CN1765116B - Electronic watermark embedding device and method thereof, and electronic watermark detection device and method thereof - Google Patents

Abstract

A digital watermark embedding apparatus for embedding digital watermark information by superimposing macroblock patterns on an image is disclosed. The digital watermark embedding apparatus: spreads input digital watermark information to obtain an embedding series having a length corresponding to a number of pixel blocks of a macroblock pattern; selects at least a frequency from among predetermined plural frequencies according to a term value of the embedding series corresponding to a position of a pixel block in the macroblock pattern, and sets a waveform pattern corresponding to the selected frequency as a pixel of the pixel block; and amplifies, with an embedding strength value, the macroblock pattern on which a waveform pattern is superimposed on each pixel block by the waveform pattern setting unit, and superimposes the macroblock pattern on an input image like a tile.

Description

Electronic watermark embedded device and method thereof and digital watermark detection and method thereof

Technical field

The present invention relates to other supplementarys are not embedded in the contents such as image, video by the people with perceiveing, and read the eletric watermark technology of this supplementary.Nowadays, copyright protection/management system, and relevant with the content service system that provides of content is provided the eletric watermark technology.

Background technology

Image, video, and the such content of audio frequency when circulation, for realize purposes such as content recognition/management, copyright protection/management, relevant information provide, have using not as the people perceive in content the method for the eletric watermark technology of other information of embedding.Generally speaking, eletric watermark detected object image is subjected to geometric transformation and produces geometry deformation, therefore need have the eletric watermark technology of anti-sex change to this geometry deformation.

As the prior art that detects described geometric transformation and proofread and correct, the technology of being put down in writing just like patent documentation 1.In the technology that patent documentation 1 is put down in writing, differently embed the pattern that usefulness is proofreaied and correct in geometric transformation with embedded images, utilize this pattern detection to go out geometric transformation.

Patent documentation 1: the spy opens flat 11-355547 communique

Summary of the invention

But in described prior art, be different with embedded images, proofread and correct the problem that produces interference between the information with the signal of pattern and embedded images so can appear at geometric transformation because the pattern of usefulness is proofreaied and correct in geometric transformation.Therefore, if improve just necessary its pattern signal of strengthening of precision that geometric transformation is proofreaied and correct, but if make the image quality deterioration keep necessarily just must weakening embedded images.And, on the contrary,, the reinforcement embedded images uses signal if just must weakening the geometric transformation correction, and in any case, the accuracy of detection that embeds information all can descend.

The present invention produces in view of described problem just, its purpose is to provide a kind of and can resembles the prior art the independent additional geometric transformation different with electronic watermark information and proofread and correct the signal of usefulness, and carries out the eletric watermark embedded technology and the eletric watermark detection technique of geometric transformation correction.

Described problem can solve by the eletric watermark device, this eletric watermark device comprises a plurality of macro block patterns that contain the block of pixels of a plurality of pixels and embeds electronic watermark information by overlapping on the image of being imported, it is characterized in that, have: the electronic watermark information expanding element, it obtains the corresponding embedding sequence of block of pixels number in length and the described macro block pattern by the electronic watermark information that expansion is imported; The waveform patterns setup unit, it is worth according to the item with the corresponding embedding sequence in position of a block of pixels in described macro block pattern, from predetermined a plurality of frequencies, select at least one frequency, will be set at pixel in the described block of pixels with the corresponding waveform patterns of selected frequency; With macro block pattern overlapping unit, it amplifies the macro block pattern of each block of pixels having been set waveform patterns by described waveform patterns setup unit with the embedment strength value, with tile-like with this macro block pattern overlapping on the image of being imported, make it fill up the image that this is imported.

Macro block pattern among the present invention is equivalent to embed electronic watermark information, and can be used for the geometric transformation correction when detecting, and the signal that therefore can independently add the geometric transformation correction usefulness different with electronic watermark information carries out geometric transformation and proofreaies and correct.

Described waveform patterns setup unit can quantize the item value of described embedding sequence, selects described at least one frequency according to the value of this quantification, also can select described at least one frequency according to the item value symbol of described embedding sequence.

And, described waveform patterns setup unit can change phase place on one side, made setting in described macro block pattern between a plurality of block of pixels of waveform patterns of same frequency, the phase place of waveform patterns is alignd on the macro block pattern, on one side each block of pixels is set waveform patterns.

And, in described macro block pattern overlapping unit during to the doubling of the image macro block pattern imported, can be at the block of pixels in the macro block pattern, Yi Bian change the phase place of waveform patterns, make the phase place of waveform patterns of same frequency on the image behind the macro block pattern overlapping, align, Yi Bian carry out overlapping.

And, described problem can solve by digital watermark detection, this digital watermark detection from overlapping by detecting electronic watermark information the image of arranging the macro block pattern that the block of pixels that is made of a plurality of waveform patterns forms, it is characterized in that, have: linear transformation distortion correcting unit, it utilizes input picture is carried out discrete Fourier transform and at least two peak information in the spectral power matrix that obtains, and predetermined reference position information, carries out the linear transformation distortion of input picture to proofread and correct; The macro block generation unit that adds, it is proofreaied and correct the linear transformation distortion and finishes image begins to be divided into described macro block pattern magnitude from position arbitrarily macro block, will cut apart the macro block addition of gained and generates the macro block that adds; Each frequency filtering image sets generation unit of finishing dealing with, it implements process of convolution with corresponding each convolution operator of predetermined multiple frequency to the macro block that adds, and obtains the Filtering Processing corresponding with each frequency and finishes image sets; The searching position setup unit, it sets a coordinate in the predetermined hunting zone; Piece cuts position response computing unit, it carries out following processing to all piece positions, with resulting maximum value and summation cut the output of position response as piece, this is treated to: a described coordinate is considered as piece cuts the position, for finishing each image in the image sets with the corresponding Filtering Processing of described each frequency, carry out cutting apart according to the piece of the big or small split image of described block of pixels, from with corresponding each image of each frequency obtain and the same corresponding block of pixels group in position, each block of pixels for the described block of pixels group relative with this piece position, try to achieve the absolute value sum of the pixel value of the whole pixels in the block of pixels, thereby obtain pixel absolute value and group, obtain the pixel absolute value and the group in maximum, as maximum value and; Piece cuts position detection unit, it cuts position response computing unit by described, obtain piece at each coordinate of the described hunting zone of setting by described searching position setup unit and cut the position response, detect the largest block that cuts in the response of position with piece and cut the corresponding coordinate of position response, cut the position as piece corresponding to each coordinate; The detected value matrix constitutes the unit, it carries out following processing to each block of pixels, thereby try to achieve the detected value matrix, this is treated to: for finishing each image in the image sets with the corresponding Filtering Processing of described each frequency, utilized and cut the detected described piece that cuts the position of position detection unit by described and cut apart, according to each image corresponding with each frequency, for with the same corresponding block of pixels group in position in each block of pixels, try to achieve the absolute value sum of the pixel value of all pixels in the block of pixels, at least utilize and obtain maximum value and the corresponding frequency of block of pixels, try to achieve the plain value of the detected value entry of a matrix corresponding with the piece position of this block of pixels; With the watermark information despread unit, it tries to achieve the detected object sequence from the detected value matrix, tries to achieve the detected electrons watermark information by the detected object sequence is carried out despreading.

According to the present invention, the macro block pattern that embeds as electronic watermark information can be used for the geometric transformation correction.

Described watermark information despread unit can have with lower unit: rotary unit, it generates plain four matrixes arranging rotation 0 degree, 90 degree, 180 degree, 270 degree and obtain of described detected value entry of a matrix, and the symbols of all elements value of each matrix that counter-rotating turn 90 degrees by revolving, 270 degree obtain; Converter unit, it is for four matrixes that obtain by described rotary unit, respectively at each element position, element is carried out matrix after cyclic shift obtains cyclic shift, make an element position of matrix become the precalculated position of matrix, finish the detected value matrix group thereby try to achieve conversion; And output unit, it tries to achieve the detected object sequence set from each matrix the detected value matrix group is finished in conversion, in the information sets that obtains in that each the detected object sequence in the detected object sequence set is carried out despreading, export the highest information of reliability as the detected electrons watermark information.

And, described a plurality of frequency be with+number and-number two kinds of corresponding frequencies, described detected value matrix constitute the unit will from the absolute value of+number corresponding pairing block of pixels of frequency and deduct with the absolute value of-number corresponding pairing block of pixels of frequency and and the value that obtains as the detected value entry of a matrix element value corresponding with the piece position of described block of pixels.And, described detected value matrix constitute the unit to the absolute value of the block of pixels corresponding with two kinds of frequencies and the absolute value of difference given with obtain bigger absolute value and the pairing symbol of the corresponding frequency of block of pixels after value as the detected value entry of a matrix plain value corresponding with the piece position of described block of pixels.

According to the present invention, can not use frame line etc. and give the correction of the affine transformation (linear transformation+parallel moving) of image, thereby carry out the detection of eletric watermark.And, need not resemble the signal of the independent additional geometric transformation correction usefulness different the prior art with electronic watermark information, and because the pattern of expression electronic watermark information double as geometric transformation correction simultaneously signal, the problem that inter-signal interference therefore can not occur has and improves the effect that detects performance.

Description of drawings

Fig. 1 is the structure chart of the electronic watermark embedded device 10 in the embodiments of the invention.

Fig. 2 is the figure of the handling process of expression electronic watermark embedded device 10.

Fig. 3 is the structure chart of macro block pattern generating unit 11.

Fig. 4 is the figure of the handling process of expression macro block pattern generating unit 11.

Fig. 5 is the structure chart of watermark information extension 16.

Fig. 6 is the figure of the handling process of expression watermark information extension 16.

Fig. 7 is the structure chart of the frequency selection portion 17 in first example.

Fig. 8 is the figure of the handling process of the frequency selection portion 17 in expression first example.

Fig. 9 is the figure of the content of expression symbol respective frequencies database 27.

Figure 10 is the structure chart of the frequency selection portion 17 in the 3rd example.

Figure 11 is the figure of the handling process of the frequency selection portion 17 in expression the 3rd example.

Figure 12 is the figure that explanation embeds the processing of sequence item quantization unit 29.

Figure 13 is the figure of the content of expression quantized value respective frequencies database 31.

Figure 14 is the figure of the processing of explanation macro block pattern configuration part 19.

Figure 15 is the figure that is used to illustrate the processing of macro block pattern generating unit 11.

Figure 16 is the structure chart of macro block pattern tiling portion 13.

Figure 17 is the figure of the handling process of expression macro block pattern tiling portion 13.

Figure 18 is the figure that is used to illustrate the processing of macro block pattern tiling portion 13.

Figure 19 is the figure that is used to illustrate the processing of overlapping 34 of tiling.

Figure 20 is the structure chart of the digital watermark detection 40 in the embodiments of the invention.

Figure 21 is the figure of the handling process of expression digital watermark detection 40.

Figure 22 is the structure chart of linear transformation distortion correction unit 41.

Figure 23 is the figure of the handling process of expression linear transformation distortion correction unit 41.

Figure 24 is the figure that is used to illustrate the handling principle of linear transformation distortion correction unit 41.

Figure 25 is the figure that is used to illustrate the handling principle of linear transformation distortion correction unit 41.

Figure 26 is the figure that is used to illustrate the processing of discrete Fourier transform portion 46.

Figure 27 is the structure chart of peak search section 47.

Figure 28 is the figure of flow process of the processing of expression peak search section 47.

Figure 29 is the figure that is used to illustrate the processing of peak search section 47.

Figure 30 is the figure that is used to illustrate the processing of peak search section 47.

Figure 31 is the figure that is used to illustrate the processing of peak search section 47.

Figure 32 is the structure chart of linear transformation matrix supposition portion 48.

Figure 33 is the figure of the handling process of expression linear transformation matrix supposition portion 48.

Figure 34 is the figure that is used to illustrate the processing of peak information conversion portion 55.

Figure 35 is the figure that is used to illustrate the principle of the last transformation matrix calculating part 56 of spectrum space.

Figure 36 is used for the figure of key diagram as the processing of transformation component 49.

Figure 37 is used for the figure of key diagram as the processing of transformation component 49.

Figure 38 adds the structure chart of macro block generating unit 42.

Figure 39 is the add figure of handling process of macro block generating unit 42 of expression.

Figure 40 is used to illustrate the figure of processing of macro block generating unit 42 of adding.

Figure 41 is used to illustrate the figure of processing of macro block generating unit 42 of adding.

Figure 42 is used to illustrate the figure of macro block of adding.

Figure 43 is the structure chart of detected value matrix generating unit 43.

Figure 44 is the figure of the handling process of expression detected value matrix generating unit 43.

Figure 45 is the figure that the corresponding Filtering Processing of each frequency of expression is finished the processing summary of image sets generating unit 62.

Figure 46 is the structure chart that piece cuts position detection part 63.

Figure 47 is the figure that the expression piece cuts the handling process of position detection part 63.

Figure 48 is the figure that is used to illustrate searching position information.

Figure 49 is the structure chart that piece cuts position response calculating part 68.

Figure 50 is the figure that the expression piece cuts the handling process of position response calculating part 68.

Figure 51 is the figure that is used for the processing of illustrated block cutting part 72.

Figure 52 is the figure that is used for the processing of illustrated block cutting part 72.

Figure 53 is the figure that is used for pixels illustrated absolute value and calculating part 73.

Figure 54 is used for the figure that illustrated block cuts the effect of position response calculating part 68.

Figure 55 is the structure chart of detected value matrix formation portion 64.

Figure 56 is the process chart of expression detected value matrix formation portion 64.

Figure 57 is the figure that is used to illustrate the processing of detected value matrix formation portion 64.

Figure 58 is the figure with explanation detected value matrix.

Figure 59 is the structure chart of watermark information despreading portion 44.

Figure 60 is the figure of the handling process of expression watermark information despreading portion 44.

Figure 61 is the structure chart of detected value matrixing portion 82.

Figure 62 is the figure of the handling process of expression detected value matrixing portion 82.

Figure 63 is used to illustrate that 90 spend the figure of the processing of rotating parts 89.

Figure 64 is used to illustrate that 90 spend the figure of the processing of rotating parts 89.

Figure 65 is the figure that is used to illustrate the processing of macro block starting position transformation component 90.

Figure 66 is the figure that is used to illustrate the processing of of one-dimensional handling part 83.

Figure 67 is the figure that is used to illustrate the desired value that has or not eletric watermark.

Symbol description

10: electronic watermark embedded device

11: macro block pattern generating unit

12: the image input part

13: macro block pattern tiling portion

14: the image efferent

16: the watermark information extension

17: the frequency selection portion

18: the waveform patterns generating unit

19: macro block pattern configuration part

21: the error correction/detection encoding section

22: Pseudo-random number generator

23: embed the sequence generating unit

25: embed sequence item symbol obtaining section

26: selection portion

27: symbol respective frequencies database

29: embed sequence item quantization unit

30: selection portion

31: quantize the respective frequencies database

33: the enlarging section

34: the overlapping portion of tiling

40: digital watermark detection

41: linear transformation distortion correction unit

42: the macro block generating unit adds

43: detected value matrix generating unit

44: watermark information despreading portion

46: discrete Fourier transform portion

47: the peak search section

48: linear transformation matrix supposition portion

49: image transform portion

51: peak-peak location finding portion

52: the second peak search section

53: the peak information output part

55: peak information conversion portion

56: the last transformation matrix calculating part of spectrum space

57: image rectification linear transformation matrix computations portion

59: macroblock partition portion

60: the macro block generating unit

62: the corresponding Filtering Processing of each frequency is finished the image sets generating unit

63: piece cuts position detection part

64: detected value matrix formation portion

65: symbol respective frequencies database

67: the searching position configuration part

68: piece cuts position response calculating part

69: piece cuts position response buffer memory

70: piece cuts determining positions portion

72: the piece cutting part

73: image absolute value and calculating part

74: maximum value and judging part

75: the maximum value and the portion that adds

77: the piece cutting part

78: pixel absolute value and calculating part

79: ceiling capacity frequency judging part

80: detected value matrix element value configuration part

82: detected value matrixing portion

83: the of one-dimensional handling part

84: the pseudo random number generating unit

85: despreading portion

86: decoder object Information Selection portion

87: the error correction/detection lsb decoder

89:90 degree rotating part

90: macro block starting position transformation component

Embodiment

Below, with reference to accompanying drawing embodiments of the invention are described.The structure of the electronic watermark embedded device in the embodiments of the invention as shown in Figure 1, the flow process of processing is as shown in Figure 2.

As shown in Figure 1, the electronic watermark embedded device in the embodiment of the invention 10 has: macro block pattern generating unit 11, image input part 12, macro block pattern tiling portion 13, image efferent 14.With reference to Fig. 2 the handling process of this structure is described.

Electronic watermark embedded device 10 input original images, embedment strength value (are used to specify the eletric watermark decorative pattern are overlapped onto parameter on the image with what kind of intensity.Its value is big more, and the image quality after the embedding is got over deterioration, but the detection of eletric watermark is stable more), as being embedded into the watermark information (step 1) of the information in the original image.

Macro block pattern generating unit 11 is passed through the expansion watermark information and is obtained to embed sequence, and by the corresponding frequency modes such as value of arrangement with the item that embeds sequence, generates the macro block pattern (step 2) as picture pattern.

And macro block pattern tiling portion 13 is tile-like for original image obtained in image input part 12 with the macro block pattern overlapping, obtains the view data (step 3) that tiling is finished.And, in image efferent 14, the view data that tiling is finished is exported (step 4) as the eletric watermark embedded images.

Below the processing of each one is described in detail.

The structure of the described macro block pattern generating unit 11 of the electronic watermark embedded device 10 in the embodiment of the invention as shown in Figure 3, handling process is as shown in Figure 4.As shown in Figure 3, macro block pattern generating unit 11 has: watermark information extension 16, frequency selection portion 17, waveform patterns generating unit 18, macro block pattern configuration part 19.Describe with reference to the flow chart of Fig. 4 handling process macro block pattern generating unit 11.

Macro block pattern generating unit 11 input watermark informations (step 21), at first, watermark information extension 16 is expanded watermark informations and is obtained to embed sequence { e _i(step 22).

Watermark information extension 16 is described in detail.

Fig. 5 represents the structure of watermark information extension 16, and Fig. 6 represents its handling process.Watermark information extension 16 has: error correction/detection encoding section 21, Pseudo-random number generator 22, and embed sequence generating unit 23.

As shown in Figure 6, at first, watermark information is transfused to (step 31), in error correction/detection encoding section 21 watermark information is carried out error correction/detection coding (step 32).As the error correction/detection Methods for Coding, can use arbitrary method such as Bose-Chaudhuri-Hocquenghem Code, Reed Solomon coding.Then, embed sequence generating unit 23 and utilize the pseudo-random number sequence that is generated by Pseudo-random number generator 22, expanding to sequence the watermark information behind the error correction/detection coding long is m * n (size that depends on macro block.As described below, be n * n in the present embodiment, but be that example describes with generally speaking m * n at this) embedding sequence (step 33), and export this embedding sequence (step 34).In addition, the initial value that generates of the pseudo random number in the Pseudo-random number generator 22 also can be as the key parameter of watermark and from the parameter of outside input.So, can obtain to realize that with different key parameters independently eletric watermark embeds, perhaps be difficult to infer effects such as embedding algorithm.

The example ((A)～(C)) of the concrete extended method of the watermark information in the embedding sequence generating unit 23 is as follows:

(A) watermark information (S bit) behind the error correction/detection coding is made as { w _i(i=0～S-1), pseudo-random number sequence is made as { r _iWhen (i=0～mn-1 (sum of block of pixels)), embed sequence { e _i(i=0～mn-1) try to achieve by following formula:

e _i＝r _i×w _i％s(i＝0～mn-1)

Wherein, w _i=1 or-1 (i bit value of the watermark information behind the error correction/detection coding be 1 o'clock be 1,0 o'clock be-1), r _iBe r _i=1 or-1 random number series.And the remainder when x%y represents that x is removed by y.

(B) watermark information (S bit) behind the error correction/detection coding is made as { w _i(i=0～S-1), generate two kinds of pseudo-random number sequences and be made as { r1 respectively _i(i=0～mn-1), { r2 _i(during i=0～mn-1), at first utilize { r1 _iGenerate the array be used to replace.For example:

Expression (numerical expression 1)

$\left(\begin{array}{cccc}0& 1& 2& 3...\mathrm{mn}-1\\ P_0& P_1& P_2& P_{3...}{P}_{\mathrm{mn}-1}\end{array}\right)$

Array { p _i(i=0～mn-1), decide in the following way.Wherein, r1 _iBe assumed to the value that obtains at random more than or equal to 0～mn-1.

For(i＝0；i＜mn-1；i++){

P _i＝i；

}

For(i＝0；i＜mn-1；i++){

r＝r1 _i％mn；

a＝p _i；

p _i＝p _r；

p _r＝a；

}

Then, try to achieve embedding sequence { e by the following method _i(i=0～mn-1).

e _i＝r2 _pi×w _pi％s(i＝0～mn-1)

Wherein, w _i=1 or-1 (i bit value of the watermark information behind the error correction/detection coding be 1 o'clock be 1,0 o'clock then be-1), r2 _iBe r2 _i=1 or-1 random number series.And the remainder when x%y represents that x is removed by y.As mentioned above, { r1 _iBe to be used for generating { p _iPseudo-random number sequence, { r2 _iBe to be used to carry out { w _iThe pseudo-random number sequence of expansion.

(C) watermark information (S bit) behind the error correction/detection coding is made as { w _i(i=0～S-1), pseudo-random number sequence is made as { r _i(during i=0～mn-1), embed sequence { e _i(i=0～mn-1) try to achieve by following formula:

e _i＝∑ _k＝0～s-1(r _(i+k)％mn×w _k)(i＝0～mn-1)

Wherein, w _i=1 or-1 (i bit value of the watermark information behind the error correction/detection coding be 1 o'clock be 1,0 o'clock be-1), r _iBe r _i=1 or-1 random number series.And the remainder when x%y represents that x is removed by y.

And, in described example, the binaryzation random number sequence is used for expanding, but also can utilizes other random number sequences, for example use based on the random sequence of Gaussian Profile and expand.

More than watermark information extension 16 is illustrated, but as mentioned above, in the present embodiment, because the macro block pattern is made of n * n piece, it is n that the sequence that therefore embeds sequence is grown ²

In the flow process of Fig. 4, then,, will specify the counter i of the item No. that embeds sequence to be set at 0 (step 23) for each piece is set the macro block pattern, frequency selection portion 17 is selected and the i item e that embeds sequence _iCorresponding frequency (step 24) such as value.Then, waveform patterns generating unit 18 generates the waveform patterns (step 25) corresponding with selected frequency.Waveform patterns is the picture element matrix of following block size.

Below the processing of frequency selection portion 17 and waveform patterns generating unit 18 is described in more detail.And the example of three kinds of processing methods described.

(1) at first, frequency selection portion 17 in first example and waveform patterns generating unit 18 are elaborated.

The structure of frequency selection portion 17 as shown in Figure 7, handling process is as shown in Figure 8.As shown in Figure 7, frequency selection portion 17 has: embed sequence item symbol obtaining section 25, selection portion 26, symbol respective frequencies database 27.

As shown in Figure 8,17 inputs of frequency selection portion embed sequence item value (step 41), embed the symbol (step 42) that sequence item symbol obtaining section 25 obtains to embed sequence item value.As symbol, desirable (+,-, 0) three values.Then, selection portion 26 is retrieved from symbol respective frequencies database 27 and is embedded the corresponding frequency of sequence item value symbol, and as the frequency output of selecting (step 43).The structure of symbol respective frequencies database 27 as shown in Figure 9.In the symbol respective frequencies database 27, at each output that embeds sequence item symbol obtaining section 25, record, the corresponding frequency of storage.For a symbol, corresponding frequency can be not only one, also can be as the symbol "-" of Fig. 9, and corresponding a plurality of group of frequencies.

And, in whole specifications of the application and claim scope, the implication of " frequency " speech not only comprises single frequency except the situation that special differentiation is used, also comprise as described symbol "-", with the corresponding a plurality of frequencies of a symbol (perhaps 0 or quantized value).

And, the frequency here is the two-dimentional frequency of the two-dimentional waveform patterns of expression, and be that expression is the relative frequency of benchmark with the block size. for example in order to make block size for X * Y pixel, the frequency of the waveform of expression two-dimensional wavelength on the x direction, with block size for 2X * 2Y pixel, the frequency of the waveform of expression two-dimensional wavelength is identical value on the x direction, carry out normalization and the value that obtains with pixel size. in first example, waveform patterns generating unit 18 generates and the corresponding waveform patterns of selecting of frequency. and the waveform patterns that waveform patterns generating unit 25 is generated is the two-dimensional pattern onesize with piece. and, when the frequency of selecting has when a plurality of, with each waveform patterns addition, and the energy that amplitude normalization is made resulting pattern when a frequency is arranged identical pattern as waveform patterns.

(2) in second example, waveform patterns generating unit 18 is amplified the absolute value of the amplitude utilization embedding sequence item value of waveform patterns, and is different with first example in this.

(3) then frequency selection portion 17 in the 3rd example and waveform patterns generating unit 18 are elaborated.

The structure of frequency selection portion 17 as shown in figure 10, handling process is as shown in figure 11.As shown in figure 10, frequency selection portion 17 has: embed sequence item quantization unit 29, selection portion 30, quantize respective frequencies database 31.

As shown in figure 11,17 inputs of frequency selection portion embed sequence item value (step 51), embed sequence item quantization unit 29 and will embed the quantification of sequence item value, obtain to quantize to embed sequence item value (step 52).And selection portion 30 is retrieved from quantized value respective frequencies database 31 and is quantized to embed the corresponding frequency of sequence item value, obtains selected frequency (step 53), and with its output (step 54).

Quantization method as embedding sequence item quantization unit 29 comprises method as shown in figure 12.The quantification of (a)～(c) shown in Figure 12 is undertaken by computing as follows respectively:

(a) binaryzation

X when embedding sequence item value, carries out binaryzation as follows

x＜0→-1

x＞＝0→1

(b) three values

X when embedding sequence item value, carries out three values as follows

x＜0→-1

x＝0→0

x＞0→1

(c) n value

X when embedding sequence item value, carries out n value as follows

(numerical expression 2)

x＜-nq/2→-n/2

(

Expression is no more than the integer of the maximum of x)

x＞(n/2-1)q→n/2-1

And, can not resembling yet and utilize (c) constant amplitude to quantize, constant amplitude quantizes and use not.

In described step 53, the structure of the quantized value respective frequencies database 31 of 30 references of selection portion as shown in figure 13.And the example when described example is described three values.As shown in figure 13, in the quantized value respective frequencies database 31, at each output that embeds sequence quantization unit 29, record, the corresponding frequency of storage.At this moment, the frequency corresponding with quantized value can not be one, but the same during with the quantized value 0 of Figure 13, also can corresponding a plurality of group of frequencies.

About waveform patterns generating unit 18, the same with first example or second example.That is, directly use the amplitude of selected frequency mode, perhaps utilize the absolute value that embeds sequence item value to amplify.

More than the example of frequency selection portion 17, waveform patterns generating unit 18 is described, but in the following description, in the present embodiment scope, being exemplified as benchmark with first describes, and, selected frequency is two kinds (f0, f1). and, will be wherein between two kinds of waveform patterns pattern behind a kind of half-twist become another kind.

In the flow process of Fig. 4, then, macro block pattern configuration part 19 buffer memorys at preprepared macro block pattern, the block of pixels of piece position that will be corresponding with the item No. i of described embedding sequence is replaced (step 26) with described waveform patterns.The macro block pattern is that the pixel of N * N pixel is arranged as shown in figure 14, is made of n piece in length and breadth.In macro block pattern configuration part 19, for example horizontal piece position m0, vertical piece position n0 are carried out i=m0+n0 * n etc., make the item No. i that embeds sequence and piece position (m0, n0) corresponding one by one, with the block of pixels of the piece position of correspondence with corresponding waveform patterns displacement.

Make and specify the counter i of the item No. that embeds sequence to increase progressively (step 27), by all being carried out repeatedly processing (step 28) from described frequency selection portion 17 to macro block pattern configuration part 19, the macro block pattern generates the pattern that has disposed the piece with different frequency pattern as shown in figure 15, and the macro block pattern (step 29) of all pieces has been set in output at last.

Macro block pattern tiling portion 13 in the described macro block pattern generating unit 11 is described in detail.The structure of the macro block pattern tiling portion 13 in the present embodiment as shown in figure 16, handling process is as shown in figure 17.

As shown in figure 16, macro block pattern tiling portion 13 has enlarging section 33 and tiles overlapping 34.With reference to Figure 17 the handling process under this structure is carried out following explanation.

The macro block pattern that macro block pattern tiling portion 13 input is generated by waveform patterns generating unit 11, and embedment strength value, original image (step 61).Enlarging section 33 utilizes the embedment strength value to amplify the amplitude (step 62) of macro block pattern, and the macro block pattern of output amplification.The macro block pattern that is exaggerated is input to the overlapping portion of tiling.

Then, overlapping 34 macro block pattern with amplification of tiling overlaps onto (step 63) on the original image with being tile-like as shown in Figure 18.At this moment, the starting position of tiling is not limited to the upper left side of original image, can begin tiling from position arbitrarily.And, when the macro block pattern in the tilings such as end of original digital image data overflows, cut away the part of overflowing, make original image all by the macro block pattern fills.And then with the macro block pattern overlapping to original pattern the time, the piece waveform patterns that constitutes macro block can be as shown in figure 19, changes the phase place of waveform patterns at each piece in the tiling process, so that have the phase alignment of waveform patterns on original image of same frequency.The phase alignment of the waveform patterns that makes same frequency here is meant, as shown in figure 19, with the waveform patterns of each piece position of the onesize waveform patterns of original image on the correspondence position on the original image, become the waveform patterns of the piece in the tiling process.So, in the linear transformation distortion treatment for correcting when following detect, can further strengthen being used to inferring the peak value of the power spectrum of transformation parameter, help to improve anti-sex change.In addition, when block size was the integral multiple of wavelength of waveform patterns, even need not change the phase place of the waveform patterns in the piece especially, the phase place of the waveform patterns on the original image after the tiling was also alignd, and therefore need not to carry out described processing.

Tiled the original image of macro block pattern from macro block pattern tiling portion 13 output, finished image (step 64) as tiling.And the image efferent 14 of electronic watermark embedded device 10 is finished image with this tiling and is embedded as eletric watermark and finish image output.

Then, to describing from finishing the digital watermark detection 40 that detects eletric watermark the image by the eletric watermark embedding of as above action generation.

The structure of the digital watermark detection 40 in the present embodiment as shown in figure 20, handling process is as shown in figure 21.

As shown in figure 20, the digital watermark detection in the present embodiment 40 has: linear transformation distortion correction unit 41, the macro block generating unit 42 that adds, detected value matrix generating unit 43, watermark information despreading portion 44.Describe for the handling process in the digital watermark detection 40 with this structure with reference to Figure 21.

Digital watermark detection 40 input detected object images (step 71).And, at first infer the parameter of the linear transformation that the detected object image is applied by linear transformation correction unit 41, proofread and correct by inverse transformation, proofread and correct and finish detected object view data (step 72) thereby obtain linear transformation.

It is macroblock size that the macro block generating unit 42 that then adds is finished the detected object image segmentation with the linear transformation correction, and the regional addition that will cut apart obtains to add macro block (step 73).Then the piece that detects in the macro block that adds of detected value matrix generating unit 43 cuts position (step 74), generates and cuts the corresponding detected value matrix (step 75) in position with piece.Then watermark information despreading portion 44 utilizes the detected value matrix to carry out the despreading processing while detecting each 90 degree rotation/macro block beginning positions, obtains to detect watermark information and output (step 76).

And in described processing, also can behind input detected object image, enlarge/dwindle conversion, make it become the image size of predetermined pixel count the detected object image, generation is with the size normalization detected object image of image size normalization, as the input of later processing.And, detect anti-sex change in order to improve, also can linear transformation be out of shape correction unit 41 or the macro block generating unit 42 that adds before pre-treatment filtering portion is set, detected object image or linear transformation distortion is proofreaied and correct finishes the detected object image and carry out the pre-treatment Filtering Processing respectively.And, also can finish detected object image enforcement pre-treatment Filtering Processing to each detected object image and linear transformation distortion correction.

The pre-treatment Filtering Processing is described.Employed filter preferably uses the band pass filter (Band-pass filter) that the frequency of the employed frequency mode of embedding is passed through in the pre-treatment Filtering Processing.Frequency band limits by band pass filter is handled, and low-frequency component reduces.Because the frequency medium and low frequency composition of original image composition is more, therefore can alleviate of the influence of original image composition to frequency mode.And,, also can use the differential circuit that the waveform of image signal data is carried out differential as pre-treatment Filtering Processing portion.On the basis of Filtering Processing, carry out slicing and handle, can further improve the detection performance.Slicing is handled and is meant, the signal section that for example surpasses predetermined higher limit and lower limit in the picture signal waveform that Filtering Processing is finished is carried out shear treatment.Consequently, the more weak signal of amplitude strengthens relatively.Therefore, the signal level that Filtering Processing can be finished is suppressed at set point with in the interior scope, thereby can improve the detection sensitivity of frequency mode.

Then each one to described digital watermark detection 40 is described in detail.

The structure of linear transformation in present embodiment distortion correction unit 41 as shown in figure 22, handling process is as shown in figure 23.As shown in figure 22, linear transformation distortion correction unit 41 has: discrete Fourier transform portion 46, peak search section 47, linear transformation matrix supposition portion 48, image transform portion 49.With reference to Figure 23 the handling process of linear transformation distortion correction unit 41 is described.

Linear transformation distortion correction unit 41 input detected object images (step 81) at first, carry out the two-dimensional discrete Fourier Tranform by 46 pairs of detected object images of discrete Fourier transform portion, obtain spectral power matrix (step 82).Then get two element positions of big peak value in the peak search section 47 search spectral power matrixs, obtain these two element positions as peak information (step 83).

Then linear transformation matrix supposition portion 48 is according to peak information and predetermined reference position, linear transformation matrix (step 84) when calculating the reference position and being transformed to peak, so with the linear transformation matrixing on the spectrum space be the inverse of a matrix matrix of the linear transformation matrix gained on the pixel space as image rectification with linear transformation matrix output (step 85).At last, image transform portion 49 is by implementing to the detected object image with the linear transformation of image rectification with the image of linear transformation matrix notation, the linear transformation that correction is implemented the detected object image obtains linear transformation distortion correction and finishes detected object view data and output (step 86).

Below the principle of treatment for correcting is carried out in the linear transformation distortion in 46 pairs of detected object images of explanation linear transformation distortion correction unit.

At first, when the image that obtains when the electronic watermark embedded device 10 that will utilize present embodiment is directly imported as the detected object image, as shown in figure 24, to the detected object image carry out discrete Fourier transform and the power spectrum that obtains with overlap piece on the corresponding frequency location of two kinds of frequency modes on have bigger peak value. this be because: in the such regional area of block unit, the frequency that expression overlaps the frequency mode on this piece has peak value, but two kinds of frequency modes are with almost equal area (by extension process) filling detected object image, and the locality that the frequency location that general image carries out the resulting power spectrum of discrete Fourier transform is not subject to frequency mode.

Then as shown in figure 25, consider the detected object image is implemented the situation of linear transformation.As the fact of generally knowing be in this case, in the linear transformation that the detected object image is implemented,, be reflected as the rotation of the equal angular on the spectrum space for the rotation composition, for the ratiometric conversion composition, be reflected as the contrary ratiometric conversion composition on the spectrum space.For example when having carried out on image space rotation θ angle and carry out the such linear transformation of s ratiometric conversion doubly, on spectrum space, rotate the θ angle accordingly, carry out 1/s ratiometric conversion such linear transformation doubly.And power spectrum has the feature that the parallel amount of movement with respect to image remains unchanged.Thus, when as Figure 25, the detected object image having been carried out linear transformation, the spectral power matrix of Figure 24 is implemented the linear transformation corresponding with this linear transformation, so also move because of conversion with two kinds of frequency mode corresponding peak value position.

Therefore, detect two peaks in the spectral power matrix that obtains by the detected object image of as Figure 25, having implemented linear transformation, as shown in figure 24, if will under the no deformation state and the peak on the spectral power matrix that obtains as the reference position, calculate the reference position and be transformed to two detected linear transformations that peak is such, then the linear transformation on the Dui Ying pixel space also can be determined, thereby can carry out the supposition of linear transformation parameter and the correction of distortion.At this moment, because the configuration of the power spectrum of image is these two reasons of method that point symmetry is arranged and do not differentiated two peak points with respect to initial point (flip-flop position), therefore in each 90 degree rotations, leave uncertainty.Promptly, the linear transformation distortion correction of carrying out the linear transformation distortion to proofread and correct and obtaining is finished the detected object image and is compared with the state that image is finished in original eletric watermark embedding, become any one states of rotation 0 degree, 90 degree, 180 degree, 270 degree, but this be after processing in solve by each 90 degree rotation correspondences.And, also have the uncertainty of the mirror transformation of image, but it solves by carrying out dual following evolution matrix supposition portion processing afterwards.

Then the processing to each one in the linear transformation distortion correction unit 41 of present embodiment is elaborated.

The processing of the discrete Fourier transform portion 46 in the present embodiment as shown in figure 26.As input, output is carried out the spectral power matrix that discrete Fourier transform obtains to the detected object image with the detected object image in discrete Fourier transform portion 46.As the size of discrete Fourier transform, can be size of detected object image itself, use FFT (fast discrete Fourier Tranform) etc. in order to realize high speed processing, therefore can big or smallly in length and breadth be set at 2 power.When the varying in size of the size of detected object image and discrete Fourier transform, as shown in figure 26, become the discrete Fourier transform object pixels from the picture centre taking-up, cut the part of the size of overflowing discrete Fourier transform, when not reaching the necessary size of discrete Fourier transform, fill with pixel value 0, perhaps fill etc. with the average pixel value of the subregion of the detected object image in the size of discrete Fourier transform.For further raising speed, also the detected object image can be reduced into predetermined multiplying power, for example be reduced in length and breadth and carry out discrete Fourier transform after 1/2 size.Like this in the spectral power matrix that obtains, owing to described reason contains the bigger peak value of 4 points (is 2 points independently if consider symmetry).

The structure of the peak search section 47 in the present embodiment as shown in figure 27, its handling process is as shown in figure 28.As shown in figure 27, the peak search section 47 in the present embodiment has: peak-peak location finding portion 51, the second peak search section 52, peak information output part 53.Followingly handling process is described with reference to Figure 28.

Peak search section 47 input power spectrum matrixs (step 91), at first, peak-peak location finding portion 51 as shown in figure 29, in the predetermined frequency band in the spectral power matrix that obtains by discrete Fourier transform portion 46, going out to obtain the frequency location (step 92) of peak-peak by searching and detecting. definite method of the frequency band here is to select to make it not contain low-frequency band. the reason of avoiding low frequency is, the concentration of energy of original image self is on low frequency, might detect the cause of peak-peak position mistakenly. and, because the symmetry of spectral power matrix is as long as search the first half frequency band just has been enough to.

In addition, when search peak-peak position, can will obtain peaked frequency location in the predetermined frequency band simply as detected peak-peak position, for example as shown in figure 30, have pulse feature in order the to find more powerful frequency location of acquisition of (and peak value), the frequency location that also can be maximum with the response of the such convolution operator of Laplce's filtering is as the peak-peak position.And then, when utilizing convolution operator, only compare, in order to detect peak value, preferred use sparse convolution operator as shown in figure 30 etc. with certain ramp rate expansion by some pulse that constitutes (for example Laplce's filtering of 3 * 3) with detection.This be because, appear at peak value in the spectral power matrix because certain nonlinear deformation of detected object image etc., the possibility of certain expansion is arranged.Peak-peak location finding portion 51 detects the peak-peak position as described above, and it is exported as the peak-peak positional information.

Then, as shown in figure 31, the second peak search section 52 is in the second peak region of search center, as the near zone of predetermined size revolve the point that turn 90 degrees with the position on the power spectrum shown in the peak-peak positional information, search obtains the frequency location of peak-peak, and detected frequency location is exported (step 93) as the second peak information.Here why the second peak region of search is set in the peak-peak position revolve the point that turn 90 degrees near, be to be that a kind of pattern that revolves after turning 90 degrees is become alternative two kinds of patterns because be used for the waveform patterns of the frequency that eletric watermark embeds.

And the peak-peak position is revolved why remarkablely turn 90 degrees o'clock as second peak, but the reason of search near zone is: the linear transformation that the detected object image is applied comprises the conversion composition that aspect ratio changes and shears, and will consider that therefore peak is from revolving the peak-peak position situation of the some dislocation that turn 90 degrees.The size of the second peak region of search pre-determines according to the degree of pairing aspect ratio change, shear transformation in the linear transformation distortion correction unit 41.And when search second peak, portion 51 is the same with the peak-peak location finding, can will obtain peaked frequency location in the second peak region of search as detected second peak, also the response of convolution operator that can Figure 30 is such becomes maximum frequency location as second peak.

Obtain peak-peak positional information, the second peak information by above action, but also can use additive method as follows: in predetermined frequency band shown in Figure 29, with the position that obtains peak-peak as the peak-peak position, with obtain second largest peak value o'clock as second peak value.

At last, peak information output part 53 with peak-peak positional information and the second peak information in groups, as peak information output (step 94).

Then the linear transformation matrix supposition portion 48 in the linear transformation distortion correction unit 41 is elaborated.The structure of the linear transformation matrix supposition portion 48 in the present embodiment shown in figure 32, the flow process of processing is as shown in figure 33.Shown in figure 32, linear transformation matrix supposition portion 48 has: peak information conversion portion 55, the last transformation matrix calculating part 56 of spectrum space, image rectification linear transformation matrix computations portion 57.With reference to Figure 33 the handling process in the linear transformation matrix supposition portion 48 is described.

The peak information (step 101) that 48 inputs of linear transformation matrix supposition portion obtain by peak search section 47, at first, conversion peak information in peak information conversion portion 55, and peak information (step 102) is finished in the acquisition conversion.And reference position information that last transformation matrix calculating part 56 utilizations of spectrum space are preserved in advance and conversion are finished the peak information calculations and are gone out the last transformation matrix (step 103) of spectrum space.At last, image rectification is transformed to the last transformation matrix of pixel space with linear transformation matrix computations portion 57 with the last transformation matrix of spectrum space, and the result is exported (step 104) as image rectification with the linear transformation matrix.Processing to each one is carried out following detailed description.

Figure 34 represents the contents processing of peak information conversion portion 55. peak information conversion portion 55 input peak information, for the peak-peak positional information in the peak information and the second peak information, judge whether the position on the power spectrum shown in each positional information is arranged in the first quartile or second quadrant, when not being positioned at the first quartile and second quadrant, the information of shifting one's position, make point-symmetric position become new position. second peak is not in the first quartile or second quadrant in Figure 34, so the point-symmetric point that becomes second peak in the peak information is as the second new peak.

Then, for the peak-peak position that all is present in the first quartile or second quadrant and second peak, coordinate figure on the x direction is less, promptly on spectral power matrix, be positioned at the left side one name changing be peak A, another changed into peak B, and the peak information B that will represent the peak information A of peak A and expression peak B in groups, exports as conversion peak information.In addition, by the generation conversion peak information that contrary processing obtains to the adquisitiones of peak A and peak B, and, can realize the mirror transformation correspondence of image with whole dual the carrying out of processing afterwards.

Then the processing of the last transformation matrix calculating part 56 of spectrum space is described with reference to Figure 35.The peak that state after the last transformation matrix calculating part 56 of spectrum space embeds eletric watermark obtains down, promptly with embed corresponding two frequency locations of employed two kinds of frequency modes and preserve in advance as reference position information.In the information of reference position, with x direction coordinate figure less, promptly on spectral power matrix, be positioned at the left side one be set at reference position a, another is set at reference position b.The position transformation matrix is that reference position a is moved to peak A, reference position b is moved to the linear transformation of peak B as shown in figure 35 on this moment spectrum space, calculates by following formula:

In the flip-flop with spectral power matrix is in the two-dimensional coordinate system of initial point,

Reference position a:(a _x, a _y)

Reference position b:(b _x, b _y)

Peak A:(A _x, A _y)

Peak B:(B _x, B _y)

If the last transformation matrix of spectrum space is

(numerical expression 3) $X=\left(\begin{array}{cc}p& q\\ r& s\end{array}\right)$

Then

(numerical expression 4) $\left(\begin{array}{cc}{A}_x& B_x\\ A_y& B_y\end{array}\right)=X\left(\begin{array}{cc}{a}_x& b_x\\ a_y& b_y\end{array}\right)$

Solve this formula,

(numerical expression 5) $X=\left(\begin{array}{cc}p& q\\ r& s\end{array}\right)=1/(a_x{b}_y-b_x{a}_y)\left(\begin{array}{cc}{A}_x{b}_y-B_x{a}_y& -A_x{b}_x+B_x{a}_x\\ A_y{b}_y-B_y{a}_y& -\mathrm{Ayby}+B_y{a}_x\end{array}\right)$

Then, the contents processing of image rectification with linear transformation matrix computations portion 57 described.The image rectification linear transformation matrix computations portion last transformation matrix in 57 input spectrum spaces.As shown in figure 25, the relation of linear transformation on the pixel space and the linear transformation on the spectrum space is:

* ratiometric conversion composition inverse transformation

* it is identical to be rotated into branch

According to this character, obtain the linear transformation matrix on the pixel space corresponding with the last transformation matrix of spectrum space.

If the linear transformation matrix on the spectrum space is

(numerical expression 6) $X=\left(\begin{array}{cc}p& q\\ r& s\end{array}\right)$

Then the linear transformation matrix on the Dui Ying pixel space is

(numerical expression 7) $Y=1/(\mathrm{ps}-\mathrm{qr})\left(\begin{array}{cc}s& -r\\ -q& p\end{array}\right)$

Because the image rectification that final hope is tried to achieve is the inverse matrix of the last transformation matrix of described pixel space with the linear transformation matrix, therefore becomes

(numerical expression 8) $Y^{-1}=\left(\begin{array}{cc}p& q\\ r& s\end{array}\right)$

The transposed matrix of the last transformation matrix of spectrum space is got final product with the output of linear transformation matrix as image rectification.

Then, the contents processing with reference to the image transform portion 49 in the distortion of the linear transformation in the digital watermark detection 40 in Figure 36, the 37 pairs of present embodiments correction unit 41 describes.

Image transform portion 49 imports the detected object images, reaches the image rectification linear transformation matrix that generates by linear transformation matrix supposition portion 48, will finish the detected object image as linear transformation distortion correction to the image that the linear transformation of detected object image obtains and export.

In the time of can obtaining outer regional of detected object image by linear transformation, use pixel value 0 to fill this moment as shown in figure 36, perhaps with the processing of the average pixel value filling of detected object image.And, as shown in figure 37, when linear transformation distortion is proofreaied and correct the image size of finishing the detected object image when having pre-determined, cut away the detected object part of overflowing from fixed size, during regional outside obtaining the detected object image, fill with pixel value 0, perhaps fill with the average pixel value of the subregion of detected object image.

Then, the macro block generating unit 42 that adds in the digital watermark detection 40 is described.The structure of the macro block generating unit 42 that adds in the present embodiment as shown in figure 38, handling process is as shown in figure 39.

As shown in figure 38, the macro block generating unit 42 that adds has macroblock partition portion 59 and macro block generating unit 60.With reference to Figure 39 the handling process in the macro block generating unit 42 that adds is described.

The macro block generating unit that adds 42 inputs are proofreaied and correct from the linear transformation distortion of linear transformation distortion correction unit 41 outputs and are finished detected object image (step 111), at first as shown in figure 40, proofread and correct the optional position of finishing the detected object image from the linear transformation distortion, be divided into the macro block of predetermined size (being identical value when embedding), obtain macro block group (step 112).This moment is in the image end etc., when the pixel region that never satisfies macro block obtains macro block, fills with pixel value 0, perhaps fills with the average pixel value of the subregion of detected object image.

Then, macro block adds portion 60 as shown in figure 41, with all the macro block additions in the macro block group, generates the macro block that adds, and with its output (step 113).

The linear transformation distortion of proofreading and correct the linear transformation distortion and obtaining proofread and correct finish the detected object image and embed after eletric watermark finish image and compare, has identical scaling, identical direction (is supposed each 90 degree anglecs of rotation, the mirror transformation uncertainty is identical), can think to follow parallel mobile state. so, with be tiled in embedding after eletric watermark embed the identical macro block of the macro block pattern magnitude finish on the image and cut apart each macro block in the macro block group that obtains, become the pixel region of macro block pattern respectively to be comprised under the parallel mobile status of same amount. therefore in the macro block that adds as the macro block summation, as shown in figure 42, the macro block pattern of dislocation is concentrated in the mode of the amount that increased the number of times correspondence that adds. because the macro block pattern when embedding is included in the macro block that adds in the mode of cyclic shift, therefore in the macro block that adds, carry out parallel mobile correspondence, the beginning of the macro block pattern when find embedding is during the position, can detect the watermark information of embedding. carry out processing as described below based on described viewpoint. in addition, also can omit and add macro block generating unit 42. promptly, also can be as Figure 42 with a piece with the identical size of macro block that cuts as input content with reprocessing. but its prerequisite is to use the macro block generating unit 42. that adds in following explanation

Then the detected value matrix generating unit 43 in the eletric watermark checking device 40 of present embodiment is described.The structure of detected value matrix generating unit 43 as shown in figure 43, handling process is as shown in figure 44.As shown in figure 43, detected value matrix generating unit 43 has: the corresponding Filtering Processing of each frequency is finished image sets generating unit 62, piece cuts position detection part 63, detected value matrix formation portion 64, symbol respective frequencies database 65.Followingly the handling process of detected value matrix generating unit 43 is described with reference to Figure 44.

The macro block (step 121) that adds that 43 inputs of detected value matrix generating unit are generated by the macro block generating unit 42 that adds, at first, finish in the image sets generating unit 62 in the corresponding Filtering Processing of each frequency, utilize and be registered in symbol respective frequencies database 65 (perhaps quantized value respective frequencies database.The identical database of employed database when using with embedding.At this is the database that contains two kinds of frequencies.) in the convolution operator of frequency correspondence, the macro block that adds is carried out convolution algorithm, the Filtering Processing of generated frequency kind number (2) is finished image, and it is finished image sets output (step 122) as the Filtering Processing corresponding with each frequency.

And piece cuts position detection part 63 and carries out the detection that piece cuts the position, detects piece and cuts position (step 123).At last, detected value matrix formation portion 64 constitutes with detected piece and cuts corresponding detected value matrix in position and output (step 124).

Below further each one is described.

Figure 45 represents that the corresponding Filtering Processing of each frequency finishes the processing summary of image sets generating unit 62.In example shown in Figure 45, these two kinds of predetermined frequencies and f0, f1 are corresponding, generate two Filtering Processing and finish image.Utilize the convolution operator corresponding, carry out convolution algorithm when each pixel of the macro block that adds is scanned successively, finish image thereby generate two Filtering Processing corresponding with two frequencies with each frequency.At this moment, as shown in figure 45, the Filtering Processing of the frequency mode alignment when pairing frequency of convolution operator and embedding is finished in the image-region, the value of big absolute value appears having, the value of less absolute value appears having when not lining up (in Figure 45, bright expression :+the value of big absolute value, dark expression :-the value of big absolute value).This is because convolution algorithm can be regarded as and carries out the part correlation computing, when being identical patterns each other, gets the correlation with big absolute value, gets the correlation with less absolute value when different pattern.These two Filtering Processing are finished image finish image sets output as the Filtering Processing corresponding with each frequency.In addition, when carrying out convolution algorithm, in the time of can't the reference pixels value in the macro block end that adds etc., with reference to the limit of the so opposite side that turns back to image of cyclic shift and the pixel value that obtains.

Then, the structure that piece cuts position detection part 63 as shown in figure 46, handling process is as shown in figure 47.As shown in figure 46, piece cuts position detection part 63 and has: searching position configuration part 67, piece cut that position response calculating part 68, piece cut position response buffer memory 69, piece cuts determining positions portion 70.With reference to Figure 47 the handling process that piece cuts position detection part 63 is described.

Piece cuts the position detection part 63 inputs Filtering Processing corresponding with each frequency and finishes image sets (step 131).At first, the searching position information (o in the predetermined hunting zone of searching position configuration part 67 generations _x, o _y) (step 132). then, piece cuts the piece that position response calculating part 68 tries to achieve when the coordinate shown in the searching position information is assumed to piece and cuts the position and cuts position response (step 133). and the piece of trying to achieve cuts position response and corresponding searching position information in groups, being stored in piece cuts in the response buffer memory of position (step 134). will handle carrying out (step 135) on all searching positions that contained in the predetermined hunting zone. and last, cut in the determining positions portion 70 at piece, will have the searching position information that largest block cuts the position response and cut positional information output (step 136) as detected.

Figure 48 represents the example of searching position information.The hunting zone is that (each image of finishing image sets with the corresponding Filtering Processing of each frequency also is identical size for upper left end points with the macro block that adds, therefore also be their upper left end points) be the near zone with a certain size at center, the coordinate (o that this near zone is contained _x, o _y) select successively as searching position information.For example, the hunting zone is the block size rectangular area at center for the left upper end with the macro block that adds.Even when the piece in the macro block that adds of input time cuts the position from the dislocation of the left upper end of the macro block that adds, have searching position under the situation that correct piece cuts the position and correct piece nearby when cutting aligned in position, as long as use piece to cut the response computational methods that position response and other searching positions are in a ratio of higher value, just can detect correct piece and cut the position.And, if use aforesaid hunting zone, then in the hunting zone, contain any correct piece certainly and cut the position, can detect correct piece effectively by detection and cut the position.Carry out following processing based on described viewpoint.

The structure that piece cuts position response calculating part 68 as shown in figure 49, handling process is as shown in figure 50.As shown in figure 49, piece cuts position response calculating part 68 and has: piece cutting part 72, pixel absolute value and calculating part 73, maximum value and judging part 74, maximum value and the portion 75 that adds.With reference to Figure 50 its handling process is described.

Piece cuts the position response calculating part 68 inputs Filtering Processing corresponding with each frequency and finishes image sets and searching position information (step 141).And value initialization maximum is absolute and the portion 73 that adds is 0 (step 142).

Then, each Filtering Processing that piece cutting part 72 is finished in the image sets for the Filtering Processing corresponding with each frequency is finished image, piece when carrying out that the coordinate shown in the searching position information cut the position as piece is cut apart, and obtains successively with each Filtering Processing that is positioned at the same block position and finishes the corresponding block of pixels group (step 143) of image.

Then, the pixel absolute value and the portion 73 that adds are for finishing each block of pixels of the corresponding block of pixels group of image with each Filtering Processing that is arranged in same position, the absolute value of the pixel value in the calculating pixel piece and, obtain pixel absolute value and the group (step 144) corresponding with each block of pixels.Then, maximum value and judging part 74 selected to obtain peaked value in pixel absolute value corresponding with each block of pixels and the group, should value as maximum value with obtain (step 145).At last at maximum value with add in the portion 75, will be that the S of 0 value adds maximum value and (step 146) as initial value.Repeatedly piece is cut apart all obtained each Filtering Processing and finish the pairing image block group of image and carry out beginning processing till the maximum value and the portion 75 that adds from described pixel absolute value and calculating part 73, the S that will obtain as its result cuts position response output (step 147, step 148) as piece.

Then, the processing that piece is cut each one in the position response calculating part 68 is specifically described.

At first, piece cutting part 72 is described.Figure 51 is the figure that is used for the processing of illustrated block cutting part 72.

The piece cutting part 72 inputs Filtering Processing corresponding with each frequency finished image sets and searching position information, finish image for each Filtering Processing that the Filtering Processing corresponding with each frequency finished in the image sets, piece when carrying out that the coordinate shown in the searching position information cut the position as piece is cut apart, and exports successively with each Filtering Processing that is positioned at same position and finishes the corresponding block of pixels group of image.The sizableness of block of pixels is in each Filtering Processing being finished the size of image (all identical sizes) when cutting apart with predetermined number.Carry out piece when cutting apart by the coordinate shown in the searching position information, under the situation of the pixel outside the reference pixels size, shown in Figure 52, with reference to the limit of an opposite side that turns back to the macro block that adds and the pixel value that obtains, by so-called cyclic shift acquisition pixel value.

Figure 53 represents the figure of the processing of pixels illustrated absolute value and calculating part 73. pixel absolute value and calculating part 73 inputs are finished the corresponding block of pixels group of image with each Filtering Processing, calculate the summation of the absolute value of its pixel value that comprises for each block of pixels, the result is exported as pixel absolute value corresponding with each block of pixels and group. the energy of corresponding frequency when this processing is equivalent to calculate with each block of pixels generation. in addition, symbol or quantized value have a plurality of frequencies at once when embedding, absolute value that will be corresponding and the summation of (energy) with each frequency as and the corresponding absolute value of these a plurality of frequencies and. in addition, " absolute value and " speech in this specification and claims scope, except the situation that special differentiation is used, it is all as the implication of described " a plurality of absolute values that addition is corresponding with a plurality of frequencies with ".

For being that correct piece describes by the reason that piece cuts the bigger value of position response calculating part 68 outputs when cutting the position when the coordinate shown in the searching position information.When the piece shown in the last figure that carries out Figure 54 cuts, searching position information is when correct piece cuts location dislocation, the difference of the piece dividing method when each block of pixels of cutting apart gained embeds owing to eletric watermark misplaces, therefore in block of pixels, just contain the pixel region outside the frequency mode overlapping when embedding, the energy of overlapping frequency is as shown below, compare when cutting, become less value with correct piece.That is, the coordinate shown in the searching position information is a correct piece when cutting the position, and piece cuts the position response and gets maximum.

Then, the detected value matrix formation portion 64 in the detected value matrix generating unit 43 in the digital watermark detection 40 is described.Figure 55 represents the structure of detected value matrix formation portion 64, and Figure 56 represents handling process.Shown in Figure 55, detected value matrix formation portion 64 has: piece cutting part 77, pixel absolute value and calculating part 78, ceiling capacity frequency judging part 79, detected value matrix element value configuration part 80.With reference to Figure 56 the handling process of detected value matrix formation portion 64 is described.

The detected value matrix formation portion 64 input Filtering Processing corresponding with each frequency finish image sets, reach piece cuts positional information (step 151).Piece cutting part 77 cuts the coordinate shown in the positional information with piece and cuts the position as piece, the Filtering Processing corresponding with each frequency finished each Filtering Processing in the image sets finish image and carry out piece and cut apart, generate successively with each Filtering Processing that is positioned at same position and finish the corresponding block of pixels group of image (step 152).

Then, pixel absolute value and calculating part 78 inputs are finished the corresponding block of pixels group of image with each Filtering Processing, calculate the summation of the absolute value of the pixel value that wherein contains for each block of pixels, the result is exported (step 153) as pixel absolute value corresponding with each block of pixels and group.The piece cutting part 77 in the detected value matrix formation portion 64 and the processing of pixel absolute value and calculating part 78 and piece cutting part 72 that piece cuts position response calculating part 68 and pixel absolute value and calculating part 73 are identical.Then, ceiling capacity frequency judging part 79 judges that the block of pixels when obtaining maximum in pixel absolute value corresponding with each block of pixels and the group is from finishing the piece that obtains the image with the corresponding Filtering Processing of which frequency, the frequency of correspondence being exported (step 154) as the frequency with ceiling capacity.

At last, detected value matrix element value configuration part 80 is according to decision of symbol respective frequencies database and the corresponding symbol of frequency with ceiling capacity, try to achieve two (because frequency have two kinds) pixel absolute values and the absolute value of difference, will to two pixel absolute values and the value of the additional described symbol of absolute value of difference as detected value, be set at the plain value (step 155) of the detected value entry of a matrix suitable with the piece position.Here, also can with from+corresponding pairing pixel the absolute value of frequency and deduct with the pixel of-corresponding frequency definitely and value as element value.And, also can be according to symbol respective frequencies database decision and the corresponding symbol of frequency with ceiling capacity, will this symbol append to described maximum pixel absolute value and value as element value.And, also can will append to value on the symbol to 1 as element value.And then, when use utilizes the method that described quantized value respective frequencies database embeds, also can determine and the corresponding quantized value of frequency with ceiling capacity, with it as element value.

Repeatedly all piece positions are carried out beginning processing (step 156) till the detected value matrix element value configuration part 80 from described pixel absolute value and calculating part 78, with its result as detected value matrix output (step 157).

With reference to Figure 57 the handling process of detected value matrix formation portion 64 is carried out more specific description. the Filtering Processing of will the Filtering Processing corresponding with each frequency finishing in the image sets is finished image, cut the position as piece and be divided into each piece to cut the corresponding coordinate of positional information with piece, generate and finish the corresponding block of pixels group of image with each Filtering Processing, at each piece position, try to achieve absolute value and (a0 among Figure 57 and the a1) of the pixel value of each block of pixels in the block of pixels group, judgement obtain the maximum pixel absolute value and block of pixels be to finish the piece that obtains the image from the Filtering Processing corresponding (to be a0＞a1) among Figure 57 with which frequency, to finish the corresponding frequency of image with Filtering Processing obtains as the ceiling capacity frequency. and according to symbol respective frequencies database decision and ceiling capacity frequency corresponding symbol (owing to be a0＞a1 among Figure 56, therefore be "+"), try to achieve two pixel absolute values and difference absolute value (among Figure 57 be | a0-a1|), the value (being "+| a0-a1| " among Figure 57) of having added described symbol as detected value, is set at the element value corresponding with the piece position of detected value matrix.

With reference to Figure 58 the detected value matrix that obtains by described action is described.The item value symbol of the embedding sequence that each piece in the macro block pattern that the left figure of Figure 58 represents to generate when embedding is corresponding, right figure are represented the symbol of each element of the detected value matrix that obtains by described action.In detected value matrix generating unit 43, detect piece and cut the position, cutting the position from piece begins to cut piece and generates the detected value matrix, in the relation of the macro block pattern when therefore adding macro block and embedding, parallel amount of movement in each piece keeps synchronously, but the parallel of the piece of arranging moved, and promptly parallel the moving of the detected value entry of a matrix element of Pai Lieing also has uncertainty.The beginning position of the arrangement of each piece in the macro block pattern that generates when the circular mark among the figure is represented to embed, but the arrangement of element is parallel mobile in the detected value matrix, therefore certain element position in the detected value matrix is corresponding with original beginning position, and the arrangement that detected value matrix integral body becomes element value is the state that benchmark carries out cyclic shift with the beginning position.And, meanwhile also have the probabilistic each 90 degree rotation compositions of conduct that linear transformation distortion correction unit 41 is left over.

If can corresponding described uncertain element, then can finally correctly detect the watermark information of embedding.Constitute the watermark information despreading portion 44 of present embodiment based on this viewpoint.Below the watermark information despreading portion 44 of present embodiment is described.

The structure of the watermark information despreading portion 44 in the present embodiment is shown in Figure 59, and handling process is shown in Figure 60.Shown in Figure 59, the watermark information despreading portion 44 in the present embodiment has: detected value matrixing portion 82, of one-dimensional handling part 83, pseudo random number generating unit 84, despreading portion 85, decoder object Information Selection portion 86, error correction/detection lsb decoder 87.Followingly the handling process of watermark information despreading portion 44 is described with reference to Figure 60.

The detected value matrix (step 161) that 44 inputs of watermark information despreading portion are generated by detected value matrix generating unit 43, at first, detected value matrixing portion 82 generates the parameter of distributing each 90 degree rotation sign-inverted/macro block starting positions and change detection value matrix and detected value matrix group (step 162) is finished in the conversion that obtains.Then 83 pairs of conversion of of one-dimensional handling part are finished the detected value matrix that each conversion in the detected value matrix group finishes and are carried out of one-dimensional, obtain detected object sequence set (step 163).

Then in despreading portion 85, pseudo-random number sequence and detected object sequence set that input is generated by Pseudo-random number generator 84, utilize pseudo-random number sequence that each the detected object sequence in the detected object sequence set is carried out despreading and handle, obtain decoder object information and eletric watermark and have or not desired value group (step 164).

Then decoder object Information Selection portion 86 is when all eletric watermarks have or not desired value to be lower than predetermined threshold value, can't detect the output of eletric watermark as watermark information despreading portion, and end process (step 168), under reverse situation, select to have the decoder object information (step 167) that the maximum electron watermark has or not index.

The decoder object information of last 87 pairs of selections of error correction/detection lsb decoder is carried out the error correction/detection decoding, obtains to detect watermark information (step 169).At this moment, during error in detecting decoder object information, if recoverable then proofread and correct and as detecting watermark information output, when detecting uncorrectable error, output can't the detected electrons watermark.

Below each one of watermark information despreading portion 44 is described in detail.

The structure of the detected value matrixing portion 82 in the present embodiment is shown in Figure 61, and handling process is shown in Figure 62.Shown in Figure 61, detected value matrixing portion 82 has 90 degree rotating part 89 and macro block starting position transformation components 90.

Shown in Figure 62, detected value matrixing portion 82 input detected value matrixes (step 171), at first in 90 degree rotating parts 89, shown in Figure 63, acquisition is with arrangement rotation 0 degree of detected value entry of a matrix element, 90 degree, 180 degree, four matrixes (step 172) of 270 degree. then in four matrixes, implement and 90 degree rotations, (step 173 is handled in the counter-rotating of the symbol of all elements value of the corresponding matrix of 270 degree rotations, 174). this is because shown in Figure 64, when revolving, image turn 90 degrees or 270 when spending, waveform patterns during embedding reverses, so the symbol of the detected value that obtains from image also can reverse. by described processing, resulting four matrixes are finished the detected value matrix group as rotation obtain (step 175).

Then in macro block starting position transformation component 90, shown in Figure 65, when with the macro block starting position during as each element position of matrix, the detected value matrix that each rotation in the detected value matrix group of finishing for rotation is finished generates the matrix of cyclic shift successively, so that this element position moves to the upper left of matrix, resulting all matrixes are finished detected value matrix group (step 176) as conversion.When the detected value matrix was n * n matrix, conversion was finished the detected value matrix group all by 4n ²The detected value matrix that individual conversion is finished constitutes.

Then, with reference to Figure 66 the processing of the of one-dimensional handling part 83 in the watermark information despreading portion 44 in the present embodiment is described.In of one-dimensional handling part 83, the detected value matrix group is finished in the input conversion, and with the identical rule in the macro block configuration part when embedding 19, the matrix element of the detected value matrix of finishing for each conversion carries out of one-dimensional and obtains the detected object sequence.After from the detected value matrix that all conversion are finished, obtaining each detected object sequence, it is summarized and exports as the detected object sequence set.

Then, below the despreading portion 85 in the present embodiment is described.Despreading portion 85 imports the detected object sequence set, reaches the pseudo-random number sequence that is generated by Pseudo-random number generator 84.Pseudo-random number sequence utilizes with electronic watermark embedded device 10 identical Pseudo-random number generator and generates.Each detected object sequence in 85 pairs of detected object sequence set of despreading portion is carried out despreading and is handled the acquisition decoded information.

As concrete despreading method, extended method that can be when embedding is corresponding, adopts method as follows:

(A) the detected object sequence is made as { d _i(i=0～mn-1), pseudo-random number sequence are made as { r _i(during i=0～mn-1), decoder object information (being made as the S bit) { c _k(k=0～S-1) try to achieve by the following method.Try to achieve

Cor _k＝(∑ _{(i＝0～mn-1)∩(i％S＝k)}(d _i×r _i))/SQRT(∑ _{(i＝0～mn-1)∩(i％S＝k)}d _i ²)(k＝0～S-1)

With the decision bit value

Cor _k＞＝0→c _k＝“1”

Cor _k＜0→c _k＝“0”

Wherein, r _iBe r _i=1 or-1 random number series.And the remainder when x%y represents that x is removed by y, SQRT () represents square root.

(B) the detected object sequence is made as { d _i(i=0～mn-1), two kinds of pseudo-random number sequences that generate are made as { r1 respectively _i(i=0～mn-1), { r2 _i(during i=0～mn-1), decoder object information (being made as the S bit) { c _k(k=0～S-1) try to achieve by the following method.At first, utilize { r1 _iSimilarly generate the array { p that is used to replace when embedding _i.

Then, try to achieve

Cor _k＝(∑ _{(i＝0～mn-1)∩(pi％S＝k)}(d _pi×r2 _pi))/SQRT(∑ _{(i＝0～mn-1)∩ (i％S＝k)}d _pi ²)(k＝0～S-1)

With the decision bit value

Cor _k＞＝0→c _k＝“1”

Cor _k＜0→c _k＝“0”

Wherein, r2 _iBe r2 _i=1 or-1 random number series. and the remainder of x%y when representing that x is removed by y, SQRT () represents square root.

(C) the detected object sequence is made as { d _i(i=0～mn-1), pseudo-random number sequence are made as { r _i(during i=0～mn-1), decoder object information (being made as the S bit) { c _k(k=0～S-1) try to achieve by the following method.

Try to achieve

Cor _k=(∑ _(i=0～mn-1)(d _i* r _{(i+k) %mn}))/SQRT (∑ _(i=0～mn-1)d _i ²) (k=0～S-1) with the decision bit value

Cor _k＞＝0→c _k＝“1”

Cor _k＜0→c _k＝“0”

Wherein, r2i is the random number series of r2i=1 or-1.And the remainder when x%y represents that x is removed by y, SQRT () represents square root.

And, in despreading portion 85, generate eletric watermark accordingly with decoder object information and have or not desired value.This is the desired value of the detection signal strength of the decoder object information that obtains by despreading of expression, in the example of described (A), (B), (C), for example calculates by the following method:

When if eletric watermark has or not desired value to be E,

E＝∑ _{(k＝0～S-1)}|Cor _k|

Having or not desired value to become effective reason with reference to Figure 67 to aforesaid eletric watermark describes.Cor when the detection of (1) expression eletric watermark of Figure 67 is correctly carried out _kThe value of being got.And, (2) expression do not embed eletric watermark or because when distortion occurring and causing detecting the detected object image of electronic watermark signal greatly, the Cor in the time of can't correctly carrying out the eletric watermark detection _kThe value of being got.Shown in Figure 67, in the time can correctly detecting, owing to the correlation between the detected object sequence that embeds employed random number series and obtain from the detected object image is bigger, so each Cor _kAbsolute value become bigger value, opposite in the time can't correctly detecting, each Cor _kAbsolute value become less value.And, among the example of despreading in the present embodiment (A), (B), (C), always after norm normalization, calculate its relevant with random number series with the detected object sequence, even therefore embed different detected object images, different watermark informations, also can assess with identical yardstick.Based on above reason, shown in the right side of Figure 67, as Cor _kAbsolute value and eletric watermark have or not desired value E under the situation of (1), to become bigger value, under the situation of (2), become less value.Therefore as described below, utilize predetermined threshold alpha, can judge whether correct the carrying out of detection of eletric watermark.And, when utilizing the fractional error of error correcting code correct detection watermark information, generally when producing the error that surpasses calibration capability, error correcting code can cause producing the danger of error correction (proofread and correct and be wrong symbol), but, can avoid the problem of error correction by as following, having or not the threshold decision of desired value to combine with eletric watermark.

Despreading portion 85 gathers all decoder object information of obtaining as mentioned above and as the decoder object information sets, and gathers all eletric watermarks and have or not desired value to have or not the output of desired value group as eletric watermark.

Decoder object Information Selection portion 86 in the present embodiment is carried out following explanation.Decoder object Information Selection portion 86 input decoder object information sets and eletric watermark have or not the desired value group, have or not each eletric watermark in the desired value group to have or not desired value for eletric watermark, utilized the threshold decision of predetermined threshold value, when all eletric watermarks have or not desired value not reach threshold value, can't detect the output content of eletric watermark as watermark information despreading portion, and end process.And, even when having a electronic watermark information more than or equal to threshold value, will determine that also the maximum electron watermark has or not desired value, from the decoder object information sets, select to have or not the corresponding decoder object information of desired value with the maximum electron watermark, and with its output.

Error correction/detection lsb decoder in the present embodiment 87 is carried out following explanation. error correction/detection lsb decoder 87 input decoder object information, decoder object information is carried out the error correction/detection decoding processing, obtain to detect watermark information. at this moment, during error in detecting decoder object information, if can proofread and correct then proofread and correct and as detecting watermark information output, when detecting uncorrectable error, output can't detect eletric watermark.

And, also can change the order of described decoder object Information Selection portion 86 and error correction/detection lsb decoder 87.That is, at first each the decoder object information to the decoder object information sets is carried out the error correction/detection decoding processing in error correction/detection lsb decoder 87, selects to comprise the decoder object information of correctable error.When detecting when all being uncorrectable error, output can't detect eletric watermark.Then in decoder object Information Selection portion 86, have or not desired value to carry out threshold decision to the eletric watermark corresponding with the decoder object information that contains correctable error, output can't detect eletric watermark when all less than threshold value, when being not this situation, will be to having or not the corresponding decoder object information of desired value to carry out the information of error correction/detection decoding acquisition as detecting watermark information output with the maximum electron watermark detection.

And, in described embodiment, put down in writing and all generated conversion in advance and finish the detected value matrix, processing afterwards also all gathers to be handled, but also can be undertaken by following manner: the generation of carrying out the detected value matrix finished from conversion successively begins the processing till the error correction/detection lsb decoder 87, when eletric watermark had or not the recoverable passed examination that threshold decision is qualified and error correction/detection is handled of desired value, Interrupt Process and output detected watermark information.So, only be the only about half of of described example to the desired value that detects the processing time till processing finishes.Breaking away from cycle criterion only has or not one in the threshold decision of desired value or the recoverable inspection that error correction/detection is handled to carry out by eletric watermark, carry out another inspection when breaking away from circulation time, passed examination is then exported the detection watermark information, detect the defective circular treatment of then proceeding, thereby the processing speed in the circulation is accelerated, and then realize high speed.

And then, also can be only to the partial sequence of detected object sequence carry out eletric watermark detect have or not desired value/despreading computing, for example carry out computing till preceding half bit of decoder object information, to and the eletric watermark that obtains has or not desired value to carry out threshold decision, when more than or equal to threshold value, the despreading of being left is handled, and detects watermark information by carrying out error correction/detection decoding output.This is to have or not desired value only with the also exportable sufficient value to judging that eletric watermark has or not of partial sequence to a certain degree because eletric watermark detects.So, be a part of sequence of detected object sequence because the eletric watermark detection has or not the treating capacity of desired value, despreading computing, therefore can realize the high speed of handling.

According to illustrated so far present embodiment, can not use frame line etc. and the correction of the affine transformation (linear transformation+parallel moving) carrying out occurring in the image, and carry out the detection of eletric watermark.And, need not resemble the signal of geometric transformation correction usefulness additional different independently the prior art with electronic watermark information, and owing to the pattern of expression electronic watermark information uses with signal as the geometric transformation correction simultaneously, the problem that inter-signal interference therefore can not occur has and can improve the effect that detects performance.

And illustrated device can be installed in the computer with CPU, storage device etc. by the program that will carry out this apparatus function and realize in the present embodiment.This program can write down, be assigned in the recording mediums such as CD-ROM, internal memory, and can distribute by network.

And, the invention is not restricted to described execution mode, within the claim scope, can carry out various changes, application.

Claims (18)

1. electronic watermark embedded device, it comprises a plurality of macro block patterns that contain the block of pixels of a plurality of pixels and embeds electronic watermark information by overlapping on the image of being imported, and it is characterized in that having:

The electronic watermark information expanding element, it obtains the corresponding embedding sequence of block of pixels number in length and the described macro block pattern by the electronic watermark information that expansion is imported;

The waveform patterns setup unit, it is worth according to the item with the corresponding embedding sequence in position of a block of pixels in described macro block pattern, from predetermined a plurality of frequencies, select at least one frequency, will be set at pixel in the described block of pixels with the corresponding waveform patterns of selected frequency; With

Macro block pattern overlapping unit, it amplifies the macro block pattern of having set waveform patterns to each block of pixels by described waveform patterns setup unit with the embedment strength value, with tile-like with this macro block pattern overlapping on the image of being imported, make it fill up the image that this is imported.

2. electronic watermark embedded device according to claim 1, wherein, described waveform patterns setup unit quantizes the item value of described embedding sequence, selects described at least one frequency according to the value of this quantification.

3. electronic watermark embedded device according to claim 1, wherein, described waveform patterns setup unit is selected described at least one frequency according to the symbol of the item value of described embedding sequence.

4. electronic watermark embedded device according to claim 1, wherein, described waveform patterns setup unit changes phase place on one side, made setting in described macro block pattern between a plurality of block of pixels of waveform patterns of same frequency, the phase place of waveform patterns is alignd on the macro block pattern, on one side each block of pixels is set waveform patterns.

5. electronic watermark embedded device according to claim 1, wherein, described macro block pattern overlapping unit is to the doubling of the image macro block pattern imported the time, at the block of pixels in the macro block pattern, change the phase place of waveform patterns on one side, make the phase place of waveform patterns of same frequency on the image behind the macro block pattern overlapping, align, Yi Bian carry out overlapping.

6. digital watermark detection, it is arranged the block of pixels that is made of a plurality of waveform patterns and detects electronic watermark information the image of the macro block pattern that forms from overlapping, it is characterized in that having:

Linear transformation distortion correcting unit, it utilizes input picture is carried out discrete Fourier transform and at least two peak information in the spectral power matrix that obtains, and predetermined reference position information, carries out the linear transformation distortion of input picture to proofread and correct;

The macro block generation unit that adds, it is proofreaied and correct the linear transformation distortion and finishes image begins to be divided into described macro block pattern magnitude from position arbitrarily macro block, will cut apart the macro block addition of gained and generates the macro block that adds;

Each frequency filtering image sets generation unit of finishing dealing with, it implements process of convolution with corresponding each convolution operator of predetermined multiple frequency to the macro block that adds, and obtains the Filtering Processing corresponding with each frequency and finishes image sets;

The searching position setup unit, it sets a coordinate in the predetermined hunting zone;

Piece cuts position response computing unit, it carries out following processing to all piece positions, with resulting maximum value and summation cut the output of position response as piece, this is treated to: a described coordinate is considered as piece cuts the position, for finishing each image in the image sets with the corresponding Filtering Processing of described each frequency, carry out cutting apart according to the piece of the big or small split image of described block of pixels, from with corresponding each image of each frequency obtain and the same corresponding block of pixels group in position, each block of pixels for the described block of pixels group relative with this piece position, try to achieve the absolute value sum of the pixel value of the whole pixels in the block of pixels, thereby obtain pixel absolute value and group, obtain the pixel absolute value and the group in maximum, as maximum value and;

Piece cuts position detection unit, it cuts position response computing unit by described, obtain piece at each coordinate of the described hunting zone of setting by described searching position setup unit and cut the position response, detect the largest block that cuts in the response of position with piece and cut the corresponding coordinate of position response, cut the position as piece corresponding to each coordinate;

The detected value matrix constitutes the unit, it carries out following processing to each block of pixels, thereby try to achieve the detected value matrix, this is treated to: for finishing each image in the image sets with the corresponding Filtering Processing of described each frequency, utilized and cut the detected described piece that cuts the position of position detection unit by described and cut apart, according to each image corresponding with each frequency, for with the same corresponding block of pixels group in position in each block of pixels, try to achieve the absolute value sum of the pixel value of all pixels in the block of pixels, at least utilize and obtain maximum value and the corresponding frequency of block of pixels, try to achieve the plain value of the detected value entry of a matrix corresponding with the piece position of this block of pixels; With

The watermark information despread unit, it tries to achieve the detected object sequence from the detected value matrix, try to achieve the detected electrons watermark information by the detected object sequence is carried out despreading.

7. digital watermark detection according to claim 6, wherein, described watermark information despread unit has:

Rotary unit, it generates plain four matrixes arranging rotation 0 degree, 90 degree, 180 degree, 270 degree and obtain of described detected value entry of a matrix, and the symbols of all elements value of each matrix that counter-rotating turn 90 degrees by revolving, 270 degree are obtained;

Converter unit, it is for four matrixes that obtain by described rotary unit, respectively at each element position, element is carried out matrix after cyclic shift obtains cyclic shift, make an element position of matrix become the precalculated position of matrix, finish the detected value matrix group thereby try to achieve conversion; With

Output unit, it tries to achieve the detected object sequence set from each matrix the detected value matrix group is finished in conversion, in the information sets that obtains in that each the detected object sequence in the detected object sequence set is carried out despreading, export the highest information of reliability as the detected electrons watermark information.

8. digital watermark detection according to claim 6, wherein,

Described a plurality of frequency be with+number and-number two kinds of corresponding frequencies,

Described detected value matrix constitute the unit will from the absolute value of+number corresponding pairing block of pixels of frequency and deduct with the absolute value of-number corresponding pairing block of pixels of frequency and and the value that obtains as the detected value entry of a matrix element value corresponding with the piece position of described block of pixels.

9. digital watermark detection according to claim 6, wherein,

Described a plurality of frequency be with+number and-number two kinds of corresponding frequencies,

Described detected value matrix constitute the unit to the absolute value of the block of pixels corresponding with two kinds of frequencies and the absolute value of difference given with obtain bigger absolute value and the pairing symbol of the corresponding frequency of block of pixels after value as the detected value entry of a matrix plain value corresponding with the piece position of described block of pixels.

10. one kind by the overlapping electronic watermark embedding method that comprises in the electronic watermark embedded device that a plurality of macro block patterns that contain the block of pixels of a plurality of pixels embed electronic watermark information on the image of being imported, and it is characterized in that having:

The electronic watermark information spread step by the electronic watermark information that expansion is imported, obtains the corresponding embedding sequence of block of pixels number in length and the described macro block pattern;

Waveform patterns is set step, be worth according to item with the corresponding embedding sequence in position of a block of pixels in described macro block pattern, from predetermined a plurality of frequencies, select at least one frequency, will be set at pixel in the described block of pixels with the corresponding waveform patterns of selected frequency; With

Macro block pattern overlapping step, with the embedment strength value macro block pattern of having set waveform patterns to each block of pixels by described waveform patterns setting step is amplified, with tile-like with this macro block pattern overlapping on the image of being imported, make it fill up the image that this is imported.

11. electronic watermark embedding method according to claim 10, wherein, the item value that described waveform patterns setting step has described embedding sequence quantizes, and selects the step of described at least one frequency according to the value of this quantification.

12. electronic watermark embedding method according to claim 10, wherein, described waveform patterns setting step has the step of selecting described at least one frequency according to the symbol of the item value of described embedding sequence.

13. electronic watermark embedding method according to claim 10, wherein, described waveform patterns is set step and is changed phase place on one side, made setting in described macro block pattern between a plurality of block of pixels of waveform patterns of same frequency, the phase place of waveform patterns is alignd on the macro block pattern, on one side each block of pixels is set waveform patterns.

14. electronic watermark embedding method according to claim 10, wherein, in described macro block pattern overlapping step, to the doubling of the image macro block pattern imported the time, at the block of pixels in the macro block pattern, change on one side the phase place of waveform patterns, make the phase place of waveform patterns of same frequency on the image behind the macro block pattern overlapping, align, Yi Bian carry out overlapping.

15. one kind from overlapping by the electronic watermark detecting method in the digital watermark detection that detects electronic watermark information the image of arranging the macro block pattern that the block of pixels that is made of a plurality of waveform patterns forms, it is characterized in that having:

Linear transformation distortion aligning step utilizes input picture is carried out discrete Fourier transform and at least two peak information in the spectral power matrix that obtains, and predetermined reference position information, carries out the linear transformation distortion of input picture to proofread and correct;

The macro block that adds generates step, the linear transformation distortion is proofreaied and correct finish image begins to be divided into described macro block pattern magnitude from position arbitrarily macro block, will cut apart the macro block addition of gained and generates the macro block that adds;

Each frequency filtering image sets of finishing dealing with generates step, and the macro block that adds is implemented process of convolution with corresponding each convolution operator of predetermined multiple frequency, obtains the Filtering Processing corresponding with each frequency and finishes image sets;

Searching position is set step, sets a coordinate in the predetermined hunting zone;

Piece cuts position response calculation procedure, following processing is carried out in all piece positions, the maximum value that obtained of output and summation cut the position response as piece, this is treated to: a described coordinate is considered as piece cuts the position, carry out cutting apart to finishing each image in the image sets according to the piece of the big or small split image of described block of pixels with the corresponding Filtering Processing of described each frequency, from with corresponding each image of each frequency obtain and the same corresponding block of pixels group in position, each block of pixels for the described block of pixels group relative with this piece position, try to achieve the absolute value sum of the pixel value of the whole pixels in the block of pixels, thereby obtain pixel absolute value and group, obtain the pixel absolute value and the group in maximum, as maximum value and;

Piece cuts the position probing step, cut position response calculation procedure by described, try to achieve piece at each coordinate of setting the described hunting zone of step setting by described searching position and cut the position response, detect the largest block that cuts in the response of position with piece and cut the corresponding coordinate of position response, cut the position as piece corresponding to each coordinate;

The detected value matrix constitutes step, each block of pixels is carried out following processing and tried to achieve the detected value matrix, this is treated to: for finishing each image in the image sets with the corresponding Filtering Processing of described each frequency, utilized and cut described detected piece that cuts the position of position probing step by described and cut apart, according to each image corresponding with each frequency, each block of pixels for the block of pixels group corresponding with same position, try to achieve the absolute value sum of the pixel value of all pixels in the block of pixels, and utilize at least and obtain maximum value and the corresponding frequency of block of pixels, try to achieve the plain value of the detected value entry of a matrix corresponding with the piece position of this block of pixels; With

Watermark information despreading step is tried to achieve the detected object sequence from the detected value matrix, try to achieve the detected electrons watermark information by the detected object sequence is carried out despreading.

16. electronic watermark detecting method according to claim 15, wherein, described watermark information despreading step has:

The rotation step generates plain four matrixes arranging rotation 0 degree, 90 degree, 180 degree, 270 degree and obtain of described detected value entry of a matrix, and the symbols of all elements value of each matrix that counter-rotating turn 90 degrees by revolving, 270 degree obtain;

Shift step, for four matrixes that obtain by described rotation step,, element is carried out matrix after cyclic shift obtains cyclic shift at each element position, make an element position of matrix become the precalculated position of matrix, finish the detected value matrix group thereby try to achieve conversion; With

The output step, from finishing each matrix the detected value matrix group, conversion tries to achieve the detected object sequence set, by each the detected object sequence in the detected object sequence set being carried out in the information sets that despreading obtains, the highest information of output reliability is as the detected electrons watermark information.

17. electronic watermark detecting method according to claim 15, wherein,

Described a plurality of frequency be with+number and-number two kinds of corresponding frequencies,

Described detected value matrix constitutes step and has following steps: will from the absolute value of+number corresponding pairing block of pixels of frequency and deduct with the absolute value of-number corresponding pairing block of pixels of frequency and and the value that obtains as the detected value entry of a matrix element value corresponding with the piece position of described block of pixels.

18. electronic watermark detecting method according to claim 15, wherein,

Described a plurality of step be with+number and-number two kinds of corresponding frequencies,

Described detected value matrix constitutes step and has following steps: will to the absolute value of the block of pixels corresponding with two kinds of frequencies and the absolute value of difference give with obtain bigger absolute value and the corresponding pairing symbol of frequency of block of pixels after value as the detected value entry of a matrix plain value corresponding with the piece position of described block of pixels.

Images