Image analysis – Applications
Reexamination Certificate
2000-02-04
2004-09-14
Au, Amelia M. (Department: 2623)
Image analysis
Applications
C382S190000, C382S289000
Reexamination Certificate
active
06792128
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention concerns methods of inserting additional information, such as a digital signature, in a digital image.
It also concerns methods of extracting this additional information.
Correlatively, the present invention concerns a device for inserting and a device for extracting additional information adapted to implement respectively the insertion methods and the extraction methods according to the invention.
2. Description of the Related Art
The present invention applies more generally to the field of watermarking digital data, and more particularly digital images.
In general terms, the watermarking of digital data can be interpreted as the insertion of a seal or signature, for example for authenticating the author or origin of a digital document.
Conventionally, such a signature must be inserted imperceptibly, that is to say in a fashion which is not detectable to the eye in the case of an image.
It must also be robust to the normal operations of digital signal processing, such as compression, for example, for fixed images, in accordance with the JPEG (Joint Photographic Expert Group) standard, filtering, digital to analogue conversion (for printing an image for example), analogue to digital conversion (for digitising a printed image for example) and geometric manipulations of the image, such as plane similarities integrating translations, rotations, changes of scale or even divisions.
The signature must also be robust to intentional attacks aimed at extracting or erasing it so as, for example, to arrogate the copyright over an image.
The majority of techniques for inserting additional information in digital data consist of choosing a set of so-called perceptually significant spatial or frequency components of the digital signal, in which the additional information can be inserted in a robust fashion notably vis-á-vis the normal compression methods, signal processing operations, intentional attacks and also, to a certain extent, geometric transformations.
Such a technique is described for example in the European patent application EP 0 766 468 filed in the name of NEC CORPORATION.
These digital image watermarking techniques are based on a prior frequency transformation of the digital image, for example by means of a discrete cosine transformation by blocks, a choice of significant components amongst the components obtained by transformation, a modulation of the components chosen for inserting the signature and an inverse frequency transformation in order to obtain the watermarked image.
When the signature is extracted, the extraction method also uses a frequency transformation of the image to be authenticated, a choice of the significant components and a demodulation of these components in order to find the inserted signature.
There also exist techniques in which the coefficients are modified according to a predetermined rule, known to the decoder. The decoding step then consists of examining whether this predetermined rule is indeed verified or not. Such a technique is described for example in the document entitled “Video-steganography: how to secretly embed a signature in a picture” by Matsui Kineo and Tamaka Kyoshi, IMA Intellectual Property Project Proceedings, Volume 1, issue 1, January 1994.
Nevertheless, when the image has undergone geometric transformations such as plane similarities, the parameters of these geometric transformations must be known a priori or estimated in a relatively precise fashion when the signature is extracted so as to re-fix the image in space.
Such a re-fixing is generally difficult to effect and requires having the original image at the decoder implementing the signature extraction method.
One means of getting around the difficulty is to use a frequency transformation which is invariant to the usual geometric transformations, such as for example the Fourier-Mellin transformation invariant to rotations and changes in scale, described in the article by J. RUANAIDH et al, “Rotation, scale and translation invariant spread spectrum digital image watermarking”, Signal Processing, 6, 1998.
Such a watermarking method is of limited application since it does not make it possible to use the conventional spectral decomposition methods in the domain of the digital images, such as Fourier transforms, discrete cosine transforms, either global or by blocks, or discrete wavelet transforms.
Another technique is described in international patent application WO 97/43736 in which a fixed reference is inserted invisibly in the image at the time of insertion of the signature. This reference is then detected in order to re-fix the image before extraction of the signature.
However, this method requires the insertion of additional information imperceptibly, independently of the signature itself.
In addition, this reference can be detected intentionally with statistical methods and be intentionally erased.
SUMMARY OF THE INVENTION
The aim of the present invention is in a first aspect to propose a method of inserting and extracting a signature in a digital image which makes it possible effectively to know and reverse the geometric transformations applied to a digital image.
To this end, the method of inserting additional information, such as a digital signal, in a digital image, comprises the following steps:
detecting the points of interest in said image;
selecting a subset of points of interest adapted to define a geometric reference frame in said image;
calculating, for the geometric reference frame, reference information necessary for determining the geometric transformation applied to the image for a class of predetermined geometric transformations;
coding said reference information in a reference key; and
inserting the additional information in said digital image.
Correlatively, the method of extracting additional information, such as a digital signature, inserted in a digital image, comprises the following steps:
detecting the points of interest in said image;
selecting a subset of points of interest adapted to define a geometric reference frame in said image in accordance with selection criteria identical to those used during the insertion of the additional information in the original digital image;
decoding a reference key in order to extract reference information necessary for determining the geometric transformation applied to the original digital image;
calculating the parameters of the geometric transformation applied to the original image from said subset of points of interest and the reference information extracted;
re-fixing the digital image by applying an inverse geometric transformation determined from the calculated parameters; and
extracting said additional information from the re-fixed image.
Thus the image can be effectively re-fixed in space before the inserted signature is extracted, by means of selected points of interest and reference information calculated when the signature is inserted.
The means used for re-fixing the digital image are consequently dependent on the image itself and do not make it necessary to use additional information.
Such a method is robust to intentional attacks which seek to modify or destroy the signature inserted by means of the use of a reference key, possibly encrypted.
By definition, the points of interest are local characteristics of the image which convey significant information on the content of the image. This is a case mainly of the points where the bidimensional variations in the light intensity are great, such as for example corners.
Advantageously, it is shown that these points of interest, detected for example by a detector known as a Harris detector, are relatively invariant to geometric transformations such as rotation, translation, change in scale and also change in point of view, and are consequently particularly well suited to defining a geometric reference frame which can be used for re-fixing the digital image before the signature is extracted.
In addition, such a method is robust to the conventional compression techniques used for c
Au Amelia M.
Canon Kabushiki Kaisha
Fitzpatrick ,Cella, Harper & Scinto
Kim Charles
LandOfFree
Method of inserting and extracting a digital signature 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 of inserting and extracting a digital signature, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of inserting and extracting a digital signature will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3196905