JP4919213B2 - Digital watermark insertion method and detection method - Google Patents

Description

   The present invention relates to a digital watermark insertion method for inserting a digital watermark when an encoded signal such as an image or sound or the compression of the signal, and a detection method for detecting a digital watermark from a signal into which the digital watermark is inserted.

  A technique for inserting a digital watermark into a moving image or still image compressed with MPEG or JPEG is disclosed. Usually, a method for operating a quantized DCT coefficient, a method for operating a motion vector, a method for operating a quantization parameter, and the like are disclosed.

In addition, as the performance of digital watermark insertion, (1) a watermark insertion method (such as Non-Patent Document 1 below) in which the amount of information does not change even when a watermark is inserted, or (2) a watermark is applied to image data into which watermark is inserted. A technique (for example, Patent Document 1 below) is disclosed in which information is removed to restore the original compressed image before watermark insertion. Also, a method of manipulating quantization parameters and subband coefficients (for example, Patent Document 2 below) is disclosed for compressed audio information such as MPEG.
JP 2007-116513 JP JP 2006-330256 A "Basic Study of Digital Watermarking for Copying Digital Content" (May 18, 1999), Masanobu Mizutani, Hitoshi Ogawa, Yoshinori Izumi, published by IPSJ, 1-6

  However, in the case of the above (1), when the digital watermark is inserted, the encoded information originally encoded under the optimal encoding condition is falsified. Therefore, as the watermark data insertion amount increases, the falsification amount increases. As a result, there has been a problem that when the encoded data into which the watermark data is inserted is decoded, the image quality is deteriorated as compared with the case where the compressed image before the watermark data is inserted.

  In the case of (2), it is possible to return to the compressed image before the watermark information insertion. However, it is necessary to extract and remove the watermark information once, and the compressed image in which the watermark information is inserted is used as it is. In the case of decoding, there is a problem that image quality deterioration is large as in the case of (1). For this reason, it has been very difficult to insert a large amount of watermark information while maintaining the image quality.

  Similarly, audio information compressed by MPEG or the like has a problem that sound quality is deteriorated by inserting watermark information.

  The present invention has been made in view of the above-described prior art, and an object of the present invention is to provide a digital watermark insertion method and a detection method that completely maintain the quality of an image or sound.

  In order to achieve the above object, the present invention provides means for inputting data compression-encoded by conversion processing (hereinafter referred to as compression-encoded data), and encoding control parameter values used in the compression-encoded data. And a means for inserting digital watermark information into the compressed encoded data by re-encoding the compressed encoded data using different encoding control parameter values. There is one feature.

  Further, a means for inputting data into which digital watermark information compression-encoded by the conversion process is inserted (hereinafter referred to as compression-encoded data), redundancy from the encoding control parameter and conversion coefficient value obtained from the compression-encoded data. There is a second feature in that a digital watermark detection method is provided which includes means for detecting the property and obtaining the watermark information value inserted depending on the presence or absence of the redundancy.

  Furthermore, the third feature is that an increase in the number of bits is suppressed by inserting a digital watermark only when the number of non-zero transform coefficients is equal to or less than a predetermined number.

  According to the present invention, even if the compressed data with the digital watermark inserted is reproduced as it is, the quality of the original data is completely maintained (not impaired at all) and the increase in the code amount can be suppressed. It is possible to realize a watermark method and a detection method.

  Hereinafter, the present invention will be described in detail with reference to the drawings. FIG. 1 shows a watermark embedding procedure according to an embodiment of the present invention, and FIG. 2 shows a watermark detection procedure for the watermark embedding shown in FIG.

  First, a watermark embedding procedure will be described with reference to FIG.

  First, it is determined whether or not the control parameter is to be embedded (step S1). If the control parameter is to be embedded, it is determined whether the watermark bit to be embedded is “0” or “1” (step S1). Step S2). If the watermark bit to be embedded is “0”, the process ends without any processing. On the other hand, if the bit is “1”, the encoding control parameter value 1a used in the compressed encoded data 1 is converted (step S1). S3) Furthermore, the transform coefficient group 1b in the compressed encoded data 1 is recompressed using the changed encoding control parameter value 1a (step S4).

  Next, a watermark detection procedure will be described with reference to FIG.

  First, it is checked whether or not the control parameter of the compression-encoded data 2 is a watermark embedding target control parameter, and if it is a watermark embedding target control parameter, the content of the control parameter 2a is checked (step) S10). As a result, if there is no evidence that the control parameter 2a has changed, a bit “0” is output and the process ends (step S11). An example of means for determining whether or not the control parameter has changed will be described later. On the other hand, if there is a possibility that the control parameter 2a has changed, the redundancy check of the conversion coefficient 2b controlled by the control parameter is continued (step S12). As a result, if it is found that all the conversion coefficients 2b are redundant, "1" is output and the process is terminated (step S13). Conversely, if all the transform coefficients 2b are not redundant, the process ends without doing anything.

  Next, a second embodiment of the present invention will be described. In this embodiment, the present invention is specifically applied to an MPEG-2 video signal. However, the present invention is not limited to this, and can be applied to JPEG or the like.

  The MPEG-2 video signal has a hierarchical structure as shown in FIG. That is, one sequence 10 includes a sequence header (SH), one or more GOPs (Group of Picture), and a sequence end code (SEC). The GOP includes a header and one or more pictures 11, and the picture 11 a includes a header and one or more slices. Further, a slice header and one or more macro blocks (MB) exist inside the slice. The MB includes a header, a luminance block of 16 pixels × 16 lines, and two color difference blocks.

  Here, the slice header existing inside the slice includes a quantization scale code (hereinafter referred to as MQ) indicating a quantization step in addition to the slice start code. This MQ indicates a quantization scale for DCT coefficients in each MB until the MQ is instructed until the MQ is instructed when the MB is viewed in the scan order.

  On the other hand, as described above, MQ can also be included in the MB header. If MQ exists in the MB header, the MQ of the MB is indicated. Further, as in the case of the MQ in the slice header, unless the MQ appears in the subsequent MB in the scan order, the MQ is inherited and used.

  Note that the presence or absence of MQ in the MB can be determined by the Coded_MQ flag. If Coded_MQ flag == True, the MQ exists, that is, the MQ of the MB is the MQ specified by the MB. Will be used. On the other hand, when Coded_MQ flag == False, there is no MQ, so the MQ of the previous MB in the scan order is inherited.

  (1) Watermark embedding method

  In MPEG-1 / 2 video encoding, Intra-MB AC coefficients and Inter-MB DC / AC coefficients are reconstructed according to Equation (1). rec [m, n] indicates the reconstructed AC coefficient or DC coefficient.

Here, 0 ≦ m and n ≦ 7, and QT _INTRA and QT _INTER are default quantization tables for Intra-MB and Inter-MB, respectively. Further, qDCTC [m] [n] indicates a DCT coefficient, and sign (qDCTC [m] [n]) indicates −1, 0, or +1.

Next, MB (k) indicates the k-th MB in the slice (k = 1, 2,..., K _max ), and the control parameter Q (k) sets the CODD_MQ flag and MQ value in MB (k). I will show you. In addition,
When Q (k) = 0, there is no MQ value because CODD_MQ == False,
When Q (k)> 0, CODD_MQ == True and Q (k) indicates the MQ value of the MB.

First, the case of an I picture will be described. In particular, consider MB (k0) where Q (k0) = α × β (where α and β are integers and β ≠ 1). Clearly α and β are factors of Q (k0). When Q (k0) is a prime number, there is one factor, and only Q (k0) itself. In this case, it is assumed that MB (k0) is converted using the following algorithm 1.

Algorithm 1 (Q conversion of MB and coefficient)
1: Q (k0) ← α
2: for Y, Cb, and Cr channel do
3: for all nonzero qDCTC [m] [n] do
4: qDCTC [m] [n] ← βqDCTC [m] [n]
5: end for
6: end for

  Here, α is a new MQ value, and β is a value selected as a multiplier for the non-zero AC coefficient qDCTC of MB (k0). However, unless it is 1, the roles of α and β can be interchanged. To embed information, MB (k0) is either converted to embed “1” or “0” is embedded and left as it is. In particular, the factor β when the original Q (k0) is divisible is compensated by a multiplier of β (as written in the fourth line of Algorithm 1). Therefore, the image quality is completely preserved until the MB is converted by embedding the watermark in the scan order.

  The subsequent MB MQ is then updated to fully preserve the image quality at the slice level. MB (k0 + 1) needs to be replaced with α × β instead of α as Q (k0 + 1). Of course, this replacement is not necessary when Q (k0 + 1)> 0 (that is, a value that is not α × β) or when the MB is the last MB in the slice.

  On the other hand, for INTER-MB, the same processing can be performed by simply replacing the fourth line of algorithm 1 as follows.

  Here, x is a non-zero DCT coefficient qDCTC (which includes a DC coefficient), and y = (β−1) / 2. Since all operations are done at the integer level, y must be an integer. Therefore, it should be noted that the multiplier constant β is an odd number in INTER-MB.

  In the digital watermark embedding method of the present invention, it is possible to extend not only binary (that is, 1 bit of 0/1) but also multi-level embedding.

For example, for Q = 16,
When (α, β) = (16, 1) is “0” (or “00”)
“1” (or “01”) when (α, β) = (8, 2)
When (α, β) = (4, 4), “2” (or “10”)
When (α, β) = (2, 8), “3” (or “11”)
By doing so, it becomes possible to embed 4-value (or 2-bit) information in one quantization coefficient.

When this is generalized and written, when Q is prime factorized, β = c (≠ 1) with Nth power as a factor,
If (α, β) = (Q, 1)
When (α, β) = (Q / c, c)
If (α, β) = (Q / c ² , c ² )
...
If (α, β) = (Q / c ^N-1 , c ^N-1 )
“N” when (α, β) = (Q / c ^N , c ^N )
(N + 1) value can be embedded.
However, when Q = cN, the case where α = Q / cN = 1 is excluded (in this case, N values can be embedded).

  FIG. 5 is a flowchart showing the watermark embedding algorithm (slice unit) in MPEG2 described above, and since the details of the flowchart have been described above, description thereof will be omitted.

  (2) Watermark detection method

  In order to detect the embedded watermark data, the modified video data is parsed in the same manner as the MPEG data parse.

  First, the case of an I picture (INTRA-MB) will be described.

When MB reaches Q = α × β, bit “0” is extracted. On the other hand, when reaching the MB where Q = α, there is a possibility that the next two MBs.
(i) MB converted by Q = α × β, and has bit “1”.
(ii) Originally MB with Q = α.

  Thus, in order to overcome the problem that one MB has two meanings, the remainder mod (x, β) when dividing each non-zero AC coefficient x by β is considered. If the result is 0 for all x, that is, if x is divisible by β (mod (x, β) = 0), it is determined that the case is (i), and bit “1” is extracted. On the other hand, if not, it is determined that the case is (ii) and nothing is taken out. Note that it is rare that all non-zero AC coefficients in MB are divisible by a specific factor β. As described above, it is possible to determine whether the converted MB has “1” or not.

If you use algorithm 2 below, you can restore the converted MB to its original state.

Algorithm 2: Repair converted MB
1: Q ← α × β
2: for Y, Cb, and Cr channel do
3: for all nonzero qDCTC [m] [n] do
4: qDCTC [m] [n] ← qDCTC [m] [n] / β
5: end for
6: end for

  Thereafter, the subsequent MB is scanned for the MQ value to be encoded next. If it is α × β, it is determined that MQ (k) was not originally encoded, so the CODD_MQ flag becomes false to restore it. Otherwise, we originally had an encoded MQ.

The above discussion can be applied to INTER-MB with simple modifications. In particular, the condition mod (x, β) = 0 in (i)
mod (x−y, β) = 0 if x> 0
mod (x + y, β) = 0 otherwise.
And just replace it. This is derived from the formula (1). Also, when restoring the INTER-MB that has been converted, the fourth line of Algorithm 2
(x−y) = β if x>0;
(x + y) = β otherwise,
Just replace it with

  FIG. 6 is a flowchart showing the above-described watermark detection algorithm (in slice units) from MPEG2. The details of the flowchart have been described above, and a description thereof will be omitted.

  Next, a specific example of watermark embedding and watermark detection will be described. Specifically, FIG. 7 shows an example of watermark embedding / detection using an I picture.

  The upper part of FIG. 7 represents the original MQ and DCT coefficients, and the lower part represents the corrected MQ and DCT coefficients. The column of cells on the left represents Q (k). The value of Q (k) is stored when Q (k)> 0, and X is stored otherwise. MB (k) in which X is stored inherits the Q (k) value of the preceding MB (k).

  First, the embedding process will be described. Here, it is assumed that a digital watermark (binary message) is embedded using all MB (k) satisfying Q (k) = α × β = 4 × 2 = 8 in the upper stage.

  If Q (k0) is converted to store bit “1” in MB (k0), Q (k0) of the lower MB (k0) becomes Q (k0) = α = 4, And all non-zero AC coefficients in the DCT coefficients are multiplied by a factor β (= 2). In this way, by multiplying the factor, the DCT coefficient increases and the bit amount also increases, but the DCT coefficient after inverse quantization is the same as the original, and it can be seen that the image quality itself is completely preserved. The Q (k0 + 1) of the next MB (k0 + 1) is Q (k0 + 1) = 10 and not 8. Therefore, the lower Q (k0 + 1) may remain at 10.

  Similarly, when Q (k0 + 2) = 8 and bit “1” is stored here, the same processing as Q (k0) is performed and Q (k0 + 2) = 4. Since Q (k0 + 3) in the next MB (k0 + 3) is X (= 0), bit “0” is stored here. If X (= 0) is moved to the lower stage, Q (k0 + 2) = 4 is inherited and treated as Q (k0 + 3) = 4, so Q (k0 + 3) is Can be changed from X to 8. Further, Q (k0 + 4) of the next MB (k0 + 4) is Q (k0 + 4) = 11 and is not 8. Therefore, Q (k0 + 4) in the lower stage may be 11 as well.

  Next, the detection process will be described. In the lower cell group in FIG. 7, since the digital watermark (binary message) detector extracts Q (k0) = α = 4, MB (k0) is recognized as the first carrier. Since all non-zero AC coefficients are divisible by β (= 2), the detector extracts bit “1”, returns MB (k0) to the original state, and moves to the next check of Q (k0 + 1). . Since Q (k0 + 1) = 10 (that is, not 8), it is recognized that no watermark is embedded in MB (k0 + 1).

  Next, MB (k0 + 2) is also recognized as the first carrier like MB (k0), and if all non-zero AC coefficients are divisible by β (= 2), bit “1” is extracted. . Therefore, the detector returns MB (k0 + 2) to the original state, and checks Q (k0 + 3). Here, since Q (k0 + 3) = 8, MB (k0 + 3) is recognized as the MB that has been taken over from the previous state. Therefore, MQ (k0 + 3) = 0 is set. Since Q (k0 + 4) = 11 is not 8, it is recognized that no watermark is embedded in MB (k0 + 4). Q (k0 + 4) remains unchanged.

  The same method can be used for INTER. However, since β must be an odd number, let us consider the case where α = 3 and β = 3. The details are shown in FIG.

First, the watermark data embedding process will be described. Here, it is assumed that a digital watermark (binary message) is embedded using all MB (k) that satisfy Q (k) = α × β = 3 × 3 = 9.
If conversion is performed to store bit “1” in MB (k0), Q (k0) = α = 3, and the following factors are used for all non-zero AC coefficients x. An operation is performed.
βx + (β-1) / 2 or 3x + 1 (x> 0)
βx-(β-1) / 2 ie 3x-1 (other)

  By performing such an operation, the coefficient increases and the bit amount also increases, but the DCT coefficient after inverse quantization is the same as the original, and it can be seen that the image quality itself is preserved.

  As the next coefficient, Q (k0 + 1) in MB (k0 + 1) must be restored to its original value, but another quantization parameter originally in the form of Q (k0 + 1) = 10 Since it is encoded as, it can be left as it is.

  Similarly, when Q (k0 + 2) = 9 and bit “1” is stored here, the same processing as Q (k0) is performed and Q (k0 + 2) = 3. Since Q (k0 + 3) = 0 in the subsequent MB (k0 + 3), it will be treated as Q (k0 + 3) = 3 as it is. Therefore, this must be restored. Therefore, Q (k0 + 3) is 9 from X.

Next, watermark data detection processing will be described. Focusing on the lower cell group in FIG. 8, since the digital watermark (binary message) detector has Q (k0) = α = 3, MB (k0) is recognized as the first carrier. All non-zero DCT coefficients are
mod (x − y, β) = 0 (x> 0)
mod (x + y, β) = 0 (Other)

Where y = (β-1) / 2
(The above three formulas are taken as formula (3))
Therefore, the detector extracts bit “1” and returns MB (k0) to the original state. Since Q (k0 + 1) = 10, MQ (k0 + 1) = 0 is not set.

  Next, MB (k0 + 2) is also recognized as the first carrier like MB (k0), and if all non-zero DCT coefficients satisfy equation (3), bit “1” is extracted. . Therefore, the detector returns MB (k0 + 2) to the original state and checks Q (k0 + 3). Here, since Q (k0 + 3) = 9, MB (k0 + 3) is recognized as the MB that has been taken over from the previous state. Therefore, MQ (k0 + 3) = 0 is set. Since MB (k0 + 4) = 11 and not 9, it remains as it is.

  (3) Method of suppressing the increasing bit amount

  Changing the MQ by embedding a watermark increases all non-zero AC coefficients. The AC coefficient encodes the number of zeros (0 run) and the magnitude (level) of the coefficient until the AC coefficient appears in the scan order. MPEG-2 run and level encoding bits are shown in FIG. As shown in FIG. 4, the bit amount generally increases as the level increases. Therefore, there arises a problem that the total bit amount is increased by embedding the watermark. In order to suppress the increase in the overall code amount with respect to the length of the bit message (digital watermark) that can be embedded, the following three methods can be taken.

  (a) A method using only a small value (eg, β = 2 or 3) as a factor.

  By performing conversion (watermark embedding) only when the factor is a small value, each non-zero DCT coefficient increases only by a factor of two or three, so that an increase in code amount for the coefficient can be minimized.

  (b) When embedding a watermark, instead of using all MB (k) with Q (k)> 0, this method is applied only when the number of non-zero DCT coefficients is less than a certain constant. Can be suppressed. For example, when comparing the case of 10 nonzero coefficients with the case of 5 cases, the conversion is performed when there are 5 non-zero coefficients, that is, when the number of nonzero coefficients is smaller than when the number of nonzero coefficients is 10 This is because the number of DCT coefficients that increase overall decreases when only watermark embedding is supported.

(c) Using a method called Matrix Encoding (ME), it is possible to suppress the amount of code that increases with respect to the length of the bit message (digital watermark). Here, if Matrix Encoding is written as ME (j) using a certain integer j, this means that if at most one of 2 ^j -1 locations changes, this change information is used to make j bits It is a method of holding information. Here, the location means the MB where Q> 0, and the correction means the change of MB. Since the number of MBs to be changed is reduced, the file size increment can be made smaller.

  Details of ME (j) are as follows.

For example, consider the case of j = 3. If bit messages are w = w1, w2, and w3, the target bit embedding location (x) is 2 ³ −1 = 7 locations. Let this be x1, x2,..., X7. An example is shown in FIG.
Here, it is assumed that the relationship between x and w is as shown in FIG. 9 and that x1 = x2 =... = X7 = 0 originally. At this time, if it is desired to set (w1, w2, w3) = (0, 1, 0), x2 = 0 → 1, and similarly (w1, w2, w3) = (1, 1, If you want to set it to 0), set x4 = 0 → 1. Thus, by changing only one place, information of three bits can be expressed.

  On the contrary, when detecting, it is only necessary to detect a bit that changes at only one place among x1 to x7. For example, if x3 = 1, it can be seen that (w1, w2, w3) = (0, 0, 1).

  The above Matrix Encoding (ME) method is disclosed in, for example, “Ron Crandall,“ Some Notes on Steganography ”, 1998”.

  (4) Watermark embedding during encoding (compression)

  Up to this point, the watermark embedding method for a compressed video signal has been described. However, it is also possible to perform watermark embedding simultaneously with encoding. The method is the same as the watermark embedding method in the compressed video signal.

  For example, when this method is applied at the time of MPEG-2 video encoding, the quantization parameter MQ is calculated for each macroblock, but the MQ of a macroblock in which “0” is embedded as a digital watermark may be α × β. On the other hand, for the macroblock in which “1” is embedded, MQ is set to β, and all the non-zero coefficients of the DCT coefficients belonging to it are set to be a multiple of β. In order to make all the DCT coefficients β times, it can be coped with by temporarily setting MQ = α × β and calculating the DCT coefficients, and then setting MQ = β to multiply all the DCT coefficients by β.

  (5) Embedding / detecting audio signals

  Although the MPEG-2 video signal has been described above, these processes can be used not only for MPEG-2 video but also for audio encoding such as MP3.

  For example, considering that this method is applied to MP3, the relationship between the scale factor and the MDCT coefficient in MP3 reconstruction is as follows.

Rec [] = c × (MDCT coefficient) ^4/3 * 2 ^-scalefactor

  (C is a constant)

  Here, the condition for the MDCT coefficient to be a multiple of the original number without changing the value of Rec [], that is, while maintaining the sound quality, while increasing the scale factor is when MDCT is 4 times. Yes, at this time the scale factor should be +8. Similarly, when MDCT is 8 times, the scale factor may be +16.

  On the other hand, the condition for MDCT coefficient to be a multiple of the original number while increasing the scale factor while maintaining sound quality to some extent, even if not perfect, is MDCT: Scale factor is +3 when the factor is twice, MDCT : In case of 3 times, the scale factor should be +5, ...

  Based on the above, a watermark embedding method will be described.

  First, do nothing if you want to fill "0" in a certain scale factor. On the other hand, if it is desired to fill in “1”, if the scale factor is a certain value (for example, α), the MDCT coefficient may be multiplied by 4 after adding +8. In this way, it is possible to embed a watermark while maintaining the sound quality. In the case of MP3, since the scale factor is encoded for each subband, it is not necessary to consider the influence on the subsequent frames in the scan order as in the case of MPEG-2 Video.

  Next, detection of watermark data. If the scale factor is α, it is determined that the watermark “0” is filled. If the scale factor is α + 8, the MDCT coefficients are all multiples of 4. Only in some cases, it is determined that the watermark “1” is buried.

  If the sound quality may be affected to some extent, it is also possible to take the method of increasing the scale factor +3, MDCT: 2 times, and increasing the scale factor +5, MDCT: 3 times. .

It is explanatory drawing which shows the outline | summary of the process sequence (watermark embedding) of one Embodiment of this invention. It is explanatory drawing which shows the outline | summary of the process sequence (watermark detection) of one Embodiment of this invention. It is explanatory drawing of this encoding structure. It is a figure which shows a part of video run level table | surface of MPEG2. It is a flowchart which shows the outline | summary of the process sequence (watermark embedding: slice unit) of 2nd Embodiment of this invention. It is a flowchart which shows the outline | summary of the process sequence (watermark detection: slice unit) of 2nd Embodiment of this invention. It is explanatory drawing of a specific example of embedding and detection (intra) of watermark information. It is explanatory drawing of a specific example of embedding and detection (inter) of watermark information. It is explanatory drawing of Matrix Encoding.

Explanation of symbols

  1, 2... Compression encoded data, 1a, 2a... Control parameters, 1b, 2b... Transform coefficient group, 10.

Claims (23)

Means for inputting data that has been compression-encoded by the conversion process (hereinafter referred to as compression-encoded data);
Means for inserting digital watermark information into the compressed encoded data by re-encoding the compressed encoded data using an encoding control parameter value different from the encoding control parameter value used in the compressed encoded data An electronic watermark insertion method characterized by comprising: The digital watermark insertion method according to claim 1,
The digital watermark insertion method, wherein the compression encoded data is an image signal compressed by JPEG or MPEG. The digital watermark insertion method according to claim 1,
The digital watermark insertion method, wherein the compression-encoded data is an audio signal compressed by MPEG. In the digital watermark insertion method according to any one of claims 1 to 3,
The encoding control parameter value is a quantization step value;
The transform coefficient obtained by partially decoding the input compressed encoded data is re-quantized using a quantization step value different from the quantization step value at the time of compression encoding. Digital watermark insertion method. In the digital watermark insertion method according to claim 4,
The transform coefficient obtained by partially decoding the input compressed encoded data is requantized using a quantization step value finer than the quantization step value at the time of compression encoding. Digital watermark insertion method. In the digital watermark insertion method according to claim 5,
The fine watermark step value is a value obtained by dividing a quantization step value at the time of compression encoding by a certain number. In the digital watermark insertion method according to claim 4,
A digital watermark insertion method, wherein a digital watermark is inserted so that all the transform coefficients quantized by the quantization step value have a certain condition. The digital watermark insertion method according to claim 7,
The condition is an electronic watermark insertion method in which the remainder when all the transform coefficients are divided by a certain number is equal. The digital watermark insertion method according to any one of claims 6 to 8,
As a data embedding method, a digital watermark insertion method is characterized in that when the embedded data is “1”, requantization is performed under the above conditions, and when the embedded data is “0”, requantization is not performed. The digital watermark insertion method according to any one of claims 4 to 9,
An electronic watermark insertion method, wherein an electronic watermark is inserted only when the number of non-zero transform coefficients is equal to or less than a predetermined number, and an increase in the number of bits is suppressed. The digital watermark insertion method according to any one of claims 4 to 10,
An electronic watermark insertion method characterized by suppressing an increase in the number of bits by applying Matrix Encoding to information on a portion where a watermark is embedded. In the digital watermark insertion method according to claim 4,
As the condition, a value obtained by adding a number to the number of quantization steps is added to the audio data encoded by MPEG as a new number of quantization steps, and requantization is performed using the number of quantization steps. A digital watermark insertion method characterized by In the digital watermark insertion method according to claim 5,
The fine quantization step value is a plurality of values obtained by dividing the quantization step of the input compression encoded data by some number, and multi-value is obtained by requantizing using the quantization step value. An electronic watermark insertion method characterized by inserting digital watermark information. The digital watermark insertion method according to claim 13,
As the data embedding method, when the embedded data is n bits, the quantization step value of the input compression encoded data is factorized to be decomposed into 2 ⁿ −1 factors, and the quantization steps corresponding to these An electronic watermark insertion method characterized in that embedded data is determined for a value. In the digital watermark insertion method according to claim 5,
The fine quantization step value is a value obtained by prime factorizing the quantization step of the input compression-coded data and dividing it by 2 or more and the minimum prime number, and requantizes using the quantization step value. An electronic watermark insertion method characterized by suppressing an increase in data amount after insertion of electronic watermark information. The digital watermark insertion method according to claim 7,
As the condition,
The same quantization step is used for MB (A) and MB (B) in one macroblock (MB (A)) and the next macroblock (MB (B)) in the scan order. If the quantization step value of MB (B) was not encoded,
When embedding a digital watermark in MB (A), when changing the quantization step value to MB (A),
An electronic device characterized in that the quantization step value of the original MB (A) is encoded as the quantization step value of the MB (B) to prevent the quantization step value changed by the watermark embedding from being inherited. Watermark insertion method. The digital watermark insertion method according to claim 7,
An electronic watermark insertion method characterized in that, as the condition, a method for changing a DCT coefficient value at the time of watermark insertion is adaptively switched based on a difference in quantization method between an Intra macroblock and an Inter macroblock. Means for inputting data into which digital watermark information compression-encoded by the conversion process is inserted (hereinafter, compression-encoded data);
An electronic watermark detection method comprising means for detecting redundancy from an encoding control parameter and transform coefficient value obtained from the compressed encoded data, and obtaining a watermark information value inserted depending on the presence or absence of the redundancy . The digital watermark detection method according to claim 18,
The encoding control parameter value is a quantization step value;
An electronic watermark detection method, wherein a watermark information value is determined based on whether or not all transform coefficient values obtained by partially decoding the input compression-coded data are divisible by a specific value. The digital watermark detection method according to claim 19,
A digital watermark detection method, wherein multi-value detection is performed based on the magnitude of the specific value. The digital watermark detection method according to claim 18,
The compression-encoded data is compression-encoded audio data;
A watermark information value is determined based on whether or not a transform coefficient value obtained by partially decoding input compression-encoded audio data is changed depending on a quantization step number different from the input quantization step number. A digital watermark detection method characterized by The digital watermark detection method according to claim 21,
An electronic watermark detection method characterized in that a smaller quantization step number than an input quantization step number is used as the different quantization step number. Means for obtaining an encoding control parameter value for compressing and encoding input data;
Means for obtaining a different encoding control parameter for the encoding control parameter value in accordance with a value to insert a digital watermark;
Means for inserting a digital watermark by compressing the input data using the determined new encoding control parameter;
Means for outputting compressed encoded data obtained by the means.

Images