared to determine by (4)
Whether candidate SYNCH patterns correspond to the true SYNCH patterns with particular trusted grade really.
Due to being unaware of position of the SYNCH patterns in content, can need to repeat step (1) in a manner of sliding window
To (4), wherein step (1) to (4) it is each repeatedly in, the treated sample of received audio content with previously weighed repeatedly
It is folded.In egregious cases, repeats step (1) and arrive (4) 44,100 times there are one the 44,100 of sample searches granularity to generate tool
The pond of a candidate's SYNCH patterns.However, in most practical applications, this radical approach is computationally infeasible, also has to detection
There are the insertion SYNCH patterns of rational reliability unnecessary.Therefore, step is repeated generally directed to each watermark bit duration
(1) (4) are arrived several times., sometimes referred to as with X position granular search's content, wherein X indicates to repeat to walk for each bit duration for this
Suddenly (1) to (4) or at least number of step (1) and (2).In some embodiments, detection program step can rough sub- position grain
It spends (for example, 3 sub- positions) to start, and repeats or restart with higher sub- position granularity once detecting potential SYNCH patterns
Search.
In the case where being distorted there is no content, the candidate SYNCH samples being spaced apart each other with nominal range that collect can be passed through
Formula place value executes SYNCH detection process referred to above.Figure 17 is the SYNCH compilations that can be executed according to one exemplary embodiment
And/or the simplification figure that template generates.As mentioned earlier, sub- position pond indicates the candidate place value generated with sub- position granularity.Figure 17's
Nominal stencil-chosen operation in exemplary pattern (or more generally, separates the son of d0 comprising every one the 5th sub- position of selection
Position).In the above example, wherein SYNCH patterns are 50 bit sequences, select 50 such sub- positions to form candidate SYNCH patterns.
Then candidate SYNCH patterns it is compared to determine with nominal (or not being distorted) SYNCH templates to what extent to match
(for example, detecting that how many mismatch position).
It, can be by be selected different from the spacing of nominal spacing in the case where there is distortion (such as time-scaling distortion)
Sub- position and form candidate SYNCH patterns.With reference to figure 17, the distortion (for example, the time extends) of the first kind can be directed to distance d1
And sub- position is selected with distance d2 for the distortion of Second Type (for example, the time shrinks).In some situations, different sons will be based on
The candidate SYNCH patterns that column pitch obtains are compared with nominal SYNCH templates to what extent to be matched with evaluating it.
In some embodiments, it substitutes and candidate SYNCH patterns is matched with nominal SYNCH templates, by candidate SYNCH
Pattern is compared with one or more predistortions SYNCH templates that may differ from nominal SYNCH templates.It can be obtained by following operation
Obtain such predistortion SYNCH templates：So that a content (or various contents) is embedded with SYNCH patterns, make content by various distortions,
And with nominal and/or modified sub- column pitch from distortion contents extraction SYNCH patterns.For example, with reference to figure 17, can by with
The N number of sub- position of d1 spacing selection obtains the predistortion template #1 of length N (for example, N=50), and can be N number of by being selected with d2 spacing
Sub- position obtains the predistortion template #2 of length N.Most had to scaling type distortion based on different sub- column pitch selection predistortion templates
Effect.It, can be by with nominal spacing d0 for some distortions (for example, acoustic propagation, intensity/amplitude modulation, Video Camera capture etc.)
It selects sub- position and obtains predistortion template.
In general, many dimensions of predistortion SYNCH template development distorted spaces can be used.Fig. 4 displayings wherein make original
(or nominal) the SYNCH templates 402 (that is, SYNCH templates in the case where being distorted there is no content) that begin pass through various potential mistakes
True A to K404,406,408,410 are to generate the example of one group of predistortion SYNCH templates A to K414,416,418,420.Fig. 4 is also
What description can together be stored with predistortion SYNCH templates A to K414,416,418,420 is not distorted SYNCH templates 412.Template
It (includes linear or non-thread that predistortion, which is applicable to modification in one or more dimensions in frequency domain/time domain/spatial domain/amplitude domain,
Property scaling (distortion) and rotation) content.The non-exhaustive list that can influence the distortion for the watermark being inserted into image includes figure
Image rotation turn, stretching, the dynamic range compression of illumination intensity signal, color balance etc. in horizontal or vertical dimension.It can influence to be inserted into
The non-exhaustive list of the damage of watermark in audio content includes the time-scaling that linear session scales, tone is constant, time
It constant tone scaling, dynamic range compression, equilibrium and resamples.
It should be noted that being easily extended to include watermark packet about the discussed above of SYNCH patterns.In addition, disclosed original
Reason may extend into comprising in two or more dimensions for being embedded in spatial domain, time domain and/or frequency domain SYNCH or
PACKET.Such SYNCH or PACKET can be indicated by binary system or non-binary symbols, and its detection can be based on binary system or non-
Binary character and/or parameter.For example, generally can be described as the parts SYNCH of embedded watermark that relative program can be used
The predefined signal pattern of step extraction.
Fig. 3 C are illustrated to collect with the similar PACKET of the identical packet of packet shown in Fig. 3 B and be operated, but in Fig. 3 C
The successful detection of embedded SYNCH is realized in operation using predistortion SYNCH templates in the case where being distorted there are content.Treating capacity
And false watermark detection probability depends on the number K of the predistortion template for detecting embedded synchronous pattern.The number of predistortion template
The range and granularity of the distortion that mesh K is adapted to depending on needs and the digital watermark sensibility to certain distortion.For example, base
Generally time-scaling distortion is compared in the digital watermark of spread-spectrum more sensitive based on the technology for replicating modulation.Therefore, when with
When duplication modulation technique is compared, the technology based on spread-spectrum needs more predistortion templates to be distorted to be scaled in existence time
In the case of watermark is effectively detected.
Constant tone scaling (TIPS) modification makes the situation that audio frequency watermark is distorted between considering when used.It is further false
If predistortion watermark template is only counted and following TIPS modified values：3%, 6% and 9%.In this case, watermark embedder can only may be used
By ground detection come the watermark of the content by about 3%, 6% or 9%TIPS damages (plus-minus tolerance value) distortion self and by its
It is reported as tentative or conclusive.In this exemplary situation, when audio frequency watermark with any other TIPS ratios (for example,
4.5%) when being distorted, there is a possibility that can't detect conclusive watermark.Clearly, the selection of the number of predistortion template is notable
Wanted reliability depending on SYNCH detections.Increase with the granularity of predistortion template, detects reliable SYNCH (for example, tool
Have low level error count or high correlation coefficient value) chance also increase, but handle load also increase.Therefore it provides predistortion is same
It is important to walk the compromise between the number of template and false watermark detection probability.
In some embodiments, granularity (and therefore, the predistortion template of the predistortion synchronization template within the scope of certain distortion
Number) be selected so that and detect at least single test watermark in being distorted content.For example, selection is pre- in such manner
It is distorted the number of synchronization template and spacing makes even if distortion between two predistortion templates centres (as in the above example), still
In the presence of the good opportunity for detecting that SYNCH patterns subsequently then detect tentative watermark.
Once finding tentative watermark, extractor can start the more search meticulously of distorted space.Fig. 5 illustrates basis
One group of operation 500 that example embodiments execute.At 502, watermarked frame is received.Institute's receiving frame may include synchronous (SYNCH) and
Watermark packet (PACKET).At 504, received SYNCH is compared with multiple predistortion SYNCH templates.At 506, really
It is fixed whether successfully to be detected SYNCH, and if so (that is, being "Yes" at 506), described group of operation 500 is at 508
Continue the PACKET that collects.On the other hand, if the determination at 506 does not indicate successfully SYNCH detections (that is, being "No" at 506),
So described group of operation 500 continues at 520, wherein abandoning PACKET.
Referring back to Fig. 5, at 508, PACKET compilations may include various operations, such as release of an interleave, descrambling and other packets
Formation operation.At 510, the number of the mistake in compiled PACKET is determined.For example, the operation at 510 may include making
It is decoded with error correction/detection code, template matches etc..At 512, the number of the mistake in compiled PACKET is determined
Whether with strong (that is, conclusive) detection of watermark associated first threshold is less than, and if so (that is, being at 512
"Yes"), then declaration detects conclusive watermark at 514.If determining error value of the instruction more than first threshold at 512
(that is, being "No" at 512), then described group of operation 500 continues at 516, wherein determining the mistake in compiled PACKET
Number whether be less than second threshold associated with the detection of tentative watermark.If being determined as "Yes", institute at 516
It states a group operation 500 at 518 to continue, wherein declaration detects tentative watermark.On the other hand, if be designated as at 516
"No" then described group of operation 500 continues at 520, and abandons PACKET.
Fig. 6 is to illustrate the component 600 that can be used for executing some or all of operations in the operation described in Fig. 5
Simplification figure.Watermark frame grabber 602 is configured to receive the watermarked frame of the parts SYNCH and PACKET comprising watermarked frame.Compare
Device 604 (a) is configured to receive SYNCH and is not distorted with predistortion synchronization template and it one or more in SYNCH templates
Person is compared.The result of the comparison can be by distortion estimator 604 (b) for obtaining to being present in one or more in content
The estimation of distortion.For example, distortion estimator is recognizable generates mould synchronous with the predistortion of the best match of received SYNCH
Plate, and one or more type of distortion associated with identified predistortion SYNCH templates and/or amount distortion are selected to indicate to be present in
Distortion in content.In some embodiments, comparator 604 (a) and distortion estimator 604 (b) are combined into single component (example
Such as, comparator 604 (a)).
Template can be stored in storage assembly 606.PACKET assembly components 608 are configured to receive at least PACKET and pass
One or more of received SYNCH and predistortion synchronization template are obtained in comparator 604 (a) or are not distorted SYNCH templates it
Between matched success or failure instruction.After detecting successfully SYNCH, PACKET assembly components 608 collect watermark at once
The PACKET of frame, the PACKET determine component 610 for generating error count by PACKET mistakes.Produced error count by
Evaluator 612 including decision logic component for determine error count be arrest limit be also above in specified packet error range or
Less than the specified packet error range.For example, evaluator 612 can be determined whether to detect conclusive or tentative watermark
(and provide and indicate its output).Evaluator 612 can also generate will be due to comparator 604 (a) and/or distortion estimator 604 (b)
Detecting SYNCH, it fails to match or the instruction of PACKET is abandoned when can't detect tentative or conclusive watermark.Evaluator 612
Output can be used for triggering operation bidirectional, such as the triggering extension search packet when detecting tentative watermark.The output of evaluator 612
It may include additional information, such as watermark value (that is, watermark state, watermark style etc. after testing), it is credible in such watermark detection
The instruction (for example, packet error count, timing error counting etc.) of grade, is present in amounts and type of one or more distortions in content
Etc..
The multistage method in conjunction with described in Fig. 5 and 6 realizes overall process load and associated vacation in watermark embedder
The efficient balance of positive watermark detection probability, and it is further improved overall robustness and the safety of watermark extracting.
In some embodiments, two step searches that can implement distorted space in such manner make predistortion SYNCH templates
Trigger extension extractor operation immediately with matching between candidate's SYNCH patterns is detected, for example, extend to synchronous pattern and/or
The position of the search of watermark packet and granularity.It may not be generated in processing load and vacation however, this of extractor operation extends immediately
Optimal compromise in terms of positive watermark detection probability.As an example it is assumed that wishing the time constant in -20% to+20% tone
It scales and improves conclusive watermark extracting performance in (PITS) distortion range.Search granularity be must be carefully selected properly to fill out
Enter predistortion SYNCH template databases 606.It, can if selecting fine search granularity (for example, there is 1%PITS spacing)
Must utilize a large amount of predistortion templates (for example, 40 predistortion SYNCH templates and one be not distorted SYNCH templates) in order to by
Detect SYNCH patterns in ground.In this scenario, a large amount of predistortion SYNCH templates can negatively affect processing load and false water
Print detection probability.In addition, if predistortion template matches have high false SYNCH detection probabilities, then every SYNCH detections can
Triggering extension extractor operation, this will increase processing load and false watermark detection probability.On the other hand, if selection rough search
Granularity (for example, there is 5%PITS spacing), then there is the possibility that can't detect embedded SYNCH and corresponding watermark PACKET.
In addition, if predistortion template matches have too low false SYNCH detection probabilities, it would be possible that cannot detect in distortion content
Many watermarks.However, this defect can be compensated by increasing the number of predistortion template using extra process load as cost.Cause
This, generates finite population predistortion synchronization template and so that improving true watermark examines using it in such manner in watermark embedder
It surveys, processing loads and the balance of false watermark detection probability is advantageous.
Fig. 7 illustrate according to one exemplary embodiment can be executed with generate one group of predistortion watermark template one group of behaviour
Make 700.It should be noted that in described group of operation 700 in describing Fig. 7, with reference to predistortion watermark template.However, it should be understood that can needle
This class template is generated to synchronous one or both of pattern and known watermark ladle sample formula.For example, there are two effectively to be born
The watermark of load can only by one of four different conditions there are senders.In such a system, unique symbol pattern can be with watermark
Each of state is associated, and is directed to each unique symbol pattern, the executable described group of operation 700 described in conjunction with Fig. 7.
With reference to figure 7, at 702, one or more pilot signals are generated.These signals can correspond to specific function, such as arteries and veins
Punching, SIN function, square-wave, flat image, edge image etc..It should be noted that can be counted using term " signal " to convey
Word or the certain types of content of analog form transmitting, reception and storage.Such signal can be by signal processing operations, such as adjust
System/demodulation, A/D and D/A are converted, are compressed, resample etc..The pilot tone of (or obtaining in other ways) is generated at 702
Signal be once embedded with synchronous pattern and/or watermark packet and distortion (or being emitted by noisy communication channel) can provide about distortion and
How the characteristic of launch channel and such distortion and/or transmitting influence seeing clearly for embedded watermark.In alternative embodiments, it substitutes
Pilot signal additionally or alternatively generates or obtains representative content (for example, concatenation of different types of film) at 702.
At 704, pilot signal is made to be embedded with watermark.At 704, embedded pilot signal undergoes one or more distortions.This
A little distortions may include, but are not limited to various time-scalings distortions, A/D and D/A distortions, cut, rotation, scaling, space and/or when
Between shift etc..At 706, the distortion is applied to host by one or more distortion parameters that may extend across particular range
Pilot signal.For example, it can be scaled according to linear session of 1% increment between -10% to+10%
(LTS) percentage parameter application LTS is distorted.At 708, watermark is extracted from distortion pilot signal.At 710, for it is each at
Work(watermark extracting generates the predistortion candidate's watermark template for corresponding to the certain distortion with certain distortion parameter.At 712,
One group of distortion watermark is received from field.In an example, the real generation by noisy communication channel has been undergone in host pilot signal
Described group at 712 is generated after boundary's transmitting, by extract watermark from host pilot signal after watermark at 704 insertion
It is distorted watermark.For example, the host pilot signal containing embedded watermark can be played back in cinema, be captured and be inputted using Video Camera
To watermark embedder to obtain distortion watermark.
At 714, distortion watermark will be received and be compared with candidate predistortion watermark template.This, which compares, can disclose candidate
Which one in predistortion watermark template is best match or correlation.At 716, by related predistortion watermark template classification (example
Such as, with related or best match order) and be stored in database.In some embodiments, the database may include producing
The predistortion watermark template of one subgroup of raw tentative watermark or strong watermark.Note that can the new mistake from field obtained
Predistortion template database is updated and/or supplemented when true watermark with new template.
As described in the exemplary pattern of Fig. 7, each distortion in particular range and the group for distortion can be directed to
Close the program step for the experimental selection for repeating predistortion watermark template.In some embodiments, in order to will be executed to pilot signal
The number of distortion be limited in actual limit, identify critical distortion vector (for example, due to the time in particular range/sky first
Between/frequency scaling is attacked, Video Camera captures, low-bit-rate is perceived and is distorted caused by compression etc.).It is then determined critical distortion to
Predistortion watermark template of amount and combinations thereof.The all other distortion vector of predistortion template test of critical distortion vector can be compareed
(for example, non-critical distortion vector) is to determine the need for additional predistortion watermark template.
Fig. 8 is the simplification figure of the component of some or all of operations in the operation 700 for can be used for executing Fig. 7.Embedded equipment
802 are used for insertion then by the pilot signal for the various distortions for using distortion processing component 804.By distortion processing component 804
The distortion insertion pilot signal of generation is input into extraction device assembly 806, corresponds to certain distortion parameter wherein generating
The predistortion watermark template of certain distortion.In fig. 8, these watermark templates are denoted as " WM templates 1 ", " WM templates 2 " etc..Figure
8 also describe the modified embedded navigation content obtained from the field for being input to extraction device assembly 808.In some embodiments, it carries
It is same physical device to take device assembly 806 and extraction device assembly 808.In other embodiments (such as distributing is used wherein
In the embodiment of processing), it can be that can be located at the not jljl of different physical locations to extract device assembly 806 and extract device assembly 808
Manage entity.
Fig. 8, which further illustrates, waits the distortion watermark from field with the predistortion generated by extraction device assembly 806
It selects watermark template to be compared and makes its matched template matches and classification component 810.The matching of various distortions and sort operation
Result be stored in database 812.
In some embodiments, it can skip any candidate to reduce processing load and false watermark detection probability, extractor
One or more of predistortion watermark template at watermark location.In following sections, the method is sometimes referred to as in distorted space
Sparse search.A disadvantage for skipping the predistortion watermark of a subgroup is the opportunity that extractor can lose watermark extracting.However,
If the number on opportunity is fully big, the general impacts on the opportunity missed can be insignificant.In addition, in some embodiments
In, if extractor at another stage of watermark extracting program step (for example, under different extraction opportunitys) find it is tentative
Watermark, then extractor can be configured with (for example) based on tentative watermark extrapolation exploitation discussed below miss when
Some opportunitys in machine.In a kind of version, only access with high successful watermark detection probability the opportunity missed and
And the not all opportunity missed.This by accessing only matching and (one or several) the associated mistake of tentative watermark after testing again
Those of true parameter opportunity and realize.
When designing the sparse search in distorted space, disclosed embodiment can utilize following two features.First, at random
(or pseudorandom) selects the predistortion watermark template of a subgroup.If this decision is deterministic, attacker can test
Mode finds certainty pattern by reverse-engineering, and then adjustment is distorted pattern to match skipped search space to escape
Watermark detection.
Secondly, if extractor knows that the possibility of existing certain distortion is advantageous.For example, if watermark passes
Up to " cinema projection " state in audio-visual content, then typical piracy situation is related to recorded content.For this attack situation, with than
The high probability selection of other predistortion templates (for example, time/frequency scaling etc.) is suitble to the predistortion template of type (for example, right
It should be in the template of cutting, stretching, rotation etc.).Similarly, if instruction content " trusting in digital encrypted form is conveyed in watermark
Source " state, then perception code conversion is that possible attack and with high probability selection predistortion mould associated with perception compression
Plate.In general, the probability of predistortion stencil-chosen is proportional to the probability for the certain distortion being present in content, as long as can estimate
Count this probability.
In some embodiments, two or more different watermarks are based on into line distortion operation.Two or more
Different watermark may include independent SYNCH patterns, routine tests watermark or the combination of SYNCH and tentative watermark.When those water
Being imprinted at embedded device has predefined correlation and the relationship when being known to extractor, and any back of the body with this relationship can be used
From this quality and quantity for carrying out operation distortion.For example, if two synchronous patterns in horizontal line embedded images and its through inspection
It surveys so that its line of connection and horizontal line are at an angle, then it can be assumed that the angle that image is detected by rotation.Similarly, if
Periodically it is embedded in audio frequency watermark with period P and extractor is found with audio frequency watermark spaced apart P ', then it can be assumed that content quilt
Time-scaling factor P '/P.
Distortion operation based on independent SYNCH patterns can have specific advantages, such as the processing load of reduction.However, making
Superior function can be provided with tentative watermark rather than SYNCH patterns, because tentative watermark usually has relatively low false detection probability
And preferable position resolution.As mentioned earlier, when compared with the detection of watermark packet, the detections of SYNCH patterns usually by compared with
High vacation detection probability characterization.This is done to balance the processing load of extraction procedure step and robustness requirement.Therefore, it is based on
Individually the distortion operation of embedded SYNCH patterns can generate incorrect or unreliable distortion estimation.However, if detecting tentative
Watermark, then it provides the additional confidence level of the presence of watermark and position after testing.
Once being based on two or more watermark operation amount distortions, next step is whether assessment distortion is reality
's.As an example it is assumed that the image rotation during video record will be not more than 20 degree (relative to horizontal or vertical baselines) or sound
The time compression of frequency track will be safe no more than 50%.Any amount distortion of limit is predefined more than those may make figure
Picture or audio can not perceive, and therefore, generate the distortion computation of the value beyond those limits and may be attributed to false watermark detection.
When detecting real amount distortion, it then may be selected and/or generate corresponding in the neighborhood being distorted through operation
The additional predistortion watermark template of distortion, and the additional predistortion watermark template is used for other watermark search.Predistortion watermark
Template can correspond to watermarked frame SYNCH part, watermarked frame PACKET part, watermarked frame combined SYNCH and PACKET
The sub-fraction of part or the parts SYNCH or PACKET of watermarked frame.Repeatably make additional predistortion watermark template and warp
Watermark correlation is extracted to increase the reliability of watermark detection and to be distorted the reliability of operation.For example, by new watermark mould
Therefore the watermark extracting of plate can be generated (and to be compared with the mistake or the correlation higher than its fewer than the tentative watermark previously detected
Its high reliability) watermark.Then, it can be used the watermark with higher reliability after testing come estimated distortion of rerunning
Amount.(for example, when during real-time playback of the watermark extracting in content occurring) in some in real time application, can be difficult to carry out pair
The repeat search of the same segmentation of content.Therefore, in some embodiments, additional predistortion watermark template is only used for subsequent content
Segmentation.
Fig. 9 illustrates the one group of operation 900 that can be executed the search to promote distortion according to example embodiments.
At 902, tentative watermark is extracted.At 904, amount distortion is estimated in the detection based on two or more tentative watermarks.It lifts
For example, as mentioned earlier, can be used two after testing spacing between watermark estimate the linear session in audio content
The amount of scaling.At 906, determine whether estimated amount distortion is less than specific threshold.Operation at 906 provides to be estimated to determination
Meter distortion whether be reality mechanism.If estimated distortion is higher than threshold value, distortion estimation is abandoned at 912.At this point, institute
Group operation 900 is stated back to 902 to wait for the distortion for becoming available for repeating at 904 until another group of feasible tentative watermark to estimate
It is calculated as stopping, or another distortion estimation is generated using another group of available tentative watermark.
If distortion estimation is selected and/or is generated at 908 and is additional pre- less than threshold value (that is, being "Yes" at 906)
It is distorted watermark template.The algorithm for allowing to generate new predistortion watermark template and associated parameter can be used to realize the pre- mistake at 908
" generation " of true watermark template.Alternatively or in addition, appropriate predistortion watermark template can be selected from database.At 910, volume is used
Outer predistortion watermark template detects watermark with higher reliability and/or improves the estimation being distorted to content.
In being distorted content, tentative watermark band is about its accurate location and the uncertainty of parameter.Therefore, distortion fortune
Calculation is also approximate.If in conjunction with mentioned by Fig. 9, additional predistortion watermark template can be selected at 908 with improve watermark detection and/
Or distortion estimation.In some embodiments, additional predistortion watermark template is selected in such manner to realize more efficient and reliable water
Print detection and/or distortion estimation.In particular, when generating predistortion watermark template, make multiple have watermark in a manner of predefined
Content distortion, detect tentative watermark, based on the distortion of tentative watermark estimation after testing, and collect related to through operation distortion
The statistics of the mistake of connection.When at watermark embedder the additional predistortion watermark template of selection/generation (for example, the operation in Fig. 9
908) when, it is based not only on estimated distortion (for example, being obtained from operation 904) and based on distortion in the range of anticipation error
The such additional predistortion watermark template of selection/generation.Therefore, for each estimated amount distortion, multiple predistortion water may be selected
Die plate, all predistortion watermark templates are near estimated distortion.By this method, extraction is significantly improved with highly reliable
Property watermark chance, without dramatically increase template matches trial number.
In some embodiments, the estimation to distortion is executed using single test watermark.As mentioned earlier, in order to subtract
The processing of small extractor loads and false watermark detection probability, and the predistortion watermark template of an only subgroup can be used to be embedded in
The initial ranging of watermark.The template of this subgroup realizes the rough search of distorted space.It is examined when using the coarse of distorted space
When measuring single test watermark, if to be searched for compared with fine granulation near the distortion parameter of detection for triggering tentative watermark
Distorted space, then in the presence of the good opportunity that can find more reliable watermark.
Figure 10 illustrates can be executed to be embedded in one group of watermark from distortion contents extraction according to one exemplary embodiment
Operation 1000.At 1002, watermarked frame is received, and at 1004, carries out the multi-step search to tentative watermark.In a reality
In example, the operation at 1002 and 1004 is executed using described group of operation described in conjunction with Figure 5.At 1006, it is determined whether inspection
Measure tentative watermark.If tentative watermark is not detected, operation 1000 continues at 1018, under wherein system waits for
One tentative watermark.If instruction detects that tentative watermark, operation 1000 continue at 1008 to determining at 1006,
Middle estimation distortion information.For example, the estimation of distortion information may include determining the specific mistake being present in received content
The very amount of (or distortion combination).In one embodiment, this estimation can be executed by following operation：Collect watermarked frame it is one or more
A parts SYNCH (or obtaining the compiled parts SYNCH in other ways)；And by itself and the certain distortion corresponding to various amounts
One or more predistortions SYNCH templates of (or distortion combination) are compared.As mentioned earlier, in some embodiments, according to
It is executed to the estimation of amount distortion and/or the detection of tentative watermark according to the rough search of distorted space.It is produced according to this rough search
Raw watermark detection provides the certain distortion that instruction is present in embedded content host or is distorted the estimated of the coarse value of combination
Distortion information.
At 1010, selection and/or additional (or supplement) the predistortion watermark template of generation are to carry out the more smart of distorted space
Fine searching.In addition the template of selection/generation corresponds to the distortion value in the neighborhood of (rough) distortion value obtained at 1008.It examines
The TIPS for considering audio content changes example, the predistortion watermark template of wherein initial group allow in 3%, 6% and 9% value when
Between scale space rough or sparse search.If making sound during the rough search of distorted space with 3.5% in this example
Frequency content is distorted, then it is follow-up then predistortion watermark template associated with 3%TIPS to can be used to obtain best SYNCH compilations
Successful tests watermark packet collects.For this example, additional distortion watermark template (that is, operation at 1010) is selected to may include selecting
Select or generate the predistortion watermark mould in the neighborhood that 3%TIPS is distorted (for example, with 2.5% to 3.5% of 0.1% stepping)
Plate.
At 1012, reappraised using the predistortion watermark template of new selection/generation associated with tentative watermark
Wrap the number of mistake.In one embodiment, in PACKET and the additional predistortion watermark template candidate restored some or
All additional predistortion watermark templates are compared to generate associated error count.In this way, new one group of predistortion PACKET
Template can be used for the repeat search of the same content section to wherein detecting tentative watermark or in subsequent content segmentation
Search.
At 1014, selection generates the packet or packet template of minimal error number.Selected packet/packet template corresponds to more acurrate
Ground indicates the amount distortion for the true amount distortion being present in received content.Therefore, in some situations, the packet through reappraising is wrong
It accidentally may indicate that the presence (and more acurrate position) of the insertion watermark with higher reliability.For example, the operation at 1014 can
Cause to detect strong or conclusive watermark.At 1016, it is used for based on the packet/template renewal for generating best error value at 1014
The distortion information of the following packet of compilation and/or synchronous pattern.
It should be noted that although the operation at 1014, which is described as selection, generates least number of errors purpose packet/template, 1012 and
Operation at 1014 should also generate the selection corresponding to low false positive detection probability.In alternative embodiments, using repetitive process
To realize the extraction of watermark and to the estimation of distortion.In this embodiment, one at a time selection/generation in the mistake previously estimated
Supplement predistortion watermark template in genuine neighborhood.For example, the first supplement predistortion watermark template can correspond to than being used for
The content that the predistortion watermark template of tentative watermark detection at 1004 is less slightly is distorted.If the first supplement predistortion watermark mould
Plate generates less bit-errors or higher related to candidate watermark, then being to correspond to by the second supplement predistortion watermark stencil-chosen
In the content distortion less than the first supplement predistortion watermark template.On the other hand, if the first supplement predistortion watermark template is led
Cause the correlation of larger bit-errors number or reduction, then by second supplement predistortion watermark stencil-chosen be corresponding to more than with
The content distortion of the associated distortion of predistortion watermark template for the tentative watermark detection at 1004.Iteration can continue directly
Until one of the predistortion watermark template best watermark detection of offer and to the best estimate of content distortion.In some examples
In, it numerical method (such as dichotomy or Newton method) can be used to promote the convergence of iteration.Iterative process as described above is more
Efficiently and with the false detection probability lower than multiple search under fine granulation obtain the most reliable estimation of watermark after testing.
In some embodiments, after detecting tentative watermark, watermark detection is improved using extrapolation technique at once
Reliability.In being distorted content, individual watermarks can be corrupted to the effort in spite of distortion operation as described above and compensation
But no conclusive watermark detection is possible degree.In these cases, extractor can near tentative watermark after testing after
The continuous search to the subsequent watermark in content.However, if finding one or more additional tests watermarks in subsequent searches, that
Can operation two or more watermarks joint " weight " and make to use it to further to reappraise watermark presence.
In some embodiments, it can be used and be embedded in pattern dependent on watermark to predict position, the Symbol Style of adjacent watermark
And the alternative of watermark parameters.Usually, it is contemplated that the repetition throughout the same watermark of content is embedded in, but more complicated embedded pattern
It is advantageous to be possible, and in some cases.For example, in some situations, adjacent watermark can be in a known way
(such as when embedded watermark includes with the incremental counter of the insertion of each additional watermark packet, or working as will with different scrambling key
When the payload scrambling of adjacent watermark packet) it is different from each other.The feelings of the insertion pattern of known subsequent and/or previous watermark wherein
In border, the detection for the tentative watermark that can extrapolate is with the presence of other embedded watermarks in the pre- discharge position in evaluation content and value.
In some embodiments, the presence of other watermark styles in predictive content is carried out using tentative watermark after testing.Then test
This prediction (for example, using test statistics method is assumed), this can lead to the refusal of the feasibility of watermark detection or confirmation.
Figure 11 A displaying, which has, is similar to the exemplary that the parts SYNCH and PACKET of part those of are presented in Fig. 3 A
Watermark frame sequence.After successfully collecting and detecting SYNCH, collect PACKET.When the distortion of significant quantity is applied to content,
It can be difficult or impossible to the parts SYNCH of detection watermarked frame.Figure 11 B shows collect after the previous SYNCH that successfully collects
PACKET#1.However, due to the parts correspondence SYNCH of watermarked frame #2 and #3 serious distortion and the PACKET#2 and #3 that do not collect.
Figure 11 C displayings are held wherein after detecting tentative watermark across watermarked frame #2 and #3 (it is expected that including PACKET#2 and #3)
The example of the extrapolation of the tentative watermark of row.In exemplary extrapolation situation, it is contemplated that some symbols in the symbol of extrapolated watermark
With the associated symbols in (that is, tentative) watermark of not extrapolating.After the symbol for extrapolated watermark of collecting, it is wrong that symbol is counted at once
Accidentally, estimate false watermark detection probability and assign weight to extrapolated watermark.Assigned weight can be added up to evaluate watermark jointly
In the presence of and value.
The extrapolation technique of disclosed embodiment includes that can improve the possibility of the embedded watermark of detection while make false watermark detection
Probability maintains the feature in desirable limit.In particular, in some embodiments, when extrapolating to search for additional watermark
When, additional SYNCH patterns need not be found so as to the next watermark packet symbol that collects.In being distorted content, many SYNCH samples
Formula is likely to be broken and can not recognize.However, even if not being directed to the portions SYNCH that all potential watermarked frames detect watermarked frame
Point, the extrapolation technique of present application still realizes compilation and the use of packet symbol.
In addition, in some extrapolation techniques in the extrapolation technique of disclosed embodiment, to detect extrapolated watermark
The number that extraction is attempted can be as low as 1.It therefore, there is no need to adjust the weight assigned to extrapolated watermark symbol to maintain wanted false inspection
Survey probability.However, in a particular embodiment, in order to count and it is associated with the distortion estimation of tentative watermark and true place not really
It is qualitative, can be used it is several start offset and several distortion estimations execute watermark extrapolation attempt.The method promote realize robustness with
Optimal compromise between false detection probability.Even if in this scenario, the number of trial is still significantly lower than is based on embedded water wherein
In the system that the independent detection triggering weight of print adds up.In this independent detection of embedded watermark, each watermark is (for example, each
Tentative watermark) detection need the detections of the associated parts SYNCH, as illustrated in Fig. 3 B and 11B.
Exemplary plot of the false watermark detection probability of Figure 12 displayings to watermark weight threshold.Exemplary graph in Figure 12
Figure corresponds to the experimental result obtained for the watermarked frame with binary character, and the binary character, which has, is embedded in audio
The parts 32 SYNCH in content and 64 parts PACKET.Curve 1220 corresponds to the sparse granular search of distorted space, and
Curve 1240 is corresponded to the fine granulation obtained using the watermark extrapolation technique of disclosed embodiment and searched for.Figure 12 also shows that correspondence
In the independent recovery operation based on watermark to realize the third curve 1260 of the false watermark detection probability of the cumulative extractor of weight.
Note that being directed to about 5 × 10-9False watermark detection probability, it is (sparse and finely search using disclosed tentative watermark extrapolation technique
Both ropes) watermark weight threshold be less than independent weight accumulation techniques used in threshold value.In this way, with based on embedded watermark
The weight accumulation techniques of independent detection are compared, the disclosed extrapolation technique of the extrapolation based on the tentative watermark previously detected compared with
Low calculation processing load is lower to generate similar or preferable false positive performance.Meanwhile it is opposite with the detection of conclusive watermark, based on previous
The detection of the extrapolation of the tentative watermark of detection needs relatively low watermark detection threshold value for the detection of tentative watermark.These features
Greatly improve the robustness performance of watermark embedder.
Figure 18 illustrates the one group of operation 1800 that can be executed with watermark of extrapolating according to one exemplary embodiment.1802
Place, extracts tentative watermark.For example, the operation at 1802 may include combining in the discussed operations of Fig. 5,9 and 10 extremely
Some few operations.At 1804, one or more extrapolated watermarks are formed.For example, embedded host can be positioned at by obtaining
Realize that this packet is formed in content relative to the symbol of the potential watermarked frame at the predefined position of extracted tentative watermark.
At 1806, extrapolated watermark and extracted tentative watermark are evaluated jointly.At 1808, evaluating whether jointly for watermark is determined
Cause to meet and wants false watermark detection probability.For example, this wants false watermark detection probability to can correspond to and conclusive watermark
Associated detection probability.
According to disclosed embodiment execute watermark extrapolation amount depend on distortion essence and distortion estimation it is accurate
Property.For example, in the situation for the audio frequency watermark that the acoustic propagation due to content host is distorted, when ambient noise and reverberation are seen
When like being main distortion sources, time-scaling distortion is usually insignificant.In the environment, it is assumed that it is periodically embedded in watermark,
It may carry out the successful extrapolation across several watermark length.On the other hand, it if content is attacked by time-scaling, can limit
System accuracy associated with estimation time-scaling distortion, and therefore, only peanut watermark is extrapolated on each time orientation
Can be feasible.
In some cases, distortion changes due to content conversion or intentional attack and in the time or spatially.For example,
Known audio tape player is generated due to the variation of the speed of the motor of tape player and is trembled or trembleed illusion.Therefore,
Generated time-scaling factor is changed over time based on the variation of motor speed.Similar techniques can be used to do in attacker
Disturb the appropriate estimation and compensation of the time-scaling distortion in the content of watermark.However, a warning is too fast and/or too big
The distortion of time change generally produces offensive perception illusion and reduces the feasibility of attack.Therefore, exploitation exists
Increase the more effective right of the robustness of watermark extracting in the watermark duration in the case of the distortion of slowly varying time change
Error resilience mistake can be beneficial.
The experiment instruction carried out by inventor, when being distorted slowly varying in the length of watermark, usually by probably suitable
The predistortion SYNCH templates for closing certain within the scope of distortion variations realize tentative watermark detection.However, being typically larger than
In the duration of the watermark PACKET of the parts SYNCH, distortion can be fluctuated to outside the detection range in predistortion watermark template
Value.This means that predistortion packet template can only match the actual distortion of the limited section of watermarked frame, causes notable symbol wrong whereby
Accidentally rate.
According to some embodiments, know embedded watermark symbol pattern (for example, by the experiment previously detected in advance wherein
Property watermark payload, pass through and consider only one group of limited payload etc.) situation in, entire watermark interval is divided into several
It is segmented and carries out watermark extracting trial for each segmentation.It can refer to Figure 11 D and more fully understand this operation.Note that in Figure 11 D
Horizontal scale those of may differ from Figure 11 A to 11C horizontal scale.To detect and/or try from the SYNCH of PACKET#1
The property tested watermark detection obtain distortion estimation start, using in the predistortion watermark template corresponding to estimated distortion value one or
More persons attempt the extraction of the first watermark segmentation SEG1.In some embodiments, SEG1 include SYNCH's it is contemplated that PACKET#1 it
Afterwards all or part of.For example, if watermarked frame is made of 20 SYNCH and 100 PACKET, in an example
In, each segmentation may include 20 positions (that is, 6 segmentation/frames).In another example, each segmentation may include 30 positions (that is, 4
A segmentation/frame).In a further example, each segmentation may include 10 positions (that is, 12 segmentation/frames).It, can in another example
The parts SYNCH are skipped, and can only execute the two stage cultivation of the parts PACKET of watermarked frame.Above example is not intended to provide different points
The exhaustive list of Duan great little and segmented configuration, but being provided to illustrate can be based on watermark detection system and intended application
It needs and ability selects different fragment sizes and configuration.
In some embodiments, range (for example, the duration, spatial dimension etc.) choosing based on extracted tentative watermark
Select the number of segmentation.For example, the watermark with smaller payload can be segmented into than the water with larger payload
Print few segmentation.Determine that the other factors of the number of segmentation are the amount distortion in embedded content host in the presence of (or expecting exist)
And/or there is the type of distortion of (or expecting exist) in embedded content host.The selection of the number of segmentation can based on (citing come
Say) amount distortion and type dynamically change, or can statically or semi-statically carry out, wherein the number being segmented is fixed.
In addition, can based in content expectation amount distortion and/or type select the length of each segmentation.
Referring back to Figure 11 D, in some embodiments, the predistortion watermark template closest to estimated distortion value is used for
The extrapolated detection of SEG1 and one or more predistortion watermark templates corresponding to the distortion near estimated distortion value.Nothing
It is matched by which and attempts to generate optimum, be associated the beginning distortion estimation that distortion levels are used as next segmentation SEG2.On
Operation mentioned by text for extrapolated watermark other segmentation SEG2 ..., SEG K continue.It by this method, can be by using proper
When predistortion watermark template effectively tracks and alleviates slowly varying distortion.
Figure 13 illustrates can be executed to extrapolate one group of watermark on the basis of by segmentation according to one exemplary embodiment
Operation 1300.At 1302, watermarked frame is received, and at 1304, carries out the multi-step search to tentative watermark.In a reality
In example, the operation at 1302 and 1304 is executed using described group of operation described in conjunction with Figure 5.At 1306, it is determined whether inspection
Measure tentative watermark.If tentative watermark is not detected, described group of operation 1300 continues at 1320, wherein system
Next frame is waited for, and then repeats the operation carried out at 1302 to 1306.If instruction detects experiment to determining at 1306
Property watermark (that is, at 1306 be "Yes"), then described group of operation 1300 continues at 1308, wherein estimating distortion information.It lifts
For example, to the estimation of distortion information may include the amount for determining the distortion (or be distorted combination) being present in received content and
Type.In one embodiment, this estimation can be executed by following operation：It (or has been converged the parts SYNCH for the watermarked frame that collects
The parts SYNCH of volume)；And by one or more predistortions SYNCH of itself and (or several) certain distortion corresponding to various amounts
Template is compared.As mentioned earlier, the estimation and/or examination to amount distortion are usually executed according to the rough search of distorted space
The detection of the property tested watermark.
At 1310, the granularity of search is determined.The granularity of particular default selection search can be initially based on (for example, pre- lose
The number really wrapped and its spacing in distorted space), estimated distortion information can be based on for initial fragment evaluation adjustment institute
It states granularity and/or the granularity can be based on error minimization criterion on the basis of by segmentation.It can be based on wanting false watermark detection
Probability, robustness and computational complexity select the criterion.In some embodiments, it is adjusted in response to observed or expecting distortion
Fit the criterion.Based on error minimization criterion, by the predistortion watermark template at segmentation stage, (it can be locally resident in watermark
At extractor or remotely reside at database) it rearranges and reclassifies to meet error minimization criterion.In addition, can
It updates for selecting the selection rule of predistortion watermark template so that watermark extrapolator can select one group of optimum pre-distortion water
Print segmentation template.In some embodiments, error minimization criterion further triggers the new steganography that can be used for extracting subsequent watermark
The selection of key.Can carry out in such manner stego-key selection with optimize be embedded in the watermark in content host extraction it is steady
At least one of property, safety and computational efficiency.In addition, the selection of one or more stego-keys may be adapted to expect content host
Transformation.For example, the transformation can be changed with the appearance of at least one watermark of the first insertion stego-key insertion so that
At least one embedded watermark is seemed with the second insertion stego-key insertion.
Referring back to Figure 13, at 1312, the watermark extrapolation of considered segmentation is executed.In some embodiments, 1312
The operation at place includes that the symbol for the segmentation that will be considered is compared with one or more predistortion watermark templates at segmentation stage.It lifts
For example, compared with "or" (XOR) operation different by turn between fragment bit and the position of predistortion watermark template can be used to execute this.Base
Granularity in search and estimated distortion information can carry out one or more extrapolations for (for example) generating one or more error values
It attempts.Operation at 1312 further comprises selecting best matching result (for example, generating least number of errors purpose result).
At 1314, processed all segmentations are optionally determined whether.Can by by the number of segmentation processed so far be based on
Including after testing the length, existing amount distortion of tentative watermark packet and the factor of type etc. and expect include watermark packet point
The number of section is compared and makes this determination.In some embodiments, it is not necessary to handle all segmentations.In this way, can jointly across
Outer postpone for crossing one or more segmentations of the range smaller than tentative watermark (for example, duration, spatial dimension etc.) is held at once
Row subsequent operation (that is, operation at 1316 and 1318).In these embodiments, the operation at 1314 may include information whether
The determination of predetermined number segmentation in the range of processing 1 to k.In some embodiments, k is to expect to include a watermark packet
The number of segmentation.In other embodiments, k be expect include more than one watermark packet segmentation number.Wherein through processing
Segmentation is jointly across in the embodiment of the range smaller than tentative watermark, if the operation at 1318 cannot generate desired result,
It can so handle and assess additional extrapolated segmentation.
Referring back to Figure 13 and assume to perform the operation at 1314, if instruction must be handled additionally determining at 1314
Segmentation (that is, being "No" at 1314), then described group of operation 1300 returns to 1308, and the knot of the extrapolation based on previous segment
Fruit obtains the new estimation of distortion information.For example, at 1308, the estimation of distorted space can be refined (for example, based on 1312
Place generates the predistortion watermark template of best matching result).
If determining the processed all segmentations (being "Yes" at 1314) of instruction, described group of operation 1300 at 1314
Continue at 1316, wherein assigning weight to each segmentation.In an example, according to the equation provided in following sections
(9) and (10) assign weight.At 1318, the presence of the common assessment watermark based on one or more segmentations and value.Citing
For, the weight assigned at 1316 can be added together to and have itself and expression the power of the detection of the watermark of specific reliable grade
Weight threshold value is compared.It should be noted that described group of operation 1300 depicted in figure 13 can be executed with different order order.Citing comes
It says, can be immediately performed simultaneously or after the watermark extrapolation for executing each segmentation to segmentation appointment weight.It should further note
The program step of meaning, distortion tracking as described above also increases false watermark detection probability.Therefore, it is necessary to carefully to ensure false knot
It is maintained in predefined limit by property watermark detection probability.
Figure 14 is some or all of behaviour in the operation that can be used for executing Figure 13 and/or Figure 18 according to disclosed embodiment
The simplification figure of the specific components of work.The component shown in Figure 14 can reside at least partly can (for example) play, remember
In record, the content disposal plant for copying, transmitting or accessing content in other ways.Content reception device assembly 1402 is configured to connect
Receive modified embedded content.Modified embedded content can (for example) be intentionally or accidentally to be lost containing embedded watermark and suffering from
Genuine content.Modified embedded content is received to provide to multi-step searching component 1404, multi-step search groups with appropriate format
Part 1404 is configured to generate one or more the tentative watermarks and segmentation that will suffer from extrapolation.Multi-step searching component 1404 can profit
With the stego-key supplied by stego-key selection component 1410.Stego-key can be stored in stego-key storage assembly 1408
On.Watermark extrapolation device assembly 1406 is configured to execute watermark extrapolation for entire watermarked frame or on the basis of by segmentation.One
In a little embodiments, stencil-chosen component 1418 is configured to fully or reception/retrieval predistortion watermark on the basis of by segmentation
Template.For example, stencil-chosen component 1418 can retrieve one or more predistortion watermark templates from database 1416 and make it
It can be used for watermark extrapolation device assembly 1406.
Referring back to Figure 14, error minimization criterion determines that component 1412 is configured to receive and is extrapolated device assembly by watermark
1406 results generated.Error minimization criterion determines that component 1412 can further be configured to from user or allow mistake most
Smallization criterion determines that component 1412 determines that the larger system for the granularity of being searched for or the entity in device receive additional input.Mistake
Minimize criterion determine component 1412 be further configured to stego-key selection component 1410, stencil-chosen component 1418 and
Template rearranges the communication of component 1414 and conveys new search granularity, update stencil-chosen rule etc. with (for example).Template
Rearrange component 1414 be configured to it is pre- based on determining that information that component 1412 provides rearranges by error minimization criterion
It is distorted watermark template.Template rearranges component 1414 can retrieve predistortion watermark template and by predistortion water from database 1416
Die plate is stored to database 1416.Assessment device assembly 1420 can perform various watermarks evaluation operation, such as to extrapolated segmentation
Or watermark assigns weight, extrapolated watermark or its segmentation is compared with predistortion template, determine extrapolated watermark or watermark
Segmentation evaluates whether to meet wanted false watermark detection probability, reports amount distortion and the type etc. being present in content host jointly
Deng.In some embodiments, assessment device assembly 1420 is incorporated into watermark extrapolation device assembly 1406.
According to some embodiments, the essential of mistake depending on being distorted perhaps content in can polluting promotes watermark extracting.
In particular, random or burst error can be injected into the watermark being embedded in content by contents processing and/or the attack of intentional content
In.For example, can be moved by unexpected Video Camera by the burst error in the video image of Video Camera capture causes, and institute
Burst error in capture audio (for example) can be caused by the undesirable sound (such as cough of people) in background.Root
According to some embodiments, there are burst error, if entire watermark estimation vacation water is segmented rather than is directed to for watermark
Detection probability is printed, then the performance of watermark extracting system can be enhanced.In an example embodiments, to realize in watermark extracting
This enhancing and reduce processing load, substitute and be directed to entire watermark itself operation weight, to entire watermark operation and cumulative individual
The weight of segmentation.
Following instance embodiment illustrates how to estimate the weight of individual watermark segmentations.Assuming that watermark segmentation is by n
The string of position indicates, and extractor finds that particular fragments contain e bit-errors.For example, can by will include segmentation position with
Indicate that the template of the actual bit in the watermark segmentation is compared and determines error number e.Weight W associated with this segmentation
(n, e) can be indicated by following equation (9).
In equation (9), wu(n) the expection weight for indicating the non-watermarked segmentation of length n, can be according to following equation
Formula (10) indicates.
The expection weight for the non-watermarked segmentation that Figure 15 displayings are generated based on equation (10) is with the segmentation as unit of position
Length n and the exemplary plot become.To illustrate the utilization of the curve graph in equation (9) and (10) and Figure 15, examine
Consider the candidate watermark PACKET in vicious 64 of wherein 20 positions.According to equation (9), weight associated with this packet is
2.3462.However, if by watermark be divided into two 32 segmentation and if error burst leads to 16 in the first segmentation just
4 bit-errors in bit-errors and the second segmentation, then the accumulated weight of two segmentations is 4.52.Therefore, by individually examining
Two segmentations and then combined result are looked into, the presence of the watermark with higher confidence level can be established.In some embodiments, with one
Mode selects section length (segmentation limit for or more generally, being such as applicable to multidimensional watermark) different with promotion (or optimization)
It is distorted the detection of the watermark in environment.For example, if by the way that there may be in the transmitting insertion of the communication channel of burst error
Hold (or its by certain types of distortion), then the section length (for example, 4≤n≤10) in particular range may be selected.At it
In its example, some in possible section length (for example, 2≤n≤(watermark frame length/2)) or institute can be used with repetitive mode
There is section length to assess potential watermark to select the section length of the generation accumulated weight of highest.
In many cases, multiple watermarks of respective carrying difference payload are embedded into same content.Citing comes
Say, content can carrying simultaneously be intended for copy control, content recognition, the identification of content owner, allocator and/or recipient
Identification watermark message, to provide and the simultaneously operating etc. of applications.In these cases, different watermarks can have
About pay load size, robustness, safety, false detection probability, the different performance requirement of processing load and other requirements.
In these systems, each watermark message mostly independently can be embedded in and extracted.
It, can be by extracting one in embedded watermark message with coordination mode when multiple watermark messages are embedded in content
Person or both and the extraction for greatly enhancing the watermark message.In particular, in some embodiments, it can reuse and be directed to
The distortion estimation that first watermark message executes promotes the second watermark with the performance requirement different from the first watermark message to disappear
The extraction of breath.In general, with watermark message (such as the content recognition watermark and/or for distinguishing with larger pay load size
Court's watermark of the different copies of content) it compares, it can more efficiently use the watermark message (example with relatively low pay load size
Such as copy control watermark) come estimate content be distorted.For example, the watermark message with the smaller payload of band can realize more have
The use of power error correction capability.In addition, the watermark message with smaller area occupied can be less susceptible to by time/spatial variations
It distortion effect and can be embedded in greater number, improve the robustness of watermark extracting whereby.In addition, can be real with small payload
The now use of the more efficient searching algorithm based on the template matches opposite with error correcting/decoding process.Therefore, in some implementations
In example, distortion estimation is carried out based on the watermark message with smaller pay load size, and have be directed to using result
The Objective extraction of the watermark message of larger pay load size is attempted.
According to some embodiments, when there is multiple watermark messages in the content, by only detecting have with smaller
Starting has the extraction of the watermark of larger payload (and relatively low robustness) and further changes when imitating the tentative watermark of load
Into the extraction of embedded watermark.It by this method, can be due to preventing to the invalid of the larger payload watermark in non-watermarked content
It searches for and realizes and significantly calculate saving.In addition, even in markd content, still eliminate in distorted space to larger effective
Load multiple search of watermark.It must carry out less search and also reduce false watermark detection associated with larger payload watermark
Probability.The experiment carried out by inventor has disclosed, and in some situations, makes the search to larger payload watermark message and knot
The insignificant improvement of watermark extracting efficiency and false positive rate is generated by the detection association of the relatively low payload watermark message of property, but
It is significantly reduced the robustness of larger payload watermark message.That is, not enough continually triggering larger payload watermark message
Extraction, and therefore waste the sufficient opportunity of the detection of larger payload watermark.Therefore, as mentioned earlier, it is based on experiment
Property smaller payload watermark message detection trigger efficiency and processing that larger payload watermark message improves extraction operation
Load, while the reasonable robust detection result of larger payload watermark message being provided.
It should be understood that can be individually or collectively real in the device being made of various hardware and/or software module and component
Apply various embodiments of the present invention.Disclosed in description when embodiment, independent assembly is illustrated as being configured to sometimes
Execute one or more operations.However, it should be understood that can be by the two in this class component or both combination of the above together and/or often
One component may include the sub-component that do not describe.In addition, being presented described in the various figures of present application with particular order order
It operates to promote the understanding to basic concept.However, it should be understood that this generic operation can be carried out with different order order, and in addition,
Additional or less step can be used to execute various disclosed operations.
In some instances, the device described in present application may include processor, memory cell, each other with communication
The interface that mode connects, and can be between from desktop PC and/or laptop computer to consumer electronics (such as matchmaker
Body player, mobile device etc.) in the range of.For example, Figure 16 diagram illustratings can be implemented various disclosed real in it
Apply the block diagram of the device 1600 of example.Device 1600 includes at least one processor 1602 and/or controller, leads to processor 1602
At least one processor unit 1604 of letter and realize directly or by communication link 1608 indirectly with other entities, device, number
According at least one communication unit 1606 of the exchange of library and the data and information of network.Communication unit 1606 can be according to one or more
Communication protocol provides wired and/or wireless communication ability, and therefore, may include appropriate emitter/receiver antenna, circuit and
Port and it can emit and/or receive necessary coding/decoding ability to the appropriate of data and other information.Described in Figure 16
Exemplary device 1600 can be integrated into that content disposal plant, watermark be embedded and/or a part for extraction element.
In the general context for the method or program step that can be implemented on by computer program product in one embodiment
Various embodiments described herein are described, the computer program product is embodied in comprising by the computer in networked environment
In the computer-readable media of the computer executable instructions (such as program code) of execution.Computer-readable media may include can
Loading and unloading type and non-removable formula storage device, including but not limited to read-only memory (ROM), random access memory (RAM), light
Disk (CD), digital versatile disc (DVD), Blu-ray Disc etc..Therefore, the computer-readable media packet described in present application
Containing non-transitory storage media.In general, program module may include executing particular task or implement the example of particular abstract data type
Journey, program, object, component, data structure etc..Computer executable instructions, associated data structures and program module indicate to use
In the example of the program code for the step of executing method disclosed herein.Such executable instruction or associated data structures
Particular sequence indicate the example of respective action for implementing the function described in such step or program step.According to institute
Disclosing the content for being embedded with watermark of embodiment can be stored on storage media.In some embodiments, including one or more not
Can perceive this stored content of embedded watermark by the content disposal plant configured with watermark embedder (for example, software or hard
Part media player) access when can trigger watermark extracting program step, associated signal processing operation and pass through present application
Disclosed in watermark embedder and/or content disposal plant subsequent operation.
The preceding description to embodiment has been presented for the purpose of diagram and explanation.What preceding description was not intended to be exhaustive
Or the embodiment of the present invention is limited to revealed precise forms, and can be made according to above-mentioned teachings or can be from various implementations
The practice of example obtains modification and version.Embodiment discussed herein is chosen and described so as to explain various embodiments and
The principle and essence of its practical application are so that those skilled in the art can be suitable in various embodiments and covering
The present invention is utilized in the case of the various modifications of special-purpose.It can be in method, equipment, module, system and computer program product
All possible combinations in combine the feature of embodiment described herein.
Claims (44)
1. a kind of method for watermark extracting comprising：
Reception includes the watermarked frame of sync section and packet portion, and the watermarked frame is obtained from the content for being embedded with watermark；
The received sync section is compared with one or more predistortion synchronization templates；
Identify best match described in watermarked frame the received sync section predistortion synchronization template, with select with it is identified
The predistortion synchronization template one or more associated type of distortion and amount distortion indicate to be present in the content
The distortion；And
Received watermarked frame described in an at least outcome evaluation based on the comparison.
2. according to the method described in claim 1, wherein at least generating one or more described pre- mistakes partially by following operation
True synchronization template：
Generate multiple navigation contents；
The multiple navigation content is set to be embedded with the watermark including sync section；
Make the multiple embedded navigation content distortion by one or more distortions；And
Candidate predistortion synchronization template is obtained from the multiple distortion navigation content.
3. according to the method described in claim 2, it further comprises：
The multiple embedded navigation content of a subgroup is received after by transmitting media transmission, wherein the transmitting media will
One or more distortions are introduced into the multiple embedded navigation content of the subgroup；
It will be in the sync section of the multiple navigation content of the receive subgroup and the candidate predistortion synchronization template
Each is compared；And
Selection is matched with the sync section of the multiple navigation content of the received subgroup in predefined tolerance
One or more candidate predistortion synchronization templates.
4. according to the method described in claim 1, wherein described relatively generate one or both of the following terms：
The error count of each of one or more predistortion synchronization templates indicates that the described of the watermarked frame is connect
Receive the number of the mismatch symbol between each of sync section and one or more predistortion synchronization templates；And
The correlation of each of one or more predistortion synchronization templates indicates the synchronous portion of the watermarked frame
Divide and to what extent matches each of one or more described predistortion synchronization templates.
5. according to the method described in claim 1, it further comprises：
Distortion and/or distortion combination for each type is with the predistortion synchronization template described in order tissue, wherein institute of presorting
Stating order of presorting keeps the with the relatively high likelihood for indicating real content distortion first candidate predistortion synchronization template ranking high
In the second candidate predistortion synchronization template with the property of may be less likely to for indicating real content distortion.
6. according to the method described in claim 1, the wherein described assessment is generated to there are the instructions of one of the following terms：
Tentative watermark indicates the candidate water for having more than the associated false watermark detection probability for wanting false watermark detection probability
Print,
Conclusive watermark indicates the associated false watermark detection for having less than or equal to false watermark detection probability is wanted after testing
The watermark of probability,
One or more type of distortion and the amount distortion being present in the content.
7. according to the method described in claim 1, wherein：
The content includes at least first and second watermark message；
First watermark message has the payload less than second watermark message；
Assessing the received watermarked frame causes first watermark message being detected as tentative watermark；And
Second watermark message is extracted using the content distortion estimation obtained for the tentative watermark.
8. according to the method described in claim 1, it further comprises：
Tentative watermark is extracted from content host, it is more than the phase for wanting false watermark detection probability that the tentative watermark, which indicates to have,
The candidate watermark of the false watermark detection probability of association；
It is positioned at by acquisition potential at the predefined position relative to the extracted tentative watermark in the content
The symbol of watermarked frame and form one or more extrapolated watermarks；And
Determine whether the extrapolated watermark described when being evaluated jointly with the watermark tentative after testing meets wanted false watermark inspection
Survey probability.
9. according to the method described in claim 8, the symbol location of the wherein described potential watermarked frame in the content
At the time predefined of the extracted tentative watermark and/or spatial position.
10. according to the method described in claim 8, the wherein described determination includes：
Determine the number of the mismark in one or more of described extrapolated watermark；
The described of mismark is based at least partially on through the one or more into the extrapolated watermark that ascertains the number to refer to
Send weight；And
It determines whether described when combine with the watermark tentative after testing weighted extrapolated watermark meets and described wants vacation
Watermark detection probability.
11. according to the method described in claim 10, wherein determining each extrapolated water by one or more of the following terms
The number of the mismark of India and China：
The comparison of the extrapolated watermark symbol and the symbol of the extracted tentative watermark；
Execute the error-correcting code decoding of the extrapolated watermark symbol；And
The comparison of the extrapolated watermark and one or more predistortion watermark templates.
12. according to the method described in claim 8, wherein：
It forms the extrapolated watermark to be included in by the symbol for obtaining the potential watermarked frame on the basis of segmentation, wherein each segmentation
Including two or more watermark symbols；And
The determination includes：
Obtain the number of the mismark in one or more segmentations；
Be based at least partially on the number of the mismark in each of one or more described segmentations to described one or
Weight is assigned in each of multiple segmentations；
The combination assigned weight associated with one or more described segmentations is to generate one or more weighted extrapolated water
Print section；And
Determine whether the weighted extrapolated watermark section described when being combined with the watermark tentative after testing meets the institute
Want false watermark detection probability.
13. according to the method for claim 12, wherein obtaining the mistake symbol in each of one or more described segmentations
Number the number include：
By the symbol and one or more predistortions of the potential watermarked frame in each of one or more described segmentations
Watermark template is compared；And
The counting of each of one or more described segmentations is generated, described count indicates described in one or more described segmentations
The number of mismatch symbol between the symbol of potential watermarked frame and one or more described predistortion watermark templates.
14. according to the method for claim 12, wherein determining the extrapolated water based on one or more of the following terms
The number of segmentation in print：
The range of the extracted tentative watermark；
The amount distortion being present in the content；And
The type of distortion being present in the content.
15. according to the method for claim 12, being wherein at least based in part on not the adding equal to particular fragments in length
The weight of particular fragments described in the expection weight assignment of watermark segmentation.
16. according to the method described in claim 8, wherein described the wanted false watermark detection probability corresponds to the inspection of conclusive watermark
It surveys.
17. according to the method described in claim 8, it further comprises that report is present in the distortion in the embedded content host
Or the estimation of the amount and type of distortion combination.
18. according to the method for claim 12, being wherein at least based in part on one or more of the following terms to determine often
The range of one segmentation：
It is expected that being present in the amount distortion in the content；And
It is expected that being present in the type of distortion in the content.
19. according to the method described in claim 1, it further comprises
From the multiple tentative watermarks of the contents extraction, it is more than to want false watermark detection probability that each tentative watermark, which indicates to have,
Associated false watermark detection probability candidate watermark；
Using in the extracted tentative watermark both at least obtain with one be present in the embedded content host or
Multiple associated estimated distortion informations of distortion；
Based on the estimated distortion information, one or more predistortion watermark templates are obtained；And
At least one of described extracted tentative watermark is reappraised using one or more described predistortion watermark templates.
20. according to the method for claim 19, wherein described reappraise leads to one or more of the following terms：
Detecting compared with the false watermark detection probability for being associated with each of described extracted tentative watermark has through changing
Into the watermark of false watermark detection probability；
It detects with conclusive less than or equal to the associated false watermark detection probability of the wanted false watermark detection probability
Watermark；And
Get the improved estimation for the distortion being present in the embedded content host.
21. according to the method for claim 20, wherein using two of described at least two extracted tentative watermarks or
More than two sync sections obtain the estimated distortion information.
22. according to the method described in claim 1, it further comprises：
From the tentative watermark of the contents extraction, it is more than the phase for wanting false watermark detection probability that the tentative watermark, which indicates to have,
The candidate watermark of the false watermark detection probability of association；
Certain distortion or the distortion that instruction is present in the embedded content host are obtained using the extracted tentative watermark
The estimated distortion information of the coarse value of combination；
Selection corresponds to one or more supplement predistortion watermark templates of one or more distortion values near the coarse value；And
The extracted tentative watermark is reappraised using one or more described selected supplement predistortion watermark templates.
23. according to the method for claim 22, wherein obtaining the estimated distortion information and including：
With coarse-grained selection corresponding to multiple rough predistortion watermarks of certain distortion or the range of the distortion value of distortion combination
Template；
Each of the extracted tentative watermark and the multiple rough predistortion watermark template are compared；And it is logical
It crosses the first rough predistortion watermark template of extracted tentative watermark described in identification best match and obtains the estimated mistake
True information.
24. according to the method for claim 22, wherein described reappraise causes to detect and wanted with being less than or equal to
The conclusive watermark of the associated false watermark detection probability of false watermark detection probability.
25. according to the method for claim 22, wherein it is described reappraise including：
One or more described supplement predistortion watermark templates are compared with the extracted tentative watermark；
It generates and indicates the corresponding supplement predistortion watermark template with each more associated error count, the error count
To what extent match the extracted tentative watermark；And
The supplement predistortion watermark template of the selection corresponding to minimum produced error count.
26. a kind of device for watermark extracting comprising：
Receiver, be configured to receive include sync section and packet portion watermarked frame, the watermarked frame is from being embedded with water
The content of print obtains；
Comparator is configured to that the received sync section is compared and is known with one or more predistortion synchronization templates
The predistortion synchronization template of the received sync section of watermarked frame described in other best match indicates to be present in described with selection
One or more type of distortion and amount distortion of the distortion in content；And
Evaluator is configured to received watermarked frame described in an at least outcome evaluation for device based on the comparison.
27. device according to claim 26, wherein the comparator be configured to generate one of the following terms or
The two：
The error count of each of one or more predistortion synchronization templates indicates that the described of the watermarked frame is connect
Receive the number of the mismatch symbol between each of sync section and one or more predistortion synchronization templates；And
The correlation of each of one or more predistortion synchronization templates indicates the synchronous portion of the watermarked frame
Divide and to what extent matches each of one or more described predistortion synchronization templates.
28. device according to claim 26, wherein one or more described type of distortion and amount distortion with it is described identified
Predistortion synchronization template is associated.
29. device according to claim 26, wherein the evaluator be configured to generate in the following terms one or
The existing instruction of more persons：
Tentative watermark indicates the candidate water for having more than the associated false watermark detection probability for wanting false watermark detection probability
Print,
Conclusive watermark indicates the associated false watermark detection for having less than or equal to false watermark detection probability is wanted after testing
The watermark of probability, and
One or more type of distortion and the amount distortion being present in the content.
30. device according to claim 26, wherein：
The content includes at least first and second watermark message；
First watermark message has the payload less than second watermark message；And
The evaluator is configured to generate instruction and content mistake to first watermark message to be detected as to tentative watermark
True estimation, wherein extracting second watermark message using the content distortion estimation.
31. device according to claim 26, further comprises
Extractor is configured to from the tentative watermark of the contents extraction, and it is more than to be wanted that the tentative watermark, which indicates to have,
The candidate watermark of the associated false watermark detection probability of false watermark detection probability；And
Watermark extrapolator is configured to acquisition and is positioned in the content relative to the extracted tentative watermark
Predefined position at potential watermarked frame symbol and form one or more extrapolated watermarks；Wherein
The evaluator is further configured to determine described extrapolated when being evaluated jointly with the watermark tentative after testing
Whether watermark meets wanted false watermark detection probability.
32. device according to claim 31, wherein the watermark extrapolator is configured to out of described content opposite
Position at the time predefined of the extracted tentative watermark and/or spatial position obtains the institute of the potential watermarked frame
State symbol.
33. device according to claim 31, wherein the evaluator is configured to：
Determine the number of the mismark in one or more of described extrapolated watermark；
The described of mismark is based at least partially on through the one or more into the extrapolated watermark that ascertains the number to refer to
Send weight；And
It determines whether described when combine with the watermark tentative after testing weighted extrapolated watermark meets and described wants vacation
Watermark detection probability.
34. device according to claim 33, wherein determining each extrapolated water by one or more of the following terms
The number of the mismark of India and China：
The comparison of the extrapolated watermark symbol and the symbol of the extracted tentative watermark；
Execute the error-correcting code decoding of the extrapolated watermark symbol；And
The comparison of the extrapolated watermark and one or more predistortion watermark templates.
35. device according to claim 31, wherein：
The watermark extrapolator is configured to obtain the symbol of the potential watermarked frame on the basis of by segmentation at least partly
Number and form the extrapolated watermark, wherein it is each segmentation include two or more watermark symbols；And
The evaluator is configured to：
Obtain the number of the mismark in one or more segmentations；
Be based at least partially on the number of the mismark in each of one or more described segmentations to described one or
Weight is assigned in each of multiple segmentations；
The combination assigned weight associated with one or more described segmentations is to generate one or more weighted extrapolated water
Print section；And
Determine whether the weighted extrapolated watermark section described when being combined with the watermark tentative after testing meets the institute
Want false watermark detection probability.
36. device according to claim 35, wherein the evaluator is configured at least partially through following operation
Obtain the number of the mismark in each of one or more described segmentations：
By the symbol and one or more predistortions of the potential watermarked frame in each of one or more described segmentations
Watermark template is compared；And
The counting of each of one or more described segmentations is generated, described count indicates described in one or more described segmentations
The number of mismatch symbol between the symbol of potential watermarked frame and one or more described predistortion watermark templates.
37. device according to claim 31, wherein the evaluator is configured to be based at least partially in length
Equal to the expection weight of the non-watermarked segmentation of particular fragments the weight is assigned to the particular fragments.
38. device according to claim 31 is present in wherein the evaluator is configured to report in the content
The estimation of the amount and type of distortion or distortion combination.
39. device according to claim 26, further comprises
Extractor is configured to from the multiple tentative watermarks of the contents extraction, and each tentative watermark, which indicates to have, is more than
Want the candidate watermark of the associated false watermark detection probability of false watermark detection probability；And
Distortion estimator is configured to be obtained using at least the two in the extracted tentative watermark and be present in insertion
One or more in content host are distorted associated estimated distortion information and obtain one or more predistortion watermark templates；Its
In
Evaluator is configured to reappraise the extracted tentative watermark using one or more described predistortion watermark templates
At least one of.
40. device according to claim 39, wherein the distortion estimator is configured to determine the estimated distortion
Whether information corresponds to over the amount distortion of certain distortion threshold value, and only corresponds in the estimated distortion information and be no more than institute
One or more described predistortion watermark templates are obtained in the case of stating the amount distortion of certain distortion threshold value.
41. device according to claim 39, wherein the distortion estimator is configured to pass through using described at least two
Two or more sync sections for extracting tentative watermark obtain the estimated distortion information.
42. device according to claim 26, further comprises
Extractor is configured to from the tentative watermark of the contents extraction, and it is more than to be wanted that the tentative watermark, which indicates to have,
The candidate watermark of the associated false watermark detection probability of false watermark detection probability；
Distortion estimator is configured to obtain the spy that instruction is present in the content using the extracted tentative watermark
The estimated distortion information of the coarse value of fixed distortion or distortion combination, and select to correspond to one or more near the coarse value
One or more supplement predistortion watermark templates of a distortion value；Wherein
The evaluator is further configured to reappraise using one or more described selected supplement predistortion watermark templates
The extracted tentative watermark.
43. device according to claim 42, wherein the distortion estimator is configured to：
With coarse-grained selection corresponding to multiple rough predistortion watermarks of certain distortion or the range of the distortion value of distortion combination
Template；
Each of the extracted tentative watermark and the multiple rough predistortion watermark template are compared；And it is logical
It crosses the first rough predistortion watermark template of extracted tentative watermark described in identification best match and obtains the estimated mistake
True information.
44. device according to claim 42, wherein the evaluator is configured to：
One or more described supplement predistortion watermark templates are compared with the extracted tentative watermark；
It generates and indicates the corresponding supplement predistortion watermark template with each more associated error count, the error count
To what extent match the extracted tentative watermark；And
The supplement predistortion watermark template of the selection corresponding to minimum produced error count.
Applications Claiming Priority (9)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| US13/288,952 | 2011-11-03 | ||
| US13/288,945 | 2011-11-03 | ||
| US13/288,937 | 2011-11-03 | ||
| US13/288,937 US8615104B2 (en) | 2011-11-03 | 2011-11-03 | Watermark extraction based on tentative watermarks |
| US13/288,958 | 2011-11-03 | ||
| US13/288,958 US8533481B2 (en) | 2011-11-03 | 2011-11-03 | Extraction of embedded watermarks from a host content based on extrapolation techniques |
| US13/288,945 US8923548B2 (en) | 2011-11-03 | 2011-11-03 | Extraction of embedded watermarks from a host content using a plurality of tentative watermarks |
| US13/288,952 US8682026B2 (en) | 2011-11-03 | 2011-11-03 | Efficient extraction of embedded watermarks in the presence of host content distortions |
| PCT/US2012/063431 WO2013067439A1 (en) | 2011-11-03 | 2012-11-02 | Watermark extraction based on tentative watermarks |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103975605A CN103975605A (en) | 2014-08-06 |
| CN103975605B true CN103975605B (en) | 2018-09-07 |
Family
ID=48192861
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201280059704.5A Active CN103975605B (en) | 2011-11-03 | 2012-11-02 | Watermark extracting based on tentative watermark |
Country Status (5)
| Country | Link |
|---|---|
| EP (1) | EP2774385A4 (en) |
| KR (1) | KR20140097306A (en) |
| CN (1) | CN103975605B (en) |
| HK (1) | HK1197126A1 (en) |
| WO (1) | WO2013067439A1 (en) |
Families Citing this family (33)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7644282B2 (en) | 1998-05-28 | 2010-01-05 | Verance Corporation | Pre-processed information embedding system |
| US6737957B1 (en) | 2000-02-16 | 2004-05-18 | Verance Corporation | Remote control signaling using audio watermarks |
| CA2499967A1 (en) | 2002-10-15 | 2004-04-29 | Verance Corporation | Media monitoring, management and information system |
| US20060239501A1 (en) | 2005-04-26 | 2006-10-26 | Verance Corporation | Security enhancements of digital watermarks for multi-media content |
| US8020004B2 (en) | 2005-07-01 | 2011-09-13 | Verance Corporation | Forensic marking using a common customization function |
| US8781967B2 (en) | 2005-07-07 | 2014-07-15 | Verance Corporation | Watermarking in an encrypted domain |
| US8259938B2 (en) | 2008-06-24 | 2012-09-04 | Verance Corporation | Efficient and secure forensic marking in compressed |
| US10664940B2 (en) | 2010-03-05 | 2020-05-26 | Digimarc Corporation | Signal encoding to reduce perceptibility of changes over time |
| US8971567B2 (en) | 2010-03-05 | 2015-03-03 | Digimarc Corporation | Reducing watermark perceptibility and extending detection distortion tolerances |
| US8477990B2 (en) | 2010-03-05 | 2013-07-02 | Digimarc Corporation | Reducing watermark perceptibility and extending detection distortion tolerances |
| US8838977B2 (en) | 2010-09-16 | 2014-09-16 | Verance Corporation | Watermark extraction and content screening in a networked environment |
| US8682026B2 (en) | 2011-11-03 | 2014-03-25 | Verance Corporation | Efficient extraction of embedded watermarks in the presence of host content distortions |
| US8615104B2 (en) | 2011-11-03 | 2013-12-24 | Verance Corporation | Watermark extraction based on tentative watermarks |
| US8923548B2 (en) | 2011-11-03 | 2014-12-30 | Verance Corporation | Extraction of embedded watermarks from a host content using a plurality of tentative watermarks |
| US8533481B2 (en) | 2011-11-03 | 2013-09-10 | Verance Corporation | Extraction of embedded watermarks from a host content based on extrapolation techniques |
| US8745403B2 (en) | 2011-11-23 | 2014-06-03 | Verance Corporation | Enhanced content management based on watermark extraction records |
| US9547753B2 (en) | 2011-12-13 | 2017-01-17 | Verance Corporation | Coordinated watermarking |
| US9323902B2 (en) | 2011-12-13 | 2016-04-26 | Verance Corporation | Conditional access using embedded watermarks |
| US9571606B2 (en) | 2012-08-31 | 2017-02-14 | Verance Corporation | Social media viewing system |
| US9106964B2 (en) | 2012-09-13 | 2015-08-11 | Verance Corporation | Enhanced content distribution using advertisements |
| US8869222B2 (en) | 2012-09-13 | 2014-10-21 | Verance Corporation | Second screen content |
| US9262794B2 (en) | 2013-03-14 | 2016-02-16 | Verance Corporation | Transactional video marking system |
| US9251549B2 (en) | 2013-07-23 | 2016-02-02 | Verance Corporation | Watermark extractor enhancements based on payload ranking |
| KR102113503B1 (en) | 2013-08-26 | 2020-06-05 | 삼성전자주식회사 | Electronic apparatus and method for providing contents in the electronic apparatus |
| US9208334B2 (en) | 2013-10-25 | 2015-12-08 | Verance Corporation | Content management using multiple abstraction layers |
| EP3117626A4 (en) | 2014-03-13 | 2017-10-25 | Verance Corporation | Interactive content acquisition using embedded codes |
| US10504200B2 (en) | 2014-03-13 | 2019-12-10 | Verance Corporation | Metadata acquisition using embedded watermarks |
| WO2016028936A1 (en) * | 2014-08-20 | 2016-02-25 | Verance Corporation | Watermark detection using a multiplicity of predicted patterns |
| EP3225034A4 (en) | 2014-11-25 | 2018-05-02 | Verance Corporation | Enhanced metadata and content delivery using watermarks |
| US9602891B2 (en) | 2014-12-18 | 2017-03-21 | Verance Corporation | Service signaling recovery for multimedia content using embedded watermarks |
| CN106921728A (en) | 2016-08-31 | 2017-07-04 | 阿里巴巴集团控股有限公司 | A kind of method for positioning user, information-pushing method and relevant device |
| CN109151483A (en) * | 2018-09-10 | 2019-01-04 | 五邑大学 | System based on image and video metadata reversible insertion and extraction |
| CN110827189B (en) * | 2019-11-01 | 2023-12-05 | 山东浪潮科学研究院有限公司 | Watermark removing method and system for digital image or video |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101751927A (en) * | 2008-10-10 | 2010-06-23 | 汤姆森许可贸易公司 | Method and apparatus for regaining watermark data in an original signal |
| WO2010135687A1 (en) * | 2009-05-21 | 2010-11-25 | Digimarc Corporation | Combined watermarking and fingerprinting |
Family Cites Families (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6516079B1 (en) * | 2000-02-14 | 2003-02-04 | Digimarc Corporation | Digital watermark screening and detecting strategies |
| US7366908B2 (en) * | 1996-08-30 | 2008-04-29 | Digimarc Corporation | Digital watermarking with content dependent keys and autocorrelation properties for synchronization |
| CA2269651A1 (en) * | 1998-05-12 | 1999-11-12 | Lucent Technologies, Inc. | Transform domain image watermarking method and system |
| US6483927B2 (en) * | 2000-12-18 | 2002-11-19 | Digimarc Corporation | Synchronizing readers of hidden auxiliary data in quantization-based data hiding schemes |
| US7840005B2 (en) * | 2002-01-22 | 2010-11-23 | Digimarc Corporation | Synchronization of media signals |
| US20060239501A1 (en) * | 2005-04-26 | 2006-10-26 | Verance Corporation | Security enhancements of digital watermarks for multi-media content |
| US7668334B2 (en) * | 2004-07-02 | 2010-02-23 | Digimarc Corp | Conditioning imagery to better receive steganographic encoding |
| JP4034776B2 (en) * | 2004-11-12 | 2008-01-16 | 株式会社東芝 | Digital watermark detection apparatus, digital watermark detection method, and program |
| WO2009031082A1 (en) * | 2007-09-03 | 2009-03-12 | Koninklijke Philips Electronics N.V. | Apparatus and methods for transferring digital content |
-
2012
- 2012-11-02 WO PCT/US2012/063431 patent/WO2013067439A1/en active Application Filing
- 2012-11-02 KR KR1020147015220A patent/KR20140097306A/en not_active Application Discontinuation
- 2012-11-02 EP EP12846172.0A patent/EP2774385A4/en not_active Ceased
- 2012-11-02 CN CN201280059704.5A patent/CN103975605B/en active Active
-
2014
- 2014-10-20 HK HK14110429A patent/HK1197126A1/en unknown
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101751927A (en) * | 2008-10-10 | 2010-06-23 | 汤姆森许可贸易公司 | Method and apparatus for regaining watermark data in an original signal |
| WO2010135687A1 (en) * | 2009-05-21 | 2010-11-25 | Digimarc Corporation | Combined watermarking and fingerprinting |
Also Published As
| Publication number | Publication date |
|---|---|
| KR20140097306A (en) | 2014-08-06 |
| HK1197126A1 (en) | 2015-01-02 |
| WO2013067439A1 (en) | 2013-05-10 |
| EP2774385A1 (en) | 2014-09-10 |
| CN103975605A (en) | 2014-08-06 |
| EP2774385A4 (en) | 2015-05-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103975605B (en) | Watermark extracting based on tentative watermark | |
| US9704211B2 (en) | Signal continuity assessment using embedded watermarks | |
| US8615104B2 (en) | Watermark extraction based on tentative watermarks | |
| US8923548B2 (en) | Extraction of embedded watermarks from a host content using a plurality of tentative watermarks | |
| US8682026B2 (en) | Efficient extraction of embedded watermarks in the presence of host content distortions | |
| US8533481B2 (en) | Extraction of embedded watermarks from a host content based on extrapolation techniques | |
| Bhattacharyya | A survey of steganography and steganalysis technique in image, text, audio and video as cover carrier | |
| US7616776B2 (en) | Methods and apparatus for enhancing the robustness of watermark extraction from digital host content | |
| AU2007327388B2 (en) | Video fingerprinting | |
| US20060239501A1 (en) | Security enhancements of digital watermarks for multi-media content | |
| US20070165851A1 (en) | Watermark detection | |
| CN1988669B (en) | Digital marking structure and verifying method in stream medium monitoring and broadcasting | |
| WO2002032031A1 (en) | Watermarks carrying content dependent signal metrics for detecting and characterizing signal alteration | |
| WO2006116270A2 (en) | Security enhancements of digital watermarks for multi-media content | |
| US8107668B2 (en) | Digital differential watermark and method | |
| GB2585773A (en) | Methods and apparatus to identify media using hash keys | |
| US20070165852A1 (en) | Watermark detection | |
| Xiang et al. | Robust and reversible audio watermarking by modifying statistical features in time domain | |
| CN101065769B (en) | Method and device for reading digital tattooing | |
| Noga | A Visual Data Hash Method | |
| CN1332354C (en) | Data content identification | |
| US20070160261A1 (en) | Watermark detection by correlation shape analysis | |
| Bhat et al. | JHUF-5 steganalyzer: Huffman based steganalytic features for reliable detection of YASS in JPEG images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1197126 Country of ref document: HK |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: WD Ref document number: 1197126 Country of ref document: HK |