Image analysis – Applications
Reexamination Certificate
1997-12-19
2002-09-17
Tran, Phuoc (Department: 2621)
Image analysis
Applications
C382S233000, C382S250000, C382S251000
Reexamination Certificate
active
06453053
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
This invention relates to a system for processing a digital image, and more particularly to a system for embedding electronic watermark data as identification data having special information into a digital image.
2. Description of the Related Art
In recent years, illegal duplication of a digital image has become a serious problem.
In order to prevent illegal duplication, a system has been proposed wherein digital image data are ciphered such that only a reproduction system having a legal decipherment key can reproduce the ciphered digital image data. However, once the cipher is deciphered, later duplication cannot be prevented.
Thus, in order to prevent illegal use and duplication of a digital image, a method has been proposed wherein electronic watermark data which are special information are embedded into the digital image itself.
As such electronic watermark data for a digital image as described above, two kinds of electronic watermark data of visible electronic watermark data and invisible electronic watermark data are available.
The visible electronic watermark data are used such that a special character or symbol or the like is combined with an image so that it may be visually recognized. While the visible electronic watermark data deteriorate the picture quality, they have an effect of visually appealins to a user of a digital image for prevention of illegal use of the digital image.
An example of embedding of visible electronic watermark data is disclosed in Japanese Patent Laid-Open Application No. Heisei 8-241403. According to the method disclosed in the document just mentioned, visible electronic watermark data are combined with an original image such that only the brightness components of those pixels which correspond to an opaque portion of the electronic watermark data are varied while color components of the pixels are not varied. In this instance, the scaling value by which the brightness components of pixels are varied is determined based on color components, random numbers, values of pixels of the electronic watermark data and so forth.
On the other hand, the invisible electronic watermark data are embedded into an image so that the picture quality may not be deteriorated. Since the invisible electronic watermark data little deteriorate the picture quality, they are characterized in that they cannot be recognized visually.
However, if special information which allows identification of an author is embedded as the electronic watermark data, then also after illegal duplication is performed, the author can be specified by detecting the electronic watermark data. On the other hand, if duplication inhibition information is embedded, then, for example, when a reproduction apparatus detects the duplication inhibition information, it is possible to notify the user that duplication of the data is inhibited or to render a duplication prevention mechanism in the reproduction apparatus operative to limit duplication of the data by a video tape recorder (VTR) or the like.
One available methods of embedding invisible electronic watermark data into a digital image is to embed special information as electronic watermark data into a portion of pixel data such as a least significant bit (LSB) which has a comparatively little influence on the picture quality. However, where this method is applied, electronic watermark data can be removed readily from an image. For example, if a low-pass filter is used, then information of the LSB of a pixel is lost. Further, electronic watermark data are lost also by image compression processing since the image compression processing reduces the data amount by removing information at such a portion which has a comparatively little influence on the picture quality as mentioned above. Consequently, it is difficult to re-detect the electronic watermark data.
Another method is disclosed in Japanese Patent Laid-Open Application No. Heisei 6-315131.
In the method of the document just mentioned, a region which can be replaced, upon reproduction, by a peripheral region without causing deterioration of an image is detected using a correlation of images of successive frames, and the level of the region of an object of conversion is converted to embed particular information. In this method, upon reproduction, the region in which identification data are embedded is specified using the signal missing portion and the replacement information, and the portion is corrected to restore the image.
A further method is disclosed in Japanese Patent Laid-Open Application No. Heisei 5-30466 wherein a video signal is frequency converted and information having a frequency signal lower than a frequency band of the video signal after the frequency conversion is embedded into the video signal. In this method, the original video signal is extracted using a high pass filter while embedded identification data are extracted using a low pass filter.
As another method which involves frequency conversion of an image, a method has been proposed wherein an image is frequency converted and electronic watermark data are embedded into a region of the video signal after the frequency conversion in which strong frequency components are included (
Nikkei Electronics
. Apr. 22, 1996, No. 660, p.13).
In this method, since electronic watermark data are embedded into frequency components, the electronic watermark data are not lost even by image processing such as compression processing or filtering. Further, random numbers which exhibit a normal distribution are adopted as the electronic watermark data to prevent interference between electronic watermark data and make it difficult to destroy electronic watermark data without having a significant influence on the entire image.
According to the electronic watermark data embedding method in the method described above, an original image
501
is converted into frequency components using a DCT (discrete cosine transform) transformer
502
as seen in FIG.
5
. Then, n data which exhibit comparatively high values in a frequency region are selected and individually represented by f(
1
), f(
2
), . . . , f(n), and electronic watermark data
503
(w(
1
), w(
2
), . . . , w(n)) are selected from within a normal distribution whose average is 0 and whose dispersion is 1. Then, F(i)=f(i)+&agr;×|f(i)|×w(i) is calculated for each i by an electronic watermark data insertion element
504
. Here, &agr; is a scaling factor. Finally, DCT transform coefficients including those frequency components in which f(i) are replaced by F(i) are inverse transformed by an inverse DCT transformer
509
to obtain an image
510
in which the electronic watermark data is embedded.
Electronic watermark data is detected using the following method. In the detection method, an original image and electronic watermark data candidates w(i) (where i=1, 2, . . . , n) must be known.
Referring to
FIG. 6
, an original image
601
and an electronic watermark data containing image
602
are converted into frequency components using DCT transformers
603
and
604
, respectively. In a frequency region, values of factors corresponding to f(
1
), f(
2
), . . . , f(n) in which electronic watermark data are embedded are represented by F(
1
), F(
2
), . . . , F(n), respectively. An electronic watermark data extractor
605
extracts electronic watermark data W(i) based on f(i) and F(i) by calculation of W(i)=(F(i)−f(i))/f(i). Then, an inner product calculator
608
calculates, based on electronic watermark data
609
, a statistical similarity between w(i) and W(i) in accordance with C=W×w/(WD×wD) using inner products of vectors. Here, W=(W(
1
), W(
2
), . . . , W(n)), w=(w(
1
), w(
2
), . . . , w(n)), WD=absolute value of the vector W, and wD=absolute value of the vector w. A statistical similarity discriminator
610
discriminates, when the statistical similarity C is higher than a particular value, that the pertaining electronic watermark
Micheal Best & Friedrich LLC
NEC Corporation
Tran Phuoc
Whitesel J. Warren
LandOfFree
Identification data insertion and detection system for... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Identification data insertion and detection system for..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Identification data insertion and detection system for... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2906098