Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
2000-01-18
2004-08-31
Philippe, Gims (Department: 2613)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
C382S251000, C348S461000
Reexamination Certificate
active
06785332
ABSTRACT:
TECHNICAL FIELD
This invention concerns marking of video signals and in particular a method for marking a compressed digital video signal by incorporating a digital signature embedded in the video signal.
The options available for transmitting images via new media such as the Internet and satellite television mean that the volume of data exchanged, especially video images, will continue to increase.
In parallel, recent developments in the television and digital video sector not only afford an appreciable improvement in image quality but also enable copies of quality identical to the original to be made, such that the copy is indistinguishable from the original.
This is why research has been undertaken with a view to enabling absolute identification of the author of a video image, to enable him to enforce his rights with regard to unauthorised broadcasters.
Prior Art
Methods are known for the protection of video images consisting in incorporating a digital signal by slightly modifying the data in order to be able to identify the said signature. This signature may then be used to identify the author of the image. Two main signature techniques have been used to date. The simplest entails modulating the brightness of a pixel selected pseudo-randomly. This technique is used in particular for grey scale images and is described in particular in the article by R. G. van Schnydel et al. (<<(Digital watermark>>,
Proceedings of the
1994 1
st
IEEE International Conference on Image Processing
, Vol. 2, pp. 86-90, 1994).
More complex techniques have also been proposed in which the image is divided into blocks with amplitude modulation and block size that are in turn modulated by local energy. Techniques suitable for black and white images have been proposed by K. Matsui et al. (<<How to secretly imbed a signature on a picture,>>
The Journal of the interactive Multimedia Association Intellectual Property Project
, Vol. 1, No. 1, pp. 187-206, January 1994). In the article by W. Bender et al. (<<Techniques for Data Hiding)>>,
Proceedings of the SPIE,
2420:40, February 1995), the difference between the brightness value of pixels is used. Whilst this latter article proposes duplication of textured areas and then use of an auto-correlation calculation, a technique very often used is modification of the DCT coefficients generated by JPEG encoders (S. Burgeft et al. <<A Novel Method for Copyright Labelling Digitized Image Data>>, IEEE Transactions on Communications, September 1994) or MPEG-2. These two techniques are well suited to colour images or animated colour images respectively.
The number of bits generated by the images in the transmission of video information requires a high compression rate. In this field, the MPEG format has rapidly taken over and has become an international standard, with most professional equipment processing video signals in this format.
An identification technique specific to MPEG uses part of the compressed signal for information designed to identify the supplier, as well as a so-called protection bit, indicating that the image is protected and that copying is prohibited.
However, these factors do not offer any security as it is very easy to delete them as their position is clearly defined in the MPEG standard.
As mentioned previously, this invention concerns a method for marking a compressed digital video signal by incorporating a digital signature embedded in the video signal.
This method is only of use if the signature cannot be deleted easily and if it can withstand handling such as compression and decompression, a zoom function and image shift, this signature obviously not altering the video signal other than imperceptibly.
It is known that the MPEG standard is based on analysis of the development in the various frames and transmission of the differences. It has been noted in particular that, from one image to another, a large amount of the information does not change, or is in a slightly different plane. To this effect the MPEG encoder breaks down the image into blocks, usually of 8 by 8 pixels, with comparisons then being made of these blocks. In particular, these operations concern DCT, DFD and motion vectors. The most recent versions of MPEG enable a set of blocks to be joined and operations then carried out on this set of blocks termed a region.
The DCT (Discrete Cosinus Transform) is a transformation into a Discrete Cosinus which enables amplitudes of the different frequencies forming an image to be obtained. Its advantage lies in the fact that it is then possible to selectively compress the different frequency bands in order to minimise the visual distortion. It is generally done on blocks of reduced size (typically 8×8). This transformation is applied to all the blocks forming the image. It is followed by a quantification and entropic coding stage.
The DFD (Displaced Frame Difference) represents the difference between the image predicted by the translational model and the actual image.
DCT coding is applied to transmit DFD type information.
Motion vectors are used when the same block is slightly offset from one frame to the next. In this case, the MPEG format provides indication that a block present in the preceding image reappears in the current image slightly offset in plane, by a value corresponding to the coefficients X and Y. This principle allows errors, i.e. if a small part of a new block recreated in this way is different, the motion vector will be accompanied by DFD data containing the differences between the displaced preceding block and the actual visual information of this new block.
Previous marking experiments used DCT coefficients to incorporated the signature in a compressed video signal. This technique is, however, highly susceptible to image framing alignment problems, and displacement of one or two pixels results in different DCT coefficients, which prevents extraction of the signature. This disadvantage is exacerbated further still in that certain video processing equipment causes offset by one pixel on the video signal produced.
SUMMARY OF THE INVENTION
The purpose of this invention is to offer a method for marking a compressed digital video signal by incorporating a signature embedded in the digital signal, which is robust, invisible and identifiable in real time.
To this effect the invention concerns a method for marking a compressed digital video signal by incorporating a digital signature embedded in the compressed video signal, the said signal representing a series of at least two video images, each of the images being divided into several regions, the said signal comprising motion vectors representing motion of the regions between the first and the second image, the method being characterised by the fact that at least one of the coefficients X or Y is modified on at least one of the said motion vectors, with a region possibly consisting of one block.
According to one embodiment, a set of motion vectors MV(i) is selected with a low visual impact, preferably a set of vectors with a standard lower than a threshold R, for example a threshold R equal to 5.
According to an initial variant, a set of vectors contained in a frame preceding a frame containing all the image information is selected (type I frame).
According to a second variant, the set of motion vectors MV(i) is changed according to a pseudo-random selection modulated by a code, the signature possibly comprising several bits S(i) and at least one of the said coefficients X or Y of the motion vectors possibly being modified according to at least one of the signature bits, with the method possibly also comprising the following steps:
generation of a random number A initialised by a parameter transmitted by the video signal, comprising the same number of bits as S;
modification of one of the coefficients X or Y such that this is signed with bit <<1>> if more than half the bits of A are identical to S or with bit <<0>> if the reverse applies.
The parameter initialising the random number may be
Ebrahimi Touradj
Jordan Frederic
Kutter Martin
Browning Clifford W.
Ecole Polytechnique Federale de Lausanne
Philippe Gims
Woodard Emhardt Moriarty McNett & Henry LLP
LandOfFree
Method for marking a compressed digital video signal does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method for marking a compressed digital video signal, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method for marking a compressed digital video signal will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3311955