ZA200711029B - Method and apparatus for watermarking a video data stream - Google Patents

Description

’ .8 5 . z 2 . h -1-

BACKGROUND OF THE INVENTION

THIS invention relates to methods and apparatus for encoding/decoding and watermarking a video data stream, and in an example embodiment, a : H264 video data stream.

The main driving force for this invention is a concern over protecting copyright. As video becomes more available in downloadable digital form "(such is the case with video on demand), the ease with which perfect copies can be made may lead to large-scale unauthorized copying, which may undermine the film industry.

The aim of the present invention is to provide a method and apparatus to address this.

. AQ 3 #°2007/11¢29 2-

SUMMARY OF THE INVENTION

According to a first aspect of the invention there is provided a method for : finding perceptually significant data carriers within a video data stream to i carry watermark data, the method comprising: : decoding the video stream into its luma components; and measuring variance of the directions of the gradient of brightness in - the luma components of the decoded video stream to identify coordinates for perceptually significant data carriers. in the video data stream.

The method may comprise selecting coordinates for perceptually significant data carriers that include a residue in luma components of the video data : stream. oo

In an example embodiment, a plurality of coordinates of perceptually ] significant data carriers are identified, with the method further comprising selecting (either randomly or algorithmically) a subset of coordinates from the identified coordinates of perceptually significant data carriers. : ~The method may further comprise storing the coordinates for the selected data carriers and uniquely associating these coordinates with the particular video data stream being processed. oo

According to a second aspect of the invention there is provided a method oo for encoding fingerprint payload data for insertion into a video data stream, the fingerprint payload data comprising a plurality of data packets, the method comprising: channel encoding each data packet of the fingerprint payload data, including the addition of error correction; : 4 t ' .® “ ®2007/ : livag rd -3- spreading each packet of fingerprint payload data using a pseudo random sequence per character; and storing the resulting fingerprint in a fingerprint store. "According to a third aspect of the invention there is provided a method for encoding a stream of video data and for inserting a watermark therein, the method comprising: Co receiving and parsing an incoming video data stream to_at least partially decode the video data stream on a frame by "frame basis so as to expose transformed residue coefficients; : oo using generated frequency domain residue coefficients to calculate an adaptive gain factor for watermark insertion; : using perceptually significant data carriers ~ coordinates together with the adaptive gain factor and frequency domain : residue coefficients to provide watermarked coefficient data; and : inserting the watermarked coefficient data directly into the video data stream. : Additionally, the frequency domain residue coefficients are generated by: : deriving a residual video data frame from the video data stream; and : transforming the residual video data frame so as to provide the generated frequency domain coefficients.

’ . 4- R .

According to a fourth aspect of the invention there is provided a method for decoding a stream of video data and for detecting a watermark therein, the” method comprising: oo : receiving and decoding both the watermarked and the original video data stream in their luma components; and oo | extracting watermark information using perceptually significant data carriers’ coordinates together and associated pseudo random sequences.

According to a fifth aspect of the invention, there is provided an apparatus for finding perceptually significant data carriers within a video data stream to carry watermark data, which apparatus comprises: Co a full video data stream decoder, the full video stream decoder being arranged to decode the video data stream thereby producing its luma components; and ) a feature extractor operable to find coordinates for perceptually ’ significant data carriers in each frame of luma components of the video stream. oo

The feature extractor may be operable to measure variance of the . directions of the gradient of brightness in the luma components of the decoded video stream thereby to find coordinates for perceptually significant data carriers in the video data stream.

The apparatus may comprise a perceptually significant carrier coordinate store operable to store coordinates for the selected data carriers that may be unique to the particular video data stream being processed.

. : PS ) . / - =

So According to a sixth aspect of the invention there is provided ‘an apparatus oo for encoding a stream of video data and for inserting a watermark therein, which apparatus comprises: : a full bit stream decoder arranged to receive and parse video stream motion vectors, transformed residue coefficients and header information from the stream of video data in order to produce an original YUV frame used together with a predicted frame to oo generate a residue video data frame which is transformed to expose - transformed residue coefficients for watermarking; : an adaptive gain module arranged to calculate an adaptive gain factor based on generated frequency domain residue coefficients; - and . a watermark insertion module arranged to receive and use perceptually significant data carriers’ coordinates from a : perceptually significant data carriers store together with the gain " factor and frequency domain residue coefficients to insert the watermarked coefficient data into the video data stream thereby to : produce watermarked video data frames. :

The apparatus may further comprise: a residue frame calculator which is operable to derive the residual video data frame from the video stream by way of at least the : exposed header information and motion vectors; and a transformation module operable to transform the residual video data frame so as to provide the generated frequency domain residue coefficients.

The apparatus may comprise a bit stream reconstruction module operable to add the originally exposed header information and motion vectors to the - video data frames so as to produce a watermarked bit stream. -

The apparatus may further comprise an encoded fingerprint store from which encoded fingerprints or watermarks are receivable by the watermark insertion module. .

According to a seventh aspect of the invention there is provided an oo apparatus for decoding a stream of video data and for detecting a watermark therein, which apparatus comprises: a full bit stream decoder arranged to receive and decode both the watermarked and the original video data stream into their luma : components; and a detector arranged to receive perceptually significant data carrier coordinates from a perceptually significant data carrier store and pseudo random sequences from an encoded fingerprint store, the detector being operable to use the received perceptually significant : data carriers coordinates and the pseudo random sequences to extract watermark information thus recovering the inserted fingerprint payload. :

- BRIEF DESCRIPTION OF THE DRAWINGS : Figure 1 shows a system to encode/decode a video data stream : according to an example embodiment; :

Figure 2 shows a method for finding perceptually significant data carriers within a video data stream, according to an example embodiment; :

Figure 3 shows a method for encoding a fingerprint payload that can then be inserted into a video data stream, according to an oo example embodiment; :

Figure 4 shows a method for encoding a stream of video data and for inserting a watermark therein, according to an example embodiment; and oo : oo Figure 5 shows a method for decoding a stream of video raw data oo and for detecting a watermark therein, according to an : example embodiment.

DESCRIPTION OF PREFERRED EMBODIMENTS oo

In broad terms, the present disclosure provides a method, system and apparatus for watermarking (also known as “fingerprinting”) video data, so that the watermarked video data can be used to detect and track copyright violators.

Watermarking, for the purposes of this specification, may be defined as the insertion of hidden data (e.g., serial numbers) or a set of characteristics that may be used to distinguish an object from another similar object. An object of the watermarking process of video data is to make the hidden data imperceptible to a viewer and at the same time impossible to erase by an attacker without degrading the video quality. In order to achieve this, and in an example embodiment, fingerprinting information may be placed on or near a border corner or edge corner between perceptual significant and insignificant parts of encoded video data in the data stream.

The border corner or edge corner where fingerprinting information is to be placed will now be described in more detail. Edges, in this context, are characterized by abrupt changes of intensity in a video frame. Edges can either be boundaries between two regions (step edges) or small elongated objects of only a few pixels width so that the boundaries on both sides of that object are too close to be separated (line edges). A video frame edge is perceptually significant and hence, if attacked or distorted, results in : notable degradation of the video data. Further selection of line edges that : : intersect to form corners may be done to ensure maximum possible : perceptible impact should an attacker try to-remove the fingerprint.

In an example embodiment, the method and system use a well known watermarking technique of spreading the watermarking information into a : spectrum of the original video data coefficients. This is sometimes referred to in the art as “information hiding in the frequency domain”. In spread spectrum communications, a narrow band signal may be transmitted over a much larger bandwidth such that the signal energy present in any single frequency is imperceptible. Similarly, watermark information may be : spread over very many frequency bins so that the resultant energy in each bin is imperceptible. oo

Video transformed into the frequency domain may be viewed as : communication channel noise, while the watermark information may be viewed as a signal. Referring first to Figure 1, a high level system 10 to encode/decode a video data stream (e.g., a H264 data stream) and insert a watermark into that data stream is shown. The system 10 comprises a watermark insertion module 12 to receive the original video data stream 13.

The watermark insertion module 12 also receives a signal from a fingerprint encoder module 14 comprising the watermark data 15 to be inserted into :

the video data stream 13. The watermark insertion module 12 also receives a signal from a perceptually significant carrier coordinate store 16, the signal comprising the location (as coordinates 17) within the video data stream that the watermark data is to be inserted. The watermark insertion module 12 then inserts the watermark data into the video data stream at a location provided by the perceptually significant carrier coordinate store 16, : so as to provide a watermarked video data stream 19.

When and where the insertion of the watermark data into the video data oo : stream occurs is implementation specific. In an example embodiment, the insertion may take place at the time of transmission of video to a set top box, for example. A media server may, for example, be configured to perform this function. Alternatively, this insertion may take place on the set top box itself before display of the video on a visual display device. : The system 10 may also comprise a watermark detector module 18 that may receive the watermarked video data stream 19 as well as the original } video data stream 13. The watermark detector module 18 may additionally oo receive the coordinates 17 from the perceptually significant carrier coordinate store 16 to enable it to locate and extract the watermark data stream from the watermarked video data stream.

Where and when watermark detection occurs may also be implementation specific. The content provider of movies on demand, for example, may be : held responsible by the studios to provide details of possible copyright - infringements. The watermark detector module 18 may be used in these : circumstances to verify from which set top box, or content provider, the movie originated. This may be possible by assigning a unique fingerprint to each set-top box or content provider that is to insert fingerprints into video data streams to be distributed or displayed. :

In an example embodiment, a frequency domain representation for the luminance component of the H264 video data stream is provided from which perceptually relevant data carriers may be selected for carrying the watermark. The video data stream is decoded into its luma or luminance components and its chrominance components, with the luma components then being extracted from the video stream. The luma represents the overall brightness in an image, i.e., the overall brightness of the amount of visible light of the spot or color space. The Y component found in the YUV video format, for example, may provide the luminance component of the : video. The chrominance is the signal used in many video systems to carry oo the color information of the picture separately from the accompanying luma : signal. :

In an example embodiment, this selection may be done on a frame by frame basis. These perceptually relevant data carriers are then converted : into the coordinates that are stored as “locations” in the perceptually significant carrier coordinate store 16. ’ ‘By employing a perceptual model that chooses perceptually significant data carriers on a frame by frame basis, there is little correlation between : chosen frame locations. Not only does this model not make use of fixed locations in video data frames for inserting the watermark so as to provide a more secure arrangement, the use of perceptually significant data carriers also adds a great deal of robustness to the watermark signal.

The selected perceptually relevant data carriers may also be used to determine the strength of insertion factors for inserting the watermark. This ~ aspect will, however, be described in more detail further on in the specification. .

Employing a perceptual model in order to choose locations for watermarking may be computationally expensive. In an example embodiment, as described above, the selection of perceptually relevant data carriers to carry may only need to be done once for each unique video . data stream, with the perceptually significant carrier coordinate store 16 then storing the associated coordinates for future use. This may reduce the computational cost of processing for subsequent watermark insertions in the same video data stream.

The watermark (also known as a payload or fingerprint) to be embedded : into the video data stream may first be channel coded and then “transformed to make it more suitable for embedding.” Channel coding is done to increase robustness against possible attacks. Channel coding is a Co broadly used term mostly referring to the error correcting codes and sometimes digital modulation methods in communication. This allows the data to travel though a noise filled channel and still be successfully recovered, hence its robustness. The watermark is transformed by applying dynamic weighting factors derived from localized frequency domain luminescent coefficients. The weighting factors change with location in each frame thus the watermark is weighted suitably for the location at which it is to be inserted. This minimizes any perceptible impact the insertion may have. : ] The digital fingerprint information to be embedded may be used to. modulate a larger pseudo random sequence, much like a spread-spectrum sequence before insertion. Spread-spectrum techniques are well known methods in which energy generated at a single frequency is deliberately oo spread over a wide band of frequencies. This is done for a variety of reasons, including increasing resistance to deliberate or natural interference and to prevent hostile detection. In the present application, the payload (fingerprint or watermark data) is used to modulate a much larger : set of data (larger pseudo random sequence) resulting in a spread payload. oo Unless the attacker knows the pseudo random sequence they will not easily detect the presence of the payload (fingerprint or watermark). -

Referring back to Figure 1, the watermark detector module 18 requires that the pseudo random sequences used to encode the watermark are known.

In addition, the watermark detector module 18 performs its task by comparing the watermarked video data stream with the original, non- marked, video data stream to take its decision.

In an example embodiment, the watermark detector module 18 is non-blind. ) In general, non-blind algorithms are more robust than blind algorithms. In . this regard, by knowing the original, non-marked, non-corrupted host signal, preprocessing may be carried out to make watermark extraction easier.

Co For example, in the case of video watermarking, rotation and magnification factors of the scaled copy can be easily estimated and compensated for if

E the non-marked video is known.

In an example embodiment, during watermarking, a subset of perceptually significant data carriers may be chosen based on information that is to be inserted in the video data stream. This ensures low location correlation between different copies of the data stream containing similar watermark payloads. For example, if two copies are watermarked, the locations chosen on a frame by frame basis for each copy will differ; thus, there is less chance of an attacker discovering a pattern of locations used during . _ watermarking. :

According to an aspect of the invention, with reference now to Figure 2, a method 20 for finding the perceptually significant data carriers within a : video data stream to carry watermark data will be described in more detail.

This information may be stored in the perceptually significant carrier coordinate store 16, shown in Figure 1, for use when the fingerprint data is inserted into the video data stream. The method 20 comprises decoding the video data stream (e.g., an H264 video data stream) into its luma and chrominance components.

In one example embodiment, the video data stream is first receiving and then parsed before being fully decoded. This may be done by using a full video data stream decoder 22. By parsing and fully decoding the video data stream a video in YUV format 24 is produced.

Each YUV image in the YUV video 24 may then be sent to a feature extractor 26 which is adapted to find coordinates for perceptually significant data carriers in the video data stream, as indicated by block 28. The coordinates are found in the spatial domain (pixel domain), however the significant carriers are marked in the frequency domain. A significant carrier (e.g., at a corner) is found in the Y component (of the YUV image), ) : then the co-ordinate of that sub-block is stored. In an example embodiment, the algorithm used by the feature extractor 26 for finding perceptually significant carriers utilizes the measuring of variance of the : : . directions of the gradient of brightness in the luma components thereby to identify coordinates for perceptually significant data carriers. : Direct corner (location of the perceptually significant carriers) detection is done without segmenting the image in advance. The algorithm determines the corner response function combining the angle and the contrast of the corner. The function has maxima at corner points. Perceptually significant carriers must lie at these corner point coordinates. Furthermore, the sub : blocks at these corner point coordinates are subjected to residue testing. A : residue, broadly, is anything left behind by a process, in this case, after motion estimation is performed. Only if the significant carrier has a residue } i in the embodiment stream, e.g., in the identified luma components, then it is stored, otherwise it is discarded. :

Thus, the coordinates chosen as having watermarkable perceptually significant carriers must pass two tests: 1. There must be a corner present at that coordinate so as to make it perceptually significant. 2. There must be a residue in the luma components at that coordinate

Co so as to ensure that watermarked coefficients will pass unaltered into the H264 video data stream.

In an example embodiment, more locations of perceptually significant data carriers than are required for the fingerprint payload are selected. This allows for a random or algorithmic subset selection of locations and ensures that there is little or no correlation between watermarkable

J oo | Co 14- locations in each sequential video frame. A perceptually significant carrier coordinate store 30, which may be similar to coordinate store 16, may be provided to store coordinates for the selected data carriers that may be unique to the particular video data stream being processed. As indicated above, the store 30 may be reused each time the particular video stream is oo watermarked.

According to another aspect of the invention, with reference now to Figure 3, a method 32 for encoding a fingerprint payload that can then be inserted

Co into a video data stream will now be described. In an example embodiment, this method 32 may be performed by the fingerprint encoder 14 shown in Figure 1. In an example embodiment, the fingerprint payload comprises a series of fingerprint symbols, e.g., a plurality of data packets which may be a plurality of 8 bit characters, indicated by reference numeral 34. ‘ :

The fingerprint symbols 34 may then be channel encoded using a channel : encoder 36 which encoder 36 also adds error correction, with the resulting data sequence containing error correction. A pseudo random sequence generator 38, which generates a pseudo random sequence for each packet or 8 bit character may also be employed during the method. The method 32 then spreads each packet or 8 bit character in the channel coded fingerprint symbols series using one pseudo random sequence per packet or character. The resulting fingerprint may then be stored in an encoded fingerprint store 40 to be used during watermarking of the video data . stream, e.g., during the method described in accordance with Figure 1.

According to yet another aspect of the invention, with reference now to

Figure 4, a detailed method 42 for encoding a stream of video data 44 and for inserting a watermark therein will now be described. In an example embodiment, the method 42 comprises a pre-processor 46 receiving, reading and parsing an incoming video data stream 44. On a frame by frame basis, the video data stream 44 may then be partially decoded, in one ‘example embodiment also by the preprocessor 46, to expose header information, motion vectors and transformed residue coefficients. Header oo and motion vectors may simply be allowed to pass through, as indicated by arrows 48 and 50, respectively. The original decoded picture frame (YUV) 52 is derived from the header information, motion vectors and transformed residue coefficients using a full bit stream decoder 54.

A predicted frame 56 is derived from the motion vectors (arrow 50) and headers (arrow 48) in the .original video data stream, as well as a watermarked frame storage module 58. The watermarked frame storage module 58 holds the YUV frame of the decoded watermarked stream that is : "obtained from the watermarked residue frame 78 and the previous predicted frame. The predicted frame is obtained by using the ' watermarked YUV frames in the watermarked frame storage module 58 as reference frames and applying the motion vectors and headers to these frames. A residue frame 60 may then be calculated by subtracting the predicted frame 56 from the original decoded picture frame (YUV) 52. The residue frame 60 may be calculated by way of a residue frame calculator, shown by reference numeral 61. E oo EE Inan example embodiment, the residue frame 60 may be transformed. and quantisized by way of a transformation module 62, to provide the frequency domain coefficients for watermark consideration. For example, the watermark may be inserted at the 3 mid band and 2 lower band frequency - domain luminescent coefficients of each four by four sub block that are marked for a given perceptually significant carrier coordinate.

A watermark insertion module 64 (which may be similar to the water mark oo insertion module 12 of Figure 1) receives an encoded fingerprint or watermark from an encoded fingerprint store 66 (which may. be similar to encoded fingerprint store 40 in Figure 3), via operation 68. Operation 68 also receives a signal from an adaptive gain module 70, which may be arranged to calculate an adaptive gain factor based on the frequency domain coefficients received from block 62.

a -16- | Co - Calculation of the adaptive gain factor for the insertion of the watermark may be done as follows. A frequency domain luminescent component sub : block is scanned to find the maximum a.c. value. Using this maximum a.c. value the weighting factor a is scaled a factor no less than one. The weighting factor is calculated for each sub block containing perceptually significant data carriers for a given frame. The perceptually significant carrier V; is added to the weighted encoded fingerprint (a*F; ) information to get the watermarked coefficient (V*). : : Vi =Vi+a'Fi (equation 1) | :

Care is taken in order not to cause saturation or brightness anomalies such as white spots. The use of a dynamically chosen weighting factor a ensures equation 1 is appropriate even when values of V;vary widely.

Using the perceptually significant data carriers coordinates retrieved from a perceptually significant carrier coordinate store 72 (which may be similar to the store 16 shown in Figure 1 or the store 30 shown in Figure 2), together with the gain factor and frequency domain coefficients, watermarked : coefficient data may be inserted directly into the video frame using the watermark insertion module 64. In an example embodiment, sequential frames in the video data stream are marked with the same data to provide resilience to temporal attacks such as frame skipping/dropping. In addition, since watermarking is performed on the transformed quantized coefficients, quantization loss and distortion during insertion is minimized.

The watermarked video data frames are then sent to a bit stream reconstruction module 74 which can add the originally stripped header information and motion vectors so as to produce a watermarked bit stream, as indicated by reference numeral 76.

The watermarked frame storage module 58 signal is formed by adding the . predicted frame 56 and an inversed transformed and quantized residual i | _ 22007711029 frame (shown by reference numeral 78), after the watermark has been inserted. : X oo ~ According to yet another aspect of the invention, with reference now to ’ Figure 5, a method 80 for decoding a stream of video raw data and for detecting a watermark therein will be described.

The method 80 may comprise the steps of receiving and parsing both the original and the watermarked video bit streams, as indicated by pre-processor blocks 82 and 84 respectively.

Both video bit streams may then be processed by a full bit stream decoder (blocks 86, 88, respectively) to produce luma components in a YUV video, as indicated by blocks 90 and 92, respectively.

A geometric/signal processor 94 may be used to process the two signals before detection can begin.

A detector 96 may use the perceptually significant data carriers coordinates from the perceptually significant carrier coordinate store 100 together with pseudo random © sequences from the encoded fingerprint store 98 to extract and channel decode watermark information, thus recovering the inserted fingerprint payload. | }

Claims (1)

1. v , Co o.oo CL : : CLAIMS:

   1. "A method for finding perceptually significant data carriers within a video data stream to carry watermark data, the method comprising: . decoding the video data stream into its luma components; and measuring variance of the directions of the gradient of . brightness in the luma components of the decoded video data stream to identify the coordinates for perceptually significant data carriers in the video data stream. :

   2. A method as claimed in claim 1, the method comprising selecting coordinates for perceptually significant data carriers that include a residue in luma components of the video data stream. :

   3. A method as claimed in claim 1 or claim 2, the method comprising identifying a plurality of coordinates for percueptually significant: “data carriers and randomly selecting a subset of coordinates from : the identified coordinates of perceptually significant data carriers. :

   4. A method as claimed in claim 1 or claim 2, the method comprising identifying a plurality of coordinates for percueptually significant data carriers and algorithmically selecting a subset of coordinates . from the identified coordinates of perceptually significant data : carriers.

   5. A method as claimed in any of the preceding claims, the method comprising: : storing the coordinates for the selected data carriers; and of uniquely associating these coordinates with the particular video data stream being processed.

   6. A method for encoding a fingerprint payload data for insertion into a video data stream, the fingerprint payload data comprising a plurality of data packets, the method comprising: : oo channel encoding and adding error correction to each data packet of the fingerprint payload data; spreading each packet of finger payload data using a’ pseudo random sequence per character; and - "storing the resulting fingerprint in a fingerprint store.

   : 7. A method as claimed in claim 6, wherein each of the data packets comprise 8 bit characters.

   8. A method for encoding a stream of video data and for inserting a watermark therein, the method comprising: ) oo receiving and parsing an incoming video data stream to at least partially decode the video data stream on a frame by frame basis so as to expose transformed residue coefficients; using generated frequency domain residue ‘coefficients to calculate an adaptive gain factor for watermark insertion; : using perceptually significant data carriers coordinates together with the adaptive gain factor and frequency domain : residue coefficients to provide watermarked coefficient data; and oo -20- : inserting the watermarked coefficient data directly into the oo video data stream.

   9. A method as claimed in claim 8, wherein the frequency domain residue coefficients are generated by: Co : deriving a residual video data frame from the video data stream; and transforming the residual video data frame so as to provide the generated frequency domain coefficients.

   10. A method for decoding a stream of video data and for detecting a : watermark therein, the method comprising: receiving and decoding both the watermarked and the original video data stream into their luma components; and extracting watermark “information using perceptually | ) significant data carriers’ coordinates and associated pseudo random sequences. : AE An apparatus for finding perceptually significant data carriers within : a video data stream to carry watermark data, which apparatus comprises: oo a full video data stream decoder, the full video data stream decoder being arranged to decode the video data stream : into its luma components; and : a feature extractor operable to find coordinates for perceptually significant data carriers in each frame of luma : components of the video data stream.

   12. An apparatus as claimed in claim 11, wherein the feature extractor is operable to measure variance of the directions of the gradient of brightness in the luma components of the decoded video data stream thereby to find coordinates for perceptually significant data carriers in the video data stream.

   13. An apparatus as claimed in claim 11 or 12, which apparatus comprises a perceptually significant carrier coordinate store operable to store coordinates for the selected data carriers that may be unique to the particular video data stream being processed:

   14. An apparatus for encoding a stream of video data and forinsertinga watermark therein, which apparatus comprises: a full bit stream decoder arranged to receive and parse oo video stream motion vectors, transformed residue coefficients and header information from the stream of video data in order to produce an original YUV frame used together with a predicted frame to generate a residue video : - data frame which is transformed to expose transformed residue coefficients for watermarking; an adaptive gain module arranged to calculate an adaptive gain factor based on generated frequency domain residue coefficients; and : a watermark insertion module arranged to receive and use perceptually significant data carriers’ coordinates from a : perceptually significant data carriers store together with the gain factor and frequency domain residue coefficients to insert the watermarked coefficient data into the video data stream thereby to produce watermarked video data frames. : 15. An apparatus as claimed in claim 14, further comprising:

   -22- Co a residue frame calculator operable to derive the residual video data frame from the video stream by way of at least the exposed header information and motion vectors; and a transformation module operable to transform the residual video data frame so as to provide the generated frequency domain residue coefficients.

   16. An apparatus as claimed in claim 14 or 15, which apparatus comprises a bit stream reconstruction module operable to add the : originally exposed header information and motion vectors to the video data frames so as to produce a watermarked bit stream.

   A7. An apparatus as claimed in either claim 12 or 13, which apparatus comprises an encoded fingerprint store from which encoded fingerprints or watermarks are receivable by the watermark insertion module. a oo 18. An apparatus for decoding a stream of video data and for detecting a watermark therein, which apparatus comprises: : : a full bit stream decoder arranged to receive and decode both the watermarked and the original video data stream into : their luma components; and a detector arranged to receive perceptually significant data carrier coordinates . from a perceptually significant data carrier store and pseudo random sequences from an encoded fingerprint store, the detector being operable to use the received perceptually significant data carriers’ coordinates and the pseudo random sequences to extract watermark information thus recovering the inserted fingerprint payload.

   vc

   19. A method for finding perceptually significant data carriers within a video data stream to carry watermark data substantially as herein described and illustrated.

   20. A method for encoding a fingerprint payload data for insertion into a video data stream substantially as herein described and illustrated.

   21. A method for encoding a stream of video data and for inserting a : watermark therein substantially as herein described and illustrated

   22. A method for decoding a stream of video data and for detecting a watermark therein substantially as herein described and illustrated.

   23. An apparatus for finding perceptually significant data carriers within a video data stream to carry watermark data substantially as herein described and illustrated.

   24. An apparatus for encoding a stream of video data and for inserting a watermark therein substantially as herein described and illustrated.

   25. An apparatus for decoding a stream of video data and for detecting : a watermark therein substantially as herein described and illustrated. a 19™ DAY OF DECEMBER 2007 SPOOR & FISHER B APPLICANTS PATENT ATTORNEYS :