U.S. patent application number 11/928831 was filed with the patent office on 2008-03-06 for system and method for embedding additional information in video data.
This patent application is currently assigned to International Business Machines Corporation. Invention is credited to Koichi Kamijo, Norishige Morimoto, Shuichi Shimizu.
Application Number | 20080056671 11/928831 |
Document ID | / |
Family ID | 17075152 |
Filed Date | 2008-03-06 |
United States Patent
Application |
20080056671 |
Kind Code |
A1 |
Kamijo; Koichi ; et
al. |
March 6, 2008 |
SYSTEM AND METHOD FOR EMBEDDING ADDITIONAL INFORMATION IN VIDEO
DATA
Abstract
A method and system for embedding an electronic watermark
directly in an MPEG stream is provided. An intra-macroblock of an
I-frame or of a P or B-frame is detected in an MPEG stream, and
following the detection of an intra-macroblock of the I-frame or of
the P or B-frame, data for one macroblock are extracted from the
MPEG stream and buffered. An embedding pattern is embedded in the
buffered macroblock without changing the length of VLC, and the
resultant macroblock is returned to the MPEG stream. More
specifically, a DC factor is extracted from the buffered
macroblock, and a pseudorandom number is generated to produce an
embedding pattern. Then, whether the bit length of the DC factor
will not be changed by embedding the obtained embedded pattern is
determined. When the bit length of the DC factor not be changed,
the embedding pattern is embedded in the buffered macroblock. When
the bit length will be changed, whether 1/2 of the embedding
pattern can be embedded is determined. If possible, 1/2 of the
embedded pattern is embedded in the macroblock.
Inventors: |
Kamijo; Koichi;
(Yokohama-shi, JP) ; Shimizu; Shuichi;
(Yokohama-shi, JP) ; Morimoto; Norishige;
(Tokyo-to, JP) |
Correspondence
Address: |
FERENCE & ASSOCIATES LLC
409 BROAD STREET
PITTSBURGH
PA
15143
US
|
Assignee: |
International Business Machines
Corporation
|
Family ID: |
17075152 |
Appl. No.: |
11/928831 |
Filed: |
October 30, 2007 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
09763988 |
Jul 5, 2001 |
7295615 |
|
|
PCT/JP98/05355 |
Nov 25, 1998 |
|
|
|
11928831 |
Oct 30, 2007 |
|
|
|
Current U.S.
Class: |
386/260 ;
375/E7.001; 375/E7.018; 375/E7.089; 386/329 |
Current CPC
Class: |
G06T 2201/0052 20130101;
G06T 2201/0053 20130101; G06T 1/0035 20130101; H04N 19/467
20141101; H04N 21/8358 20130101; H04N 7/24 20130101; H04N 21/23892
20130101 |
Class at
Publication: |
386/094 |
International
Class: |
H04N 5/91 20060101
H04N005/91 |
Foreign Application Data
Date |
Code |
Application Number |
Aug 27, 1998 |
JP |
10-241493 |
Claims
1. A system for controlling the copying of digital data, said
system comprising: means for detecting CCI in input data; means
for, when said CCI is detected, detecting ECCI in said input data;
means for, when said ECCI is detected, inhibiting the copying of
the digital data; and means for, when said ECCI is not detected
permitting the embedding of said ECCI in said digital data and the
copying of the resultant digital data.
Description
CROSS-REFERENCE TO RELATED APPLICATIONS
[0001] This application is a divisional application of copending
U.S. patent application Ser. No. 09/763,988, which is a national
phase entry of International Application No. PCT/JP98/05355 having
an International Filing Date of Nov. 25, 1998, the contents of
which are hereby incorporated by reference in its entirety.
FIELD OF THE INVENTION
[0002] The present invention relates to a technique for the
embedding of additional information in a video data stream, and in
particular to a method and a system for electronically embedding
information in MPEG2 video data without changing the length of the
VLC (Variable Length Code) that constitutes the MPEG video data
portion.
BACKGROUND OF THE INVENTION
[0003] The data hiding technique (trademark of IBM) is known as a
technique for embedding invisible and inseparable data in still
picture data, and moving picture data or audio data. This technique
is also called electronic watermarking and has received attention
as a technique that can be used to protect copyrights for the
contents of multimedia applications. Recently, since digital data
have come to be widely employed for moving pictures which have in
turn been widely distributed, and since the need for copyright
protection has increased for the high value added contents of such
applications as movies, the use of data hiding for the control of
copying and replaying has been discussed. For embedding additional
information in a digital image, generally the information is
embedded in a baseband and then JPEG/MPEG compression is performed
on the resultant data. Here it should be noted that JPEG is used to
represent the Joint Photographic Coding Experts Group, and that
MPEG is used to represent the Moving Picture Experts Group.
[0004] An electronic watermarking technique for still picture data,
moving picture data or audio data is disclosed in, for example,
Japanese Patent Application No. Hei 8-159330 (Our docket No.
JA9-96-044), Japanese Patent Application No. Hei 8-273551 (Our
docket No. JA9-96-076), Japanese Patent Application No. Hei
8-348426 (Our docket No. JA9-96-090), Japanese Patent Application
No. Hei 9-088493 (Our docket No. JA9-97-045), Japanese Patent
Application No. Hei 9-248272 (Our docket No. JA9-97-156), and
Japanese Patent Application No. Hei 8-272721 (Our docket No.
JA9-96-074). In all the cited prior art, embedding of image data in
a baseband is assumed. Even for the embedding of the data contents
that are compressed by JPEG/MPEG, generally the data are embedded
in the baseband and then the resultant data are compressed, and the
data are not directly embedded in a MPEG2 video data stream. This
is because the length of the VLC (Variable length Code) would be
changed by directly embedding image data in a compressed MPEG2
video data stream, and thus a content packet can not be coped
with.
[0005] More specifically, since an AC or DC factor is changed using
the VLC, the demultiplexing and the multiplexing of the VLC is
required before and after the embedding process, so that the entire
length of a bit stream is changed. Therefore, a large buffer is
required and reconstruction of the bit stream is difficult.
Particularly when DVD data are formed into a packet having a fixed
length, the reconstruction of the packet is very difficult.
Proposed in Japanese Patent Application No. Hei 8-272721 (Our
docket No. JA9-96-074) is a method for directly embedding data in a
MPEG2 domain using the forward prediction and backward prediction
for a B or P frame. This method also has the same shortcomings in
that the length of the VLC is changed and moving picture data
having a few B or P frames can not be coped with.
PROBLEMS TO BE SOLVED BY THE INVENTION
[0006] It is, therefore, one object of the present invention to
provide a method and system for embedding additional information
directly in video data.
[0007] It is another object of the present invention to provide a
method and system for embedding additional information in a packet
of video data.
[0008] It is an additional object of the present invention to
provide a method and system for embedding additional information in
MPEG data without changing the length of the VLC.
[0009] It is a her object of the present invention to provide a
method and system for embedding additional information in a video
data stream without changing the overall length of the video data
stream.
[0010] It is still another object of the present invention to
provide a method and system for embedding additional information in
video data without a large buffer being required.
[0011] It is a still additional object of the present invention to
provide a method and a system, for embedding an electronic
watermark in MPEG data, that can cope with a moving picture having
a few B or P frames.
SUMMARY OF THE INVENTION
[0012] To achieve the above objects, to embed additional
information in video data, first a video frame is detected in the
video data, and data in a small domain are extracted from the
detected video frame and buffered. Additional information is
embedded in the buffered small domain without changing the length
of the video data stream, and the resultant small domain is
returned to the video data.
[0013] As for MPEG video data, an intra-macroblock of an I-frame or
of a P or B-frame is detected in an MPEG stream, and following the
detection of an intra-macroblock of the I-frame or of the P or
B-frame, data for one macroblock are extracted from the MPEG stream
and buffered. An embedding pattern is embedded in the buffered
macroblock without changing the length of VLC, and the resultant
macroblock is returned to the MPEG stream. More specifically, a DC
factor is extracted from the buffered macroblock, and a
pseudorandom number is generated to produce an embedding pattern.
Then, whether the bit length of the DC factor will not be changed
by embedding the obtained embedded pattern is determined. When the
bit length of the DC factor not be changed, the embedding pattern
is embedded in the buffered macroblock. When the bit length will be
changed, whether 1/2 of the embedding pattern can be embedded is
determined. If possible, 1/2 of the embedded pattern is embedded in
the macroblock.
[0014] To detect additional information in video data, first, a
video frame is detected in video data, and data in a small domain
are extracted from the detected video frame and buffered. Then, the
additional information in the buffered small domain is detected. To
detect an electronic watermark in an MPEG stream, first, an
intra-macroblock of an I-frame or of a P or B-frame is detected.
When the intra-macroblock of an I-frame or of a P or B-frame is
detected, data for one macroblock are extracted form the MPEG
stream and buffered. An embedding pattern that is generated using a
pseudorandom number is detected in the DC factor in the buffered
macroblock. It should be noted that the pseudorandom number used
for embedding and detection is generated by using a common secret
key M.
[0015] With this arrangement, additional information can be
embedded in the video data or can be detected therein without
changing the size of a data stream. In addition, additional
information can be embedded by an inexpensive method that does not
require a high-cost process such as DCT calculation. Since only a
small buffer is required, accordingly, any delay is small. Further,
this method can be applied to a packetized MPEG2 stream.
BRIEF DESCRIPTION OF THE DRAWINGS
[0016] FIG. 1 is a diagram showing example embedding patterns used
for the present invention.
[0017] FIG. 2 is a flowchart showing the embedding processing for
the present invention.
[0018] FIG. 3 is a table showing example values for Len(x).
[0019] FIG. 4 is a block diagram showing an embedding apparatus
according to the present invention.
[0020] FIG. 5 is a flowchart showing the processing for an example
copying control system when the present invention is applied for
digital video.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
Preferred Embodiment
[0021] An explanation will be given for the preferred embodiment
wherein the method of the present invention is employed to embed
additional information in MPEG video data. A system for
embedding/detecting additional information in an MPEG2 stream
mainly comprises two components: an embedding apparatus, and a
detection apparatus. The embedding apparatus and the detection
apparatus employ a common secret key M. A detailed explanation will
be given below for these two apparatuses.
[0022] Embedding Apparatus
[0023] First, an embedding method will be explained. In this
invention, as is shown in FIG. 1, there are four embedding patterns
(P0, P1, P2 and P3), which are embedded as macroblock (16.times.16
pixels) units in a DC luminance factor of MEG2 video data. For
example, to embed P0, the DC factor for Y0 of the luminance factor
of the macroblock is incremented by 1, the DC factor for Y1 is
decremented by one, and the DC factors for Y2 and Y3 are unchanged.
A P that corresponds to each macroblock is obtained using a
pseudorandom number M(i,j). That is, embedding pattern P, which is
allocated for slice_number=i(0 origin) and macroblock_number=j, is
obtained as P=PM(i,j). Further, for MPEG2, the DC factor that is
actually changed in the MPEG2 domain is .DELTA.(M(i,j),Y) because
of a difference in the luminance factors in the preceding sub-block
(Y0 for Y1 and Y3 in the preceding macroblock for Y0). The
embedding processing will now be described while referring to the
flowchart in FIG. 2.
[0024] First at step 100, an intra-macroblock of an I-frame or of a
P or B-frame is detected. When an MPEG2 video stream is detected in
a data stream and the intra-macroblock of the I-frame or of the P
or B-frame is detected, program control moves to step 200.
[0025] At step 200 data for one macroblock are buffered. Of the bit
stream for one macroblock, the portion extending from the DC factor
for Y0 to the DC factor for Y3 is stored. At steps 300 and 400, the
bit lengths for DDSL (Variable length code)+DDD
(dct_dc_size_luminance) before and after the embedding process are
calculated. When the bit lengths are equal, i.e., when for Y=0, 1,
2, 3 the following equation is established,
Len(D(i,j,Y))=(Len(D(i,j,Y)+.DELTA.(M(i,j),Y)) (1),
[0026] PM(i,j) is embedded in the macroblock. D(i,j,Y) represents a
difference in DC factors when slice_num=i, macroblock_num=j, and
subblock_num=Y; .DELTA.(k,Y) represents the increase in the factor
difference between the embedding pattern k and Y; and Len(x)
represents the bit length of DDSL+DDD when the DC factor difference
is x. Len(x) is shown in FIG. 3. The subblock in which
Len(0)=Len(1)=3 and the DC factor difference is "0" can be changed
to "1" or "-1" because the bit length will not be changed. However,
the subblock having the DC factor difference of "4" can not be
changed to "3" because the bit length will be changed. When
equation (1) is not established, program control moves to step
500.
[0027] When the embedding can not be performed because the bit
length differs before and after the embedding, at steps 500 to 700,
whether half of the quantity of information can be embedded is
determined. If possible, half of the quantity of information is
embedded. That is, when P1 in FIG. 1 can not be embedded, whether
or not P'.sub.1 or P''.sub.1 can be embedded is determined. If
either P'.sub.1 or P''.sub.1 can be embedded, it is embedded. In
other words, when for Y=0, 1, 2, 3 the following equation is
established, Len(D(i,j,Y))=Len(D(i,j,Y)+.DELTA.'(M(i,j),Y)) or
Len(D(i,j,Y))=Len(D(i,j,Y)+.DELTA.''(M(i,j),Y)) (2).
[0028] P'M(i,j) or P''M(i,j) is embedded in the macroblock.
.DELTA.'(K,Y) and .DELTA.''(K,Y) are increases in the factor
difference between the embedding pattern K at Y for P'.sub.K and
P''.sub.K. When equation (2) is not established, program control
moves to step 700, whereat the macroblock embedding is abandoned.
At step 800 a check is performed to determine whether the
intra-macroblock of the I-frame or of the P or B-frame has been
processed. When the process has not been completed, program control
returns to step 200.
[0029] FIG. 4 is a block diagram illustrating the arrangement of
the embedding apparatus. An input controller 100 receives data from
a data bus, such as an ATA, and stores them in a line buffer 200.
The line buffer 200 is used to store data received by the input
controller 100. An intra-macroblock detector 300 for an I-fame or
for a P or B-fame detects an MPEG2 video stream in data stored in
the line buffer 200, and further detects an intra-macroblock for an
I-frame or for a P or B-frame. A CCI detector 400 is used only when
the present invention is mounted in a data hiding detection chip
(DataHiding.TM. detection chip), and detects the CCI after the
I-frame is detected by the intra-macroblock detector 300. The CCI
detector 400 detects the CCI and issues an interrupt to a CPU. The
above described blocks 100 to 400 function as the data hiding
detection chip.
[0030] A macroblock buffer 500 is used to store data for one
macroblock. A DC luminance factor detector 600 detects a DC
luminance factor (Y=0,1,2,3) in a macroblock. A pseudorandom number
generator 700 generates a pseudorandom number (M(i,j)) for an
embedding pattern. Two bits are allocated for one macroblock. For
MP@ML (720.times.480), 720/16*490/16*2=2700 bits are generated. A
DC huffman table 800 (Table B.12 for ISO/IEC 31818-2) is used to
store a DC luminance factor for MPEG2. A DC huffman length
comparator 900 determines whether or not the bit length of DDSL+DDD
for the DC luminance factor will be changed by embedding data. A DC
re-embedding unit 1000 embeds a pattern that is obtained by the
pseudorandom number generator 700 when the comparator 900
determines that the bit length will not be changed. An output
controller 1010 returns the obtained macroblock to the bit stream
and outputs the resultant bit stream. When the output controller
1010 does not perform re embedding, the data received from the
input controller are transmitted unchanged. The above described
blocks 500 to 1010 have been newly added to the data hiding
detection chip.
[0031] Detection Apparatus
[0032] The detection processing is performed in the following
manner. When an I-frame is detected, assuming that DC(i,j,Y) is
defined as a DC luminance factor (i, j and Y being defined in the
same manner as for the embedding apparatus), that P(M(i,j),Y) is
defined as an embedding value of embedding pattern M(i,j) for Y,
and that .DELTA.DC(i,j)=DC(i,j0)-DC(i,j,1), M(i,j)=0
DC(i,j,0)-DC(i,j,2), M(i,j)=1 DC(i,j,1)-DC(i,j,0), M(i,j)=2
DC(i,j,2)-DC(i,j,0), MC(i,j)=3.
[0033] Then, when the embedding is not performed, it is expected
that the following expression will be established:
.SIGMA..sub.i.SIGMA..sub.j.DELTA.DC(i,j)/N-->0
[0034] (It should be noted that N represents the number of
ADC(i,j)s that are employed for the calculation of .SIGMA..SIGMA..)
When embedding is performed, the value approaches a specific
positive value A: .SIGMA.i.SIGMA.j.DELTA.DC(i,j)/N-->A>0.
[0035] When the standard deviation .sigma.of .DELTA.DC(i,j) is
.sigma..sup.2=.SIGMA.i.SIGMA.j(.DELTA.DC(i,j)-<.DELTA.DC(i,j)>).sup-
.2/N, and
[0036] when threshold T<z while
z=.SIGMA.i.SIGMA.j{DC(i,j)/(.sigma.*N),
[0037] the embedding is "Yes," and when T.gtoreq.z, the embedding
is "No."
Embodiment
[0038] The control of digital video recording and replaying has
been discussed by employing the data hiding technique. Of demands
from the film business and the public appliance business, the
implementation of a system is sought that can permit recording only
one time for charge broadcasting via a satellite or a cable (sea
DVD CYTWG DHSG CFP Ver 1.). A system for controlling the
frequencies for the copying of distributed digital data can be
proposed as one embodiment of the present invention. This system
can be provided by employing the following method.
[0039] When a recording device detects CCI=(1,0) (Copy Once) in
digital data, an examination is made to determine whether new CCI
(extended CCI, hereinafter referred to as ECCI) is present. When
the result "ECCI is not present" is obtained, ECCI is embedded and
recording is permitted. When the result "ECCI is present" is
obtained, recording is not permitted. When the method of the
present invention is employed for the embedding and the detection
of ECCI, the system that permits copying to be performed only once
can be provided. FIG. 5 is a flowchart showing example processing
performed by a system for the distribution of digital video
data.
[0040] At step 510 digital video data are received by an STB (Set
Top Box). When the data are transmitted to a recording device 570,
at step 520 CCI is detected, and when the CCI is detected at step
530 whether or not ECCI is present is determined. When ECCI is
present (YES), at step 560 recording is inhibited. When ECCI is not
present, at step 540 ECCI is embedded in the video data and at step
550 recording is performed.
[0041] The advantages of this system are:
[0042] 1. ECCI can be embedded in the MPEG2 domain in real time,
and a packet having a DVD format can be coped with;
[0043] 2. ECCI can be detected even in a baseband after the MPEG2
data have been developed, or in an analog domain for which DA
conversion has been performed;
[0044] 3. ECCI can be embedded directly in the baseband; and
[0045] 4. the accuracy of the ECCI detection is not deteriorated by
the occurrence of an error across a communication path, which is a
problem that afflicts the Token method disclosed in Japanese Patent
Application No. Hei 9-083493.
ADVANTAGES OF THE INVENTION
[0046] According to the present invention, an MPEG electronic
watermarking system can be provided whereby the size of a data
stream after additional information has been embedded is the same
as it was before the additional information was embedded. The
method of the present invention can be applied for a packet in an
MPEG2 stream, and can also be performed at a low cost, without a
high-cost process, such as DCI calculation, being required. Since
only a small buffer is required, any delay is accordingly small. In
addition, an electronic watermark can be detected even in a
baseband image for which MPEG2 data have been developed. Since the
information is embedded in the baseband image, the watermark can be
directly detected even when MPEG2 compression has been performed
for the image data. The method of the present invention does not
conflict with the conventional marking technique for an AC factor,
and can coexist with it.
* * * * *