Classifications

• • G—PHYSICS
  • G11—INFORMATION STORAGE
  • G11B—INFORMATION STORAGE BASED ON RELATIVE MOVEMENT BETWEEN RECORD CARRIER AND TRANSDUCER
  • G11B20/00—Signal processing not specific to the method of recording or reproducing; Circuits therefor
  • G11B20/10—Digital recording or reproducing
• • 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
  • G06T1/0064—Geometric transfor invariant watermarking, e.g. affine transform invariant
• • 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
  • G06T1/0028—Adaptive watermarking, e.g. Human Visual System [HVS]-based watermarking
• • 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
• • 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
  • H04N1/32149—Methods relating to embedding, encoding, decoding, detection or retrieval operations
  • H04N1/32203—Spatial or amplitude domain methods
  • H04N1/32229—Spatial or amplitude domain methods with selective or adaptive application of the additional information, e.g. in selected regions of the image
• • 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
  • H04N1/32149—Methods relating to embedding, encoding, decoding, detection or retrieval operations
  • H04N1/32288—Multiple embedding, e.g. cocktail embedding, or redundant embedding, e.g. repeating the additional information at a plurality of locations in the image
  • H04N1/32299—Multiple embedding, e.g. cocktail embedding, or redundant embedding, e.g. repeating the additional information at a plurality of locations in the image using more than one embedding method
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N2201/00—Indexing scheme relating to scanning, transmission or reproduction of documents or the like, and to details thereof
  • H04N2201/32—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
  • H04N2201/3201—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title
  • H04N2201/3225—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title of data relating to an image, a page or a document
  • H04N2201/3233—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title of data relating to an image, a page or a document of authentication information, e.g. digital signature, watermark
• • H—ELECTRICITY
  • H04—ELECTRIC COMMUNICATION TECHNIQUE
  • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
  • H04N2201/00—Indexing scheme relating to scanning, transmission or reproduction of documents or the like, and to details thereof
  • H04N2201/32—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
  • H04N2201/3201—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title
  • H04N2201/3269—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title of machine readable codes or marks, e.g. bar codes or glyphs
  • H04N2201/327—Display, printing, storage or transmission of additional information, e.g. ID code, date and time or title of machine readable codes or marks, e.g. bar codes or glyphs which are undetectable to the naked eye, e.g. embedded codes

Definitions

• the invention relates to multimedia signal processing, and specifically, steganography, digital watermarking and data hiding.
• Digital watermarking is a process for modifying physical or electronic media to embed a machine-readable code into the media.
• the media may be modified such that the embedded code is imperceptible or nearly imperceptible to the user, yet may be detected through an automated detection process.
• digital watermarking is applied to media signals such as images, audio signals, and video signals. However, it may also be applied to other types of media objects, including documents (e.g., through line, word or character shifting), software, multi-dimensional graphics models, and surface textures of objects.
• Digital watermarking systems typically have two primary components: an encoder that embeds the watermark in a host media signal, and a decoder that detects and reads the embedded watermark from a signal suspected of containing a watermark (a suspect signal).
• the encoder embeds a watermark by altering the host media signal.
• the reading component analyzes a suspect signal to detect whether a watermark is present. In applications where the watermark encodes information, the reader extracts this information from the detected watermark.
• a robust watermark refers to a watermark that is designed to survive typical and even malicious processing of the watermarked signal that distorts the watermarked signal and makes it more difficult to reliably detect and read the watermark.
• a fragile watermark refers to a watermark where the watermark degrades in response to certain forms of processing like printing copying, scanning, compression, etc.
• Fragile watermarks are typically used in authentication application to detect tampering of a signal.
• Semi-fragile watermarks combine the concepts of fragile and robust watermarks. These types of watermarks are designed to survive certain types of processing like compression, yet detect tampering like cropping or swapping of signals. Fragile and semi-fragile watermarks may be used to trigger certain actions or control usage of the watermarked content when degradation of the fragile watermark is detected.
• the music track may be rendered using high fidelity audio equipment, or lower quality equipment, giving rise to different perceptual quality constraints, hi particular, lower quality rendering enables the watermark to be embedded more robustly because perceptibility constraints on the watermark are less stringent.
• video signals like movies, television programming, advertisements, etc.
• an image may undergo transformations, such as compression, color conversion, halftoning, etc. before it is finally printed or rendered.
• transformations such as compression, color conversion, halftoning, etc.
• Such art imagery may include a collection of a raster images that are combined to form a final image.
• the graphic artist creates a piece of graphic art for a customer, typically including a collection of constituent images in different formats.
• Some of the images may be line art, vector graphics, color halftone or color multi-level per pixel images (in color formats like RGB, CMYK or YUN).
• the entire image product is described in a job ticket that encapsulates the rendering functions to control the assembly of the constituent images and the printing process.
• the customer may want to apply a watermark to the final image product for a variety of applications, such as inserting a customer identifier for tracking purposes, linking the image to the customer's web site, etc.
• One problem occurs with the content flow and timing of adding the watermark flow.
• Another problem occurs with adding watermarks to vector graphics.
• the stage at which the watermark message payload and embedding parameters are defined may not always be the appropriate stage to embed the watermark in the host signal.
• One place to embed the message payload of the watermark into the graphic art is in the raster interface processing (RIP) stage. In this stage, the constituent images are assembled and converted to a particular halftone image format compatible with the printer.
• RIP raster interface processing
• the halftone image format includes one or more color planes of pixel elements that specify the presence or absence of ink at corresponding pixel locations.
• the RIP stage usually occurs at the Pre-Press house or Printer, and requires the person with the most critical eye for color. In addition, this stage, by definition, results in a complete raster image.
• the watermark can be defined for vector graphics (or line-art), but is ultimately embedded in a raster image when printed with common modern equipment. The customer doesn't usually interact with the Pre-Press house or Printer, except to possibly proof the image. In addition, these locations are under terrible time and cost constraints and do not want to deal with inefficient and costly customer interactions.
• the graphic artist must rasterize the image. This causes two problems. First, the graphic artist must now deliver a file consisting of a large number of bits (i.e. size). Second, the graphic artist is not the best person to deal with the color management required to produce a quality image.
• the difficulty is that the customer is already working with the graphic artist and wishes to define the contents of the watermark, but the watermark is ultimately embedded in the rasterized image in the Pre-Press house or Printer.
• the image file is a vector graphic, whether rendered for printing as described above, or distributed electronically such as on the web
• a participant such as the owner, may want to watermark the vector graphic.
• the participant wants that watermark to be embedded in the rendered image whenever the vector file is rendered, such as on a computer screen, possible within a wed browser or printer. This allows illegitimate copies, such as copies made with a print screen function, to be identified.
• a method for controlling watermark embedding in a media object through the use of a watermark embedding command is described below.
• the method includes a watermark embedding command among a set of one or more rendering commands that specify how the media object is to be rendered.
• certain media signal formats like PCL, PDF, or postscript for images, MIDI and structured audio for audio signals, and MPEG-4 and MPEG-7 for audio and video signals, include descriptors that control how a particular media signal is to be rendered.
• the watermark embedding command includes a combination of the following items: an identifier used to link to customer or related content information, the customer's web site or store, the intensity at which to embed the watermark, areas not to embed, batch processing options, printing preferences for images, watermarking embedding methods to use on different media types, formats, or different parts of the media object, and desired rendering quality.
• the watermark embedding command enables the customer or creator to specify watermark message payload and embedding parameters and preferences, and enables the rendering device to embed the watermark appropriately for a particular rendering process.
• the customer can preview the watermarked content on the graphic artist's monitor or inexpensive printer, which rasterizes the image for display, embeds the watermark in response to the command, and renders the watermarked image.
• the Pre-Press house or Printer can add and modify the watermark without interacting with the customer, thereby saving time and money.
• the watermark embedding command includes the message payload to be embedded and rules or links to how to embed these bits.
• the watermark function is implemented according to the desired embedding method when the graphic art is rendered, such as on the screen, printed proofs or final printing plates. This method is extended to other types of media objects, including audio or music tracks, video sequences, etc.
• watermarks can be embedded in rendering description content, such as vector graphics, MIDI, and structured MPEG audio and video.
• rendering description content such as vector graphics, MIDI, and structured MPEG audio and video.
• watermarks can be embedded at a time and location separate from where and when the watermark and content is rendered. This reduces costs by allowing proper interaction between the content owner and creators, who have different responsibilities and skills.
• the invention further provides methods and related systems, devices and software for transmarking media signals.
• Transmarking relates to converting auxiliary data embedded in a media signal from one digital watermark format to another. It is used in processes that transform the media signal, such as compression, broadcast, editing, rendering, etc., to change the characteristics of the embedded watermark so that the watermark has improved robustness or perceptibility characteristics for its new environment.
• transmarking can be extended to. cases where out-of-band data file the header or footer of a media file, or other metadata provided with the media file is transmarked into a watermark or is derived from a watermark. Thus, the watermarks appear to be robust to all transformations.
• One aspect of the invention is a method of transmarking a media signal previously embedded with a first digital watermark using a first digital watermark embedding method. This transmarking method detects the first digital watermark in the 1 ⁇
• Another aspect of the invention is another method of transmarking a media signal. This method detects the first digital watermark in the media signal, converts the media signal to a different format, and embeds message information from the first digital watermark into a second digital watermark in the converted media signal.
• the second digital watermark is adapted to robustness or perceptibility parameters associated with the new format.
• Figure la Diagram for embedding a feature-based watermark to ease searching a large image for a small watermarked area.
• Figure la Diagram for retrieving a feature-based watermark to ease searching a large image for a small watermarked area.
• Figure 2 This figures shows a pseudo-random noise array that can be used to determine the scaling and rotation of an image via autocorrelation-
• Figure 3 This figure demonstrates the art of slowing the transition between embedding auxiliary 0's and l's.
• Figure 4a This figure shows the grid used to embed an autocorrelation-based watermark.
• Figure 4b This figure demonstrates how to skip embedding data in some blocks to find the orientation of an autocorrelation-based watermarked image.
• X's represent watermarked blocks; thus, blocks without Xs are not watermarked.
• FIG. 5 is a diagram illustrating a transmarking process where a first digital watermark in a media signal is transmarked into a second digital watermark in the media signal.
• Fig. 6 is a diagram illustrating a watermark embedding function and rendering description file.
• Fig. 7 is a diagram illustrating a process for embedding watermarks in media objects using watermark embedding commands.
• noise reduction techniques such as Weiner filtering or spectral subtraction
• This noise represents the sum of all the watermark layers.
• This noise can be re-scaled and embedded in other images such that they impersonate the original image.
• the first suggestion is that a different group of global PN sequences could be used in this new version than with earlier versions.
• the second suggestion is to add a layer of noise defining the version.
• the third is to use different spacing or position in the grid used to determine scaling and rotation of the embedded data. h addition, when trying to find a watermarked area of a large image, feature- based watermarking is advantageous. As well known, searching the whole image for the small watermark is slow.
• the process is to use a feature of the picture, such as the peak of the derivate, to embed a space-limited data, such as a local PN sequence, that provides information about the location of the picture's corner and the scaling.
• a feature of the picture such as the peak of the derivate
• a space-limited data such as a local PN sequence
• the whole block structure of the watermark such as P by Q pixel areas for embedding (e.g., P and Q are preferably the same and multiples of two)
• P and Q are preferably the same and multiples of two
• This embedded local-feature PN sequence will intrinsically inform the decoder that the feature is part of the picture by its existence.
• This local-feature PN sequence should also include a grid layer so that once it is found the scaling coefficient can be determined. Instead of a grid layer, the watermark decoder could employ the autocorrelation and cross-correlation scaling methods for compensating for scaling and rotation discussed in this document.
• This local-feature PN sequence should also include a few layers to provide where the lower-left (or other corner) of the picture is located. For example, two layers could inform the decoder which quadrant the feature was located. With the scaling and quadrant information, finding the global PN sequence, which carries the message, will be easier and faster. Scaling This method is illustrated through the following two embodiments.
• auto-correlation of an image and self-similar noise layer is used to determine the image's scaling and rotation.
• Figure 2 shows the self-similar noise array layer that can be embedded within an image, or sequentially within audio, to determine the time scaling and rotation, for 2D images only.
• the PN variable is, for example, a 10x10 array of noise, where each PN sequence is identical.
• the 0 variable is, for example, a 10x10 array of zeros.
• the second embodiment includes estimating the image transformation by cross- correlating an original PN noise layer with an image which previously had the PN noise layer added and has been modified. Assuming the image has only been linearly transformed, such as by rotation or scaling, the PN noise layer is white, and the PN noise layer is orthogonal to the image, the result of the cross-correlation is the impulse function of the transformation. This impulse function can be used to improve recovery of the watermark. Finally, concepts from spectral estimation can be applied to increase the accuracy of the estimation since the assumptions are usually only partially true. Transitions audio applications, the transition between embedding a 0 and 1 bit of auxiliary information occur by immediately changing the phase of the PN sequence, i.e. switch from multiplying by -1 and 1 and visa-versa.
• the transition between 0 and 1 bit of auxiliary information should have a transition period where the phase of the noise sequence is slowly changed. Although this will lower the embedded bit rate, it should decrease the perception of the watermark. .
• the transition period length could be from 1 to 1 several hundreds of a milliseconds.
• a utocorrelation Watermarks In general, a problem with reading watermarks via digital cameras, such as CCD or CMOS based cameras, is that the cameras integrate over space to get a color value. This integration is used since each camera receiving-element, such as a CCD, takes up space and a RGB or CMYK color grid is used. This integration does not degrade the picture quality since real- world pictures have data points that are correlated to its neighbor.
• white noise-based watermarks where the value changes every pixel, the camera not only removes the noise but also produces incorrect data since every pixel is independent in white noise.
• a current solution is to use noise where the value changes in blocks of pixels.
• An alternative solution uses an autocorrelation based watermark, defined as taking a copy of the image, lowering its level, and placing it slightly offset from the original image. Either the offset value or copy level can be used to transfer 0's and l's. For example, up and left shifts represent l's, whereas down and right shifts represent 0's.
• the watermark is retrieved by calculating the autocorrelation function and finding the offset value of the peak, which is provided by the embedded low-level and shifted copy of the image.
• This type of watermark survives integration since, as with real-world data, the neighboring will be related to each other and survive the camera's integration.
• This watermark will also be invisible since it intrinsically places the data where it can be hidden. In other words, an offset copy of the image is already prepared to be hidden in the image.
• the block size is a balance between the number of embedded bits versus amount of noise embedded to retrieve one bit.
• the smaller the block size more information is lost in edge patterns.
• the shift used in embedding the low level copy of the image should be minimal so as not to degrade quality, such as blurring the edges. It appears desirable to have the shift larger than a single cameral pixel element, i.e. one CCD grid.
• the dynamic media scrambling process is to re-encrypt or re-scramble the content using a new technique or key each time the content is rendered (assuming the device is re-writeable), or using some other interval, possibly regular or not. This technique is invisible to the consumer. In addition, when keys are found to be broken, the removal of that key from the system will happen over time without any inconvenience to the legitimate consumer.
• the encryption routine When content is rendered on the user's machine, the encryption routine decrypts the content using the current key. Then a new key is created, and the encryption routine encrypts the content for storage on the user's machine. To generate a new key, the encryption routine changes part or. all of the previous key. In particular, part of the key may be based on something unique to the machine or software running on the machine, such as a processor ID, or date of the trash can or recycle bin in the operating system. The remainder of the key changes with each rendering according to a random or pseudorandom function. When the new key is created, it is stored in a secure, encrypted and tamper resistant file on the user's machine. This key is used the next time the content is rendered. The key not be changed each time the content is rendered.
• the key may be changed based on some external event trigger, such as the receipt of a new key from a local or remote key management system, or the receipt of a key update flag from a key management system or registry database that instructs the encryption routine on the user's device to update the key the next time the content is rendered.
• some external event trigger such as the receipt of a new key from a local or remote key management system, or the receipt of a key update flag from a key management system or registry database that instructs the encryption routine on the user's device to update the key the next time the content is rendered.
• This process of key updating enables encryption keys to be updated over time, and eventually move old or broken keys out of the system.
• Transmarking many applications a digital watermark signal embedded in media signals like audio, video and still images can be changed when the signal is transformed. Transmarking of the digital watermark may be used to change the embedded digital watermark technique at signal transformation points to be compatible with the new signal.
• the watermark when playing DVD audio over the radio, analog or digital radio, the watermark can be retrieved and re-embedded at a higher level or using a different technique at the broadcast location. Additionally, the watermark could be modified at a repeater station due to the increased noise level in the signal. This way an audio application can retrieve the watermark, while the original DND can have the lowest change in perception due to the watermark as possible. More specifically, the audio application may be retrieving the watermark in a noisy room and artist won't complain that the DND watermark ruins their recording.
• This method also applies to video signals.
• DND video is transferred to low bandwidth Internet video, such as provided by Real Networks
• the DVD watermark is read and increased in amplitude or re-embedded to survive the massive compression needed to stream video over low bandwidth.
• This watermark may be used for copy protection, but could also be used to enable links or information about the video.
• Transmarking may include converting an out of band identifier like a tag in a header/footer to a watermark or vice versa. It may also involve converting a message in one watermark format to another.
• the process involves a decoding operating on an input media object, and an encoding of the decoded information into the media object. It may also involve a process for removing the mark originally in the input object to avoid interference with the newly inserted mark.
• transmarking There are a variety of reasons to perform transmarking. One is to make the embedded information more robust to the types of processing that the media object is likely to encounter, such as converting from one watermark used in packaged media to another watermark used in compressed, and electronically distributed media, or a watermark used in radio or wireless phone broadcast transmission applications.
• This type of transmarking process may be performed at various stages of a media object's distribution path.
• An identifier in a watermark or file header/footer may be encoded at the time of packaging the content for distribution, either in an electronic distribution format or a physical packaged medium, such as an optical disk or magnetic memory device.
• the media signal may be converted from one format to another.
• This format conversion stage is an opportunity to perform transmarking that is tailored for the new format in terms of robustness and perceptibility concerns.
• the new format may be a broadcast format such as digital radio broadcast, or AM or FM radio broadcast.
• the identifier may be transmarked into a watermark or other metadata format that is robust for broadcast applications.
• the new format may be a compressed file format (e.g., ripping from an optical disk to an MP3 format). In this case, the identifier may be transmarked into a file header/footer or watermark format that is robust and compatible with the compressed file format.
• the transmarking process may leave an existing embedded identifier in tact and layer an additional identifier into the media object. This may include encoding a new watermark that does not interfere with an existing watermark (e.g., insert the new watermark in unmarked portions of the media object or in a non-interfering transform domain). It may also include adding additional or new identifier tags to headers or footers in the file format.
• Fig. 5 is a flow diagram illustrating a process of transmarking.
• the input to the transmarking process is a digitally watermarked signal 20, such as an audio signal (e.g., a music track), a video signal, or still image.
• the digital watermark carries a message payload of one or more symbols (e.g., binary or M-ary symbols) conveying information such as a content identifier, transaction identifier, database index, usage or copy control parameters (flags instructing a device or process not to copy, copy once, not to transfer, etc.).
• symbols e.g., binary or M-ary symbols
• Some examples include: to increase the robustness of the watermark as it undergoes a format change (such as for compression, transmission, digital to analog conversion, up-sampling or down-sampling, printing, display, etc.), to reduce the perceptibility of the watermark before playback, or to balance the trade-off of perceptibility levels vs. robustness levels of the watermark signal for a new as the host signal undergoes a change from one format to another.
• a format change such as for compression, transmission, digital to analog conversion, up-sampling or down-sampling, printing, display, etc.
• the transmarking process illustrated in Fig. 5 begins by detecting a first watermark in the watermarked signal (22).
• a watermark detector employs a watermark key to identify the presence of a watermark. The specific operation of the detector depends on the watermarking process employed. In many techniques, the watermark key specifies the spatial, time, and/or frequency domain location of the watermark signal. It may also specify how to decode a message that has been modulated with a pseudo-random number (e.g., frequency or phase hopping, spread spectrum modulation). To simplify the search for the watermark, the watermark detector searches for reference signal attributes of the embedded signal, such as a known sequence of embedded symbols, or a known signal pattern in a particular time, space, or transform domain. These attributes enable the detector to determine whether a watermark is present in a suspect signal, and to determine its position within the time, space and/or transform domain.
• the watermark detector searches for reference signal attributes of the embedded signal, such as a known sequence of embedded symbols, or a known signal pattern in
• the watermark detector may optionally decode an embedded message (26), such as copy control parameters, content identifiers, owner identifiers, transaction identifiers, etc.
• an embedded message such as copy control parameters, content identifiers, owner identifiers, transaction identifiers, etc.
• This step is optional because the initial detection operation may convey enough information to trigger the remainder of the transmarking operation. For example, the mere detection of the presence of a watermark signal at a particular time, space, or transform domain location may convey one or more bits of message information.
• One type of watermark embedding process encodes symbols by inserting scaled- amplitude, shifted versions of the host signal.
• the shift may be a combination of time, frequency, and/or spatial shifts of the host signal depending on the nature of the signal (e.g., time-frequency for audio, spatial frequency for imagery).
• This shifted version conveys message symbol values by the presence or absence of the shifted version or versions at a particular shift relative to the host, and/or by the amount of change effected to a statistical characteristic of the host signal by the embedding of the shifted version.
• Another type of embedding process embeds a watermark by modulating perceptual domain samples and/or transform domain frequency coefficients
• the message may be randomized by applying a pseudo randomizing process (e.g., spreading a message by multiplying or XORing with a PN sequence) before making the changes to the host to hide the resulting message sequence in the host signal.
• the message may be embedded by an additive process of a modifying signal and/or by a quantization of sample values, frequency coefficient values, or statistical characteristic values.
• the detector looks for attributes of the watermark signal, such as by using correlation or a statistical analysis to detect the shifted versions or modulated samples/coefficients.
• the detector determines whether a watermark signal is present. In some cases, the watermark detector determines that an additional message payload message is present based on the detection of certain watermark signal attributes. It then proceeds to decode additional signal attributes and map them to message symbols. Further error correction decoding may be employed, such as BCH, turbo, Reed Solomon, and convolution decoding, to extract the message payload.
• the transmarking process removes the first watermark signal (28). Again, this process is optional because the transmarking process may proceed by embedding a second watermark without specifically attempting to remove or mitigate the effects of the first.
• the watermark detector Once the watermark detector has detected the watermark and determined its temporal, spatial, and/or frequency domain position, it can remove the watermark or mitigate its effect. It can substantially remove the watermark in cases where the embedding function is invertable, such as a reversible addition operation, by performing the inverse of the embedding function using the watermarking key to specify the attributes and location of the first watermark. It can also remove the watermark without knowing the inverse function, such as using a whitening filter with PN sequence based watermarking.
• the watermarked signal may be converted to another format, such as compressing the signal before the transmarking process proceeds.
• the transmarking process proceeds by embedding a second watermark into the host signal after the format change has occurred. This enables the watermark embedding process to adapt the watermark to the perceptual quality and robustness parameters of the signal in the new format.
• the transmarking process proceeds to embed a second watermark and adapts the watermark to the robustness and perceptual quality parameters appropriate for the new format of the signal before the format change occurs.
• the transmarking process encodes the second watermark (44) using the same or some different embedding process as the first watermark (30).
• This second watermark can be added before the transformation, after the transformation, or during the transformation with a feedback loop.
• the first watermark may be embedded by adding a shifted version of the host signal
• the second watermark may be embedded by adding a perceptually adapted pseudo random carrier signal in the perceptual or some transform domain (like Fourier, DCT, wavelet, etc.), or vice versa.
• the second watermark may modify different temporal, spatial or frequency portions of the host signal than the first, or the two watermarks may overlap in one or more of these portions of the signal.
• this embedding process is specifically adapted to the perceptibility and robustness constraints of the new format or environment.
• This watermark embedding process uses robustness parameters (32) (e.g., watermark signal gain, extent of redundancy, frequency domain locations) to specify the watermark strength, redundancy and frequency domain locations that particularly adapt the watermark for survival in the new format.
• This second watermark may add new information about the transformation where this information can be used for forensic tracking.
• the information could include any combination of the following: an identifier of the transformation device (such as an MPEG encoder device or manufacturer), and an identification of the distribution system, such as an identifier of the broadcast network or cable system. This new information augments the original information embedded into the first watermark and does not alter its meaning, but instead, adds additional payload information.
• the embedding process optionally applies a feedback path that applies the watermarked signal to a degradation process, then measures the number of errors incurred in decoding a known message, and selectively increases the gain of the watermark signal in the portions (temporal, spatial or frequency portions) of the signal where the errors occurred.
• the degradation operations may include a compression operation, or an operation that models degradation likely to be encountered in the new format, such as digital to analog conversion, printing/scanning, broadcast transmission, time scale changes, etc. This process repeats until the measured error rate falls below an acceptable threshold.
• the embedding process uses perceptual quality parameters 33 that specify constraints on perceptual quality of the signal for the new format. These parameters may specify limits on the watermark strength, or. define a perceptibility threshold that can be measured automatically, like Peak Signal to Noise Ratio, typically used in analysis of digital watermarking methods.
• the embedding process optionally includes a feedback path that measures the perceptual quality of the watermarked signal and selectively reduces the gain of the signal in the portions of the signal (temporal, spatial or frequency portions) where the watermarked signal exceeds the perceptibility threshold.
• Fig. 5 graphically depicts the interaction between the watermark embedding process 30, on the one hand, and the rendering/editing environment or transmission environments (34, 36) on the other.
• This diagram depicts how the embedder adapts the new watermark to the environment in which the transmarked signal will be used. For example, if the signal is a still image that is being used in a photo editing software environment, the robustness of the watermark can be adapted to the image processing operations in the editing tool. If the watermark is going to need to survive printing, then the transmarking process embeds the signal with a new watermark designed to survive that process and be recoverable via an image scanned from the printed image.
• the watermark embedder may include additional calibration signal information as set forth in US Patent 5,862,260 to ensure that the watermark can be detected despite geometric distortion.
• the operations in the editing tool can be modified so as to improve the survivability of the watermark, i this case, the image editing operations such as blurring, color transformation, etc. are adapted to preserve the watermark signal to the extent possible.
• the image editing operations such as blurring, color transformation, etc. are adapted to preserve the watermark signal to the extent possible.
• a low pass filter or blur operation that typically reduces high frequency components may be implemented so as to pass selected high frequency components to maintain the watermark signal in those components.
• the operation of adding guassian noise may be modified by shaping or reducing the noise at certain frequencies to reduce interference with the watermark signal at those frequencies, hi cases where watermarks are inserted by modifying a particular color channel such as luminance, the color transform operations may be designed to preserve the luminance of the watermarked image.
• the signal editing tool may be integrated with the transmarking process to decode the watermark before an operation, and then re-encode the watermark after an operation to ensure that it is preserved.
• the wateramark may be re- applied after the image editing tool is used to make an affine transform of an image, or after the image is cropped.
• the watermark may be transmarked at points in the communication channel where the signal (audio, video, or image signal) is transformed.
• the signal audio, video, or image signal
• the signal is un-compressed and re-compressed in another format, where the signal is transformed in a router or repeater (e.g., when the signal is amplified in a router or repeater node in the communication path, the watermark is transmarked at higher intensity), where the signal is transformed into packets in a switching network, the watermark signal may be decoded and re-encoded in the individual packets, or re- encoded after the signal is re-combined.
• the re-encoding is effected by transferring a watermarking command in the header of the packets specifying the watermark payload and watermark embedding protocol to be used in the re-combined signal.
• the transmarking process may be integrated into the compression codec. This enables the codec to modify the compression operation or modify the bitrate to ensure that the watermark survives.
• the compression codec may be designed to preserve certain frequency components that would otherwise be substantially reduced to preserve the watermark. In the latter case, the codec selects a bit rate at which the watermark survives, yet the signal has been compressed to an acceptable level.
• the second watermark can be embedded so as to have less impact on perceptibility.
• the second watermark can be embedded so that it is more robust while staying within the perceptual quality parameters of the rendering device, h addition, the watermark can be changed if DND audio masters are converted to CDs or cassette tapes. If the watermarked signal is going to be transmitted, such as in the broadcast environment, the embedding process encodes the second watermark with robustness to survive the broadcast and maintain the perceptual fidelity within the less rigid constraints of the broadcast environment.
• the transmarking process can be used to encode triggers used in interactive video or audio.
• the triggers may be originally encoded in one format and transmarked into another format before broadcast, or at some node in the broadcast process.
• the trigger can transmarked in video when it is compressed into MPEG2 format for broadcast, or when the content is received at a cable head-end or node in the content distribution channel.
• the trigger may be a network address of interactive content like an IP address or URL, or an index to a network address, interactive content like HTML or XML, etc.
• triggers for interactive content in radio broadcasts can be transmarked when the content is transferred from a packaged medium, such as an optical disk, and prepared for broadcast over traditional radio broadcast, digital satellite broadcast, or Internet streaming broadcast. Like the first watermark, this second watermark employs a watermarking key
• the message decoded from the first watermark such as an identifiers 40, copy control parameters 42 are embedded.
• the result of the transmarking process is a new watermarked signal 46.
• the information or function of the watermark may be transmarked to out-of-band data like a file header or footer, such as an ID3 tag in MP3 audio.
• out-of-band data may be transmarked into in-band data that is embedded into the host signal using a digital watermarking process.
• a watermarking function (e.g., a PostScript-like command) can be provided in the tools. This function is called with parameters specifying the desired features of the watermark information, e.g., payload, robustness level, masks to be used.
• the watermark is actually added as digital data.
• the embedder knows the properties of the rendering device, such as the printer, and appropriately adjust its embedding accordingly. With this concept, watermarks are not lost during composite operations, and watermarks can be embedded in vector (or line) art.
• the color manager at the ripping stage may be the best entity to add the watermark. This idea likewise extends to video - especially MPEG-4 object video, audio - especially MIDI or MPEG-4 structured audio language, and virtual advertisements.
• PostScript-like function to embed a watermark is further detailed in application 09/629,401.
• An alternate method is that no desktop tool has watermarking capability, but instead an on-line watermarking server is available to support common image formats.
• a variety of tools are enabled to submit images to the server with information regarding the desired parameters of the watermark. The server then returns the image to the application. In this way, the burden of integration is virtually eliminated and the registration and marking take place simultaneously.
• a solution is to embed watermark functions in the line-art file, in a similar fashion to how fonts are described with a Bezier curve.
• the watermark function contains the bits to embed as well as rules how to embed these bits in different elements.
• the watermark function could be considered as a command in the popular expanded postscript (EPS) format.
• EPS expanded postscript
• the watermark bits when producing text and a watermark is contained in the line-art, the watermark bits could be embedded by slightly adjusting the position, either vertical, horizontal or both, of each letter. Alternatively, the watermark could be embedded by adding or removing bumps, which are too small to see but can be read digitally, on the edges of the letters.
• any data embedding method can be used according to the bits and rules of the watermark function.
• the watermark function when producing drawing objects, the watermark function could be implemented by embedding the bits in bumps along the edges of the object. Alternatively, when putting a gradient fill inside an object, the watermark function could be implemented by adding more traditional PN sequences within the gradient fill, or modulating halftone dots. In general, the watermark function contains the bits to be embedded and rules or links to how to embed these bits.
• the watermark function is implemented according the desired embedding method when the line-art is rendered, such as on the screen, printer or printing plates.
• Figs. 6 and 7 illustrate a framework for implementing and using the watermark embedding function as a rendering command.
• Fig. 6 is a diagram illustrating a watermark embedding command (100) and insertion of the command into a rendering description file (102).
• the watermark embedding command is specified in a text format or some other binary form compatible with the rendering description file in which it is inserted.
• the user specifies the watermark embedding command and associated parameters. Later, at the time of rendering, the rendering device invokes a watermark embedding module to embed the watermark in the media object according to the watermark embedding command.
• the watermark command parameters include a combination of parameters describing the watermark message payload, the watermark protocol, the watermark embedding method, the payload specification, the embedding locations, the robustness parameters, and the perceptual quality parameters. Any combination of these and other parameters may be used depending on the application.
• the watermark message comprises some number of binary or M-ary symbols.
• the watermark protocol specifies how the watermark message is to be embedded and the meaning of the various symbols in the watermark message.
• the protocol may be specified using one or more parameters. These protocol parameters include a parameter that specifies the embedding method, such as a pointer to a embedder module or plug-in to be used in the rendering device to embed the watermark.
• the method may include a spatial or frequency domain spread spectrum watermark embedder, a watermark embedder that encodes symbols by adjusting samples or features to quantization levels associated with symbols to be embedded, halftone modulation methods (varying halftone dot shapes, screens, error diffusion thresholds, dot cluster sizes or widths according to changes associated with message symbols, etc.).
• the method may include temporal or frequency domain spread spectrum watermark embedder, a watermark embedder that encodes symbols by adjusting samples or features to quantization levels associated with symbols to be embedded, a watermark embedder that encodes a collection of masked tones or time/frequency shifted versions of the host signal corresponding to symbols to be embedded, etc.
• the method may be left unspecified so that the rendering device or transmission channel may optimize the watermark method and protocol for that rendering device or channel.
• the rendering device or channel has a compatible decoder associated with that device or channel for decoding the watermark.
• a universal watermark signal or metadata may be used to specify the watermark type for decoding.
• the protocol parameters may also include more detailed information about the watermark payload, namely a payload specification.
• the payload specification may includes items such as the type of error correcting codes to employ, the type of error detection to employ, the number of message symbols (e.g., binary bits) in the payload, encryption keys for encrypting the payload, etc.
• the protocol may also specify where to embed the watermark, which is referred to as the "embedding locations" in Fig. 6.
• the embedding locations include, and are not limited to, spatial, temporal, and transform domain locations to embed the watermark in the host media signal.
• the transform domain locations refer to transform domain coefficients or sets of coefficients in particular block size of content. Examples of transform domains include Fourier domain, wavelet domain, DCT, etc.
• the embedding locations may specify, for example, that the watermark is to be confined to certain frequency ranges in the signal. Also, for images and video, the embedding location may also specify the color plane or planes in which to embed the watermark signal, such as the luminance channel, the blue channel, or some other color channel.
• the watermark embedder will embed different message payloads in different parts (spatial, temporal, frequency, transform domain portions) of the host media signal.
• the watermark embedding command specifies the parameters for each of the different message payloads, including its embedding location, intensity, fragility (for fragile watermarks), robustness parameters, perceptual quality parameters, redundancy, etc. This enables the watermark embedder module (or modules) to embed combinations of different robust watermarks, robust and fragile watermarks, or fragile watermarks at varying degrees of fragility.
• the message payload may be a single bit, which reduces to the presence or absence of a watermark signal.
• the embedding locations may be specified in terms of spatial, temporal or transform domain masks that specify the areas for embedding the watermark.
• the mask is an array of elements each corresponding to an embedding location. For each element, the mask may be associated with other parameters, such as the payload for that location, the robustness for that location, and the perceptual quality for that location.
• the mask may be designed by the creator of the media object to specify where to, and conversely, where not to embed the watermark, and also to specify the watermark intensity for the areas where it will be embedded.
• the robustness and perceptual quality parameters enable the user or application that inserts the embedding command to control the trade-offs between robustness of the watermark and perceptibility.
• the robustness parameters may be specified in terms of intensity (e.g., watermark signal gain for a particular embedding location), redundancy (e.g., the extent to which the message payload is redundantly encoded across embedding locations to increase its robustness), and frequency locations (e.g., the extent to which the watermark signal is concentrated in lower frequency areas that are more likely to survive transformations of the host signal).
• intensity e.g., watermark signal gain for a particular embedding location
• redundancy e.g., the extent to which the message payload is redundantly encoded across embedding locations to increase its robustness
• frequency locations e.g., the extent to which the watermark signal is concentrated in lower frequency areas that are more likely to survive transformations of the host signal.
• Each of these parameters may be specified as a preferred range to enable the embedding
• the watermark embedding command may also specify the level of fragility of the watermark at particular locations in the media signal.
• Such fragile watermarks are embedded in response to the embedding command. Later at watermark decoding, the presence of the fragile watermark, or its measured strength (e.g., as measured by the error detection rate of a known embedded symbol set, or by threshold levels of detected watermark strength), are used to detect tampering or processing of the watermarked signal.
• This type of robustness and perceptual quality specification enables the watermark embedder module to perform iterative embedding with a feedback path to optimize embedding for a particular rendering or transmission device.
• the embedder initially embeds the watermark payload according to the command parameters at lowest robustness and highest perceptual quality, applies a model of degradation for the particular rendering device or transmission channel to the watermarked signal, and then decodes the watermark to measure the detection error rate for the message payload (e.g., the detection error is quantified using a measure of the difference between decoded symbols and expected symbols before error correction decoding is applied). It then repeats another iteration of this process, increasing the robustness slightly with each iteration until the detection error rate is at a satisfactory level.
• the model of the degradation maybe a compression operation, or a signal transformation that simulates the distortion due to digital to analog - and analog to digital conversion, time scaling, affine transformation, etc.
• the perceptual quality parameters may be specified using automated measures such as peak signal to noise ratio, which quantifies the distortion of the watermarked signal relative to the un-watermarked signal.
• the perceptual quality parameter may be specified as an allowable range or as a threshold which should not be exceeded.
• a media object creation program inserts the watermark embedding command into the rendering description file 102 as another rendering command.
• the rendering description file includes a collection of rendering commands (104, 106, 108) that reference media signals (110, 112) or descriptions of media signals (e.g., 114, such as the case for vector graphics file) to which the rendering commands are to be applied. This file may then be stored for later use, sent to a rendering device, or distributed over a transmission channel.
• There are a variety of potential formats for the rendering description file such as postscript, PCL, EPS, PDF, job tickets, vector graphics, etc. for images and documents, structured audio and MIDI for audio, and MPEG-4 or MPEG-7 for video and audio.
• Fig. 7 is a process for embedding watermarks in media objects using watermark embedding commands.
• the process begins when a user or application program inserts the watermark embedding function as a rendering command (120) into the rendering description file (122). Later, when the media object described in the rendering description file is prepared for rendering, the rendering process (124, 126, 128) reads • the watermark embedding command, and invokes the appropriate watermark embedding module (e.g., 130, 132) to embed the watermark according to the parameters specified in the embedding command (120).
• the watermark embedding module is adapted for the particular rendering device (134, 136, 138) that will render the signal or the transmission channel (140) that will communicate the signal.
• the rendering process may be implemented in a display driver, printer driver, or plug-in to the display or printer driver. It may also be implemented in the printer hardware and specifically integrated into the halftoning process so that the watermark is particularly adapted to the halftone process and is embedded into the image after or while it is rasterized to a halftone image.
• This technique applies to a variety of halftone processes including ordered dithering (e.g., blue noise masks, clustered dot halftones, etc.), error diffusion, stochastic screening, etc.
• ordered dithering e.g., blue noise masks, clustered dot halftones, etc.
• error diffusion e.g., stochastic screening, etc.
• stochastic screening e.g., stochastic screening, etc.
• Adding a perceptually adapted spread spectrum watermark signal to an image in multi-level per pixel format at the halftone dot resolution before converting the image to a halftone image is created by convolving or multiplying the message payload with a pseudorandom carrier signal, and then scaling the carrier signal based on the masking attributes of the image;
• the signal transformation process selects the embedding method and parameters that adapt the robustness of the embedded watermark and perceptual quality of the rendered watermarked signal for the particular rendering process or transmission channel. For example, an audio processor renders a music signal and embeds the watermark payload at a robustness level appropriate for the distribution, broadcast or transmission format. Similarly, a video processor renders a video signal and embeds the watermark payload at a robustness level appropriate for the distribution, broadcast or transmission format.
• the watermark function can specify that the watermark be embedded as part of the signal formatting process, such as part of the process of compressing the image, video or audio signal.
• This enables the watermark module to interact with the compression process to embed the watermark so that it is adapted to that format, e.g., embedding in the compressed data stream or partially compressed stream.
• the compression rate of the signal can be adaptively set by determining the greatest extent of compression where the watermarked signal still survives based on an error detection measure.
• the perceptual quality parameters may be used to tune the compression process so that the compression rate is selected that maintains the desired perceptual quality of the signal and the robustness level of the watermark signal.
• the watermark function can specify that the watermark be embedded after it is converted to a particular format for rendering or transmission (e.g., embedded after compression, or conversion to a broadcast format).
• the rendering or transmission channel provides robustness and perceptual quality parameters about that rendering process or transmission channel to the embedder module so that it can optimize the watermark embedding for the particular rendering process or transmission channel.
• it specifies the watermark robustness, e.g., intensity, or quality constraints that the watermark embedder must adhere to while embedding the payload specified in the watermark embedding command.
• the watermark embedder module queries the rendering process, device or transmission channel for its robustness and perceptual quality attributes. If the quality requirements are lower, then the embedder can increase the robustness of the watermark within an allowable range specified by the watermark embedding command parameters. Conversely, if the quality requirements are higher, then the embedder can select the lowest allowable robustness level specified in the watermarking command to embed the watermark so as to minimize degradation to perceptual quality due to the watermark. The same process can be applied to tune the embedding operation based on the robustness attributes of the rendering process or transmission channel. If the rendering process is expected to substantially degrade the watermark's detectability, then the embedder can select the most robust level for the watermark within the allowable range of the watermark embedding command.
• the watermark embedding command can be designed to select automatically the preferred watermark embedding method for that device or channel.
• the watermark embedding function is particularly well suited for controlling the embedding of watermarks in vector graphics used in virtual advertising for streaming media, like streaming video.
• the virtual advertising is a vector graphic such as a logo that is superimposed on a video sequence when the streaming video is rendered in a receiving device, such as television equipped with a set top box or a personal computer on the Internet.
• This vector graphic file defining the virtual advertising can include a watermark embedding command as described above.
• a watermark embedder module at the receiver embeds a watermark onto the vector graphic. This vector graphic can be used as a trigger for interactive TV applications wherever that video travels.
• the user clicks on (or otherwise selects the logo displayed on the video screen with a cursor control device) to request interactive information such as a web page or to order a product or service when playing previously recorded or live content through a personal video recorder like a Tivo machine.
• the watermark in the logo is then decoded and a payload is extracted from it that indexes a database entry.
• the database returns the interactive information (URL, HTML, web page, etc.) or some other programmatic code that executes on the user's set-top box or computer and enables the user to buy the advertised product.
• the watermark embedding command may be specified for content that includes a combination of different media signals like video, vector graphics, and audio, that get combined at rendering time in the receiving device.
• auxiliary data encoding processes may be implemented in a programmable computer or a special purpose digital circuit.
• auxiliary data decoding may be implemented in software, firmware, hardware, or combinations of software, firmware and hardware.
• the methods and processes described above may be implemented in programs executed from a system's memory (a computer readable medium, such as an electronic, optical or magnetic storage device).

Applications Claiming Priority (7)

Publications (2)

Family

ID=27392750

Family Applications (1)

Country Status (5)

Families Citing this family (30)

Citations (3)

Family Cites Families (5)

• 2001
  • 2001-03-16 JP JP2001570009A patent/JP4785317B2/ja not_active Expired - Fee Related
  • 2001-03-16 EP EP01920399A patent/EP1266475A4/de not_active Withdrawn
  • 2001-03-16 WO PCT/US2001/008315 patent/WO2001071960A1/en active Application Filing
  • 2001-03-16 KR KR1020027012294A patent/KR20020084216A/ko not_active Application Discontinuation
  • 2001-03-16 AU AU2001247457A patent/AU2001247457A1/en not_active Abandoned

Patent Citations (3)

Non-Patent Citations (1)

Also Published As

Similar Documents

Legal Events