Image analysis – Applications
Reexamination Certificate
1999-07-20
2004-11-30
Mehta, Bhavesh M. (Department: 2621)
Image analysis
Applications
C283S017000, C283S071000, C283S072000, C283S113000, C345S629000, C345S638000, C345S641000, C380S054000, C380S201000, C382S250000, C382S294000, C382S305000, C713S176000
Reexamination Certificate
active
06826289
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Technical Field of the Invention
The present invention relates to a system for changing an attribute of image by electronic watermarks, wherein the attribute of image with an inserted watermark is changed by adding another electronic watermark.
2. Description of the Prior Art
Recently, it has become important to prevent reproduction of image information. Therefore, Image information are accompanied very often by some kind of attributes which are identifiers to instruct commands such as “reproduction prohibited”, or “reproduction allowed only once”.
One of the conventional techniques for adding the attribute to image information is disclosed, for example, in JP 08275127 A, 1996 (Video signal processing apparatus). Referring to this document, video identifier, or video blanking identifier (VBI) is inserted in vertical blanking periods to control the reproduction of image information. VBIs are inserted in the 20-th vertical blanking period of the first field and 283-th vertical blanking period of the second field. Further, in effective video signal areas after horizontal synchronizing signals and color burst signals, 2 bit reference signal and 20 bit information signal are inserted. The two bits from the above-mentioned 20 bit information signal are used for copy generation management system (CGMS) which controls the reproduction of image information.
VBI and CGMS are applicable for analogue video signals in 525 line NTSC, 625 line PAL, and 1125 line High Definition Video System.
However, VBI can be easily eliminated by filtering, because VBI are inserted in fixed positions. Therefore, illegal reproduction can not be prevented perfectly.
Accordingly, visible or invisible electronic watermark becomes an alternative.
Here, the visible electronic watermark has an effect to appeal visually not to reproduce and misappropriate the displayed image information, because prescribed charactes and figures are added to image information by using the visible electronic watermark.
The visible electronic watermark is disclosed, for example, in JP 08241403 A, 1996 (Digital watermarking without color change in image). Referring to this document, only brightness is changed in picture elemnts corresponding to opaque part of the visible electronic watermark data. The brightness is determined by such parameters as color components, random numbers, and values of picture elements of the visible watermark data.
On the other hand, the invisible electronic watermark is recognized visually, because it is imbedded in image information without degrading the image quality.
For example, author of image information may be identified by the invisible electronic watermark. Further, the invisible electronic watermark may notify users that the image information is prohibited to reproduce, or actuate reproduction prohibition mechanism in VTR.
The invisible electronic watermark may be inserted, for example, in the least significant bit (LSB) in picture element, because the LSB does not affect so much the image quality. However, the LSB is eliminated easily by low pass filtering, or image information compression. Therefore, in this case, it becomes difficult to detect filtered or compressed invisible electronic watermarks.
Another electronic watermark is disclosed in Nikkei Electronics 1996, 4.22, No. 660, 13 p). Referring to this document, Image information is transformed in frequency space, and the electronic watermark is imbedded in the dominant frequency components. Therefore, the electronic watermark is not easily eliminated by filtering or image compression. Furthermore, when random numbers, for example, under the normal distribution, are employed, the electronic watermarks rarely interfere with each other, do not affect image quality so much, and are not easily destroyed.
Here, the electronic watermark by using the frequency transformation is explained, referring to
FIGS. 7
to
11
.
As shown in
FIG. 7
, original image
701
is transformed by discrete cosine transformation (DCT)
703
. “n” data f(1),f(2), . . . ,f(n) is chosen among the frequency components of which amplitudes are great. Further, an electronic watermark of which frequency components are w(1),w(2), . . . ,w(n) is assumed to follow the normal distribution with average 0 and variance 1. Watermark imbedding means
704
calculates for each “i”
F
(
i
)=
f
(
i
)+&agr;·|
f
(
i
)|·
w
(
i
),
where &agr; is a scaling factor
Then, output image
710
with watermark is obtained through inverse DCT means
709
.
The imbedded watermark is detected on the basis of known original image data and known w(i) where i=1 to n.
As shown in
FIG. 8
, object image
802
with w(i) is transformed by DCT
804
to obtain its frequency components F(1), F(2), . . . ,F(N). At the same time, original image
801
is transformed by DCT
803
to obtain its frequency components f(1), f(2), . . . ,f(N). Then, the watermark data W(i) is calculated. Here, W(i) is defined as (F(i)−f(i))/f(i). Further, statistical similarity C is defined as (W·w)/(WD·wD), where (W·w) is a scalar product, WD and wD are the absolute values of W and w, respectively. Finally, statistical similarity determination means
810
determines that electronic watermark is imbedded in object image
802
, when C is greater than a prescribed value.
The watermark detection by the statistical similarity requires the original image. Therefore, the author of the original image can use this method. However, general users of playback apparatus such as, for example, VTR, can not use this method. Accordingly, the following method is proposed for the general users of MPEG (Moving Picture Experts Group) system which handles compressed moving pictures, according to ISO (International Standardization Organization) protocol.
As shown in
FIG. 9
, in the above-mentioned method for MPEG system, electronic watermark is embedded in blocks which are the areas of 8 pixels×8 pixels divided from the original image.
At first, the original image is transformed by DCT
903
to obtain f(1),f(2), . . . ,f(n) where f(1),f(2), . . . ,f(n) are selected in order of amplitude of the obtained frequency components and f(1) is minimum. Again, electronic watermark is assumed to follow the normal distribution with average 0 and variance 1. Then, electronic watermark embedding means
905
calculates for each “i”
F
(
i
)=
f
(
i
)+&agr;·
avg
(
f
(
i
))·
w
(
i
)
Here, &agr; is a scaling factor, and avg(f (i)) is an average of |f (i−1)|, |f (i)|, and |f (i+1)|. In the following, avg(f (i)) is referred to as partial average. Then, MPEG data is obtained through quantization and coding.
The original image data is not required. Only the data w(i) is required, where “i” is 1 to n.
As shown in
FIG. 10
, after inverse quantization of MPEG data, F(1),F(2), . . . ,F(n) are selected in order of amplitude of the obtained frequency components, where F(1) is minimum. Then, electronic watermark data W(i) defined as F(i)/avg(F(i)) is calculated for each “i”. Further, WF(i) defined as a summation of W(i) over one frame (one picture) of the expanded MPEG image
1101
is calculated for each “i”. Next, statistical similarity C defined as (WF·w)/(WFD·wD), where (WF·w) is a scalar product, WFD and wD are the absolute values of WF and w, respectively. Finally, statistical similarity determination means
1109
determines that electronic watermark is imbedded in the expanded MPEG data, when C is greater than a prescribed value.
As shown in
FIG. 11
, electronic watermark is also detected from expanded MPEG image. After DCT of expanded MPEG data
1101
, F(1),F(2), . . . ,F(n) are selected in order of amplitude of the obtained frequency components of image data
1101
, where F(1) is minimum. Then, electronic watermark data W(i) defined as F(i)/avg(F(i)) is calculated for each “i”. Further, WF(i) defined as a summation of W(i) over one frame (one picture) of the original image is calculated for each “i”. Next, statistical similarity C defined a
NEC Corporation
Scully Scott Murphy & Presser
LandOfFree
System for changing attribute of image by electronic watermark does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System for changing attribute of image by electronic watermark, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System for changing attribute of image by electronic watermark will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3291773