JP2824910B2 - How to combine data into an image - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a method for synthesizing data into an image for mixing or separating other data such as a document into an image signal.

(Prior Art) In recent years, the form of use of OA (office automation) equipment has been networked with the progress of technology, and the information to be handled has been diversified to include document data as well as audio and image data.

For example, a face photograph as information on a person is handled as image data, and in addition, attribute data such as a name, age, and address are also indispensable and are generally handled as document data.

However, since the image data and the document data have different signal processing forms, they are generally transmitted and stored separately. In this case, there are drawbacks that two files are managed in spite of only one target, which makes not only complicated handling but also errors such as a mismatch between the two.

Also, in transmitting information between networked OA devices, concealment measures to prevent leakage to third parties are indispensable.

In view of such circumstances, various methods for mixing other information such as document data into an image signal have been conventionally proposed.

One such method is CSXydeas, B. Kostic, and R. Steele, et al., “Method of Combining Data into Images by Modulo Masking” (IEEE Trans. Cnmmun. COM-32 NO1, 1984),
Shimomura, Hasegawa, Nakagawa, et al., “Improvement of Data Synthesizing Method by Modulo Masking” (IEEE Trans.C)
ommun COM-36, 1988). The method using the Modulo Masking method proposed in these papers is
It utilizes the fact that there is a strong correlation between adjacent pixels in a multi-valued image, and replaces the value of the intermediate pixel with the code corresponding to the data to be mixed for three adjacent pixels.

By using such a method, both the pixel and the document can be handled collectively, which is not only extremely convenient but also makes it possible to conceal and transmit more important information as if the image were transmitted. Therefore, it can be used as a kind of encrypted communication means.

However, although the modulo masking method can completely reconstruct the original image as an output image, it necessarily holds multi-level information, making it unsuitable for data compression, and is not practical as an image processing method itself. Lack of sex.

On the other hand, the inventors of the present application reported the Journal of the Institute of Image Electronics Engineers of Japan (1988
The method proposed in “Embedding attribute information in dithered image by multi-stage division quantization” published in 1993) is slightly output image, although data is compressed from multi-valued to binary by signal processing by dithering. There is room for improvement due to deterioration of quality.

In addition, a “data synthesizing method to an image” using the density pattern method already proposed by the present applicant (Japanese Patent Application No. 62-47310)
In this example, other data is embedded in an image by controlling a density pattern with other data when converting multi-value density information of each image into binary density information. According to this method, an extremely large amount of data can be synthesized without any disturbance of the image, and this method is very effective as a method of synthesizing data into an image. However, the size of an original image depends on the size of a density pattern (n × n pixels). for representing information for one pixel, there is a disadvantage that the data amount increases ^twice n compared as systematic dither image.

(Object of the Invention) The present invention has been made in order to eliminate the above-mentioned drawbacks of the conventional method of synthesizing data into an image, and it is possible to further reduce the data amount while hardly impairing the image quality. An object of the present invention is to provide a method for synthesizing data into an image in which a lot of information can be mixed.

(Summary of the Invention) In order to achieve this object, in the present invention, each element of a dither matrix used in expressing the density of an original image in a plurality of gradations, in particular, an organized dither method in which data compression is easy. Attention is paid to a set of threshold values of which are a predetermined value and an output dither image corresponding to the set, and other data is embedded using the statistical characteristics of the dither image. That is, it is necessary to determine the arrangement of the two pixels in the dither image based on the data to be mixed, utilizing the visual integration effect that the overall density is the same even if the two pixels in the dither image are replaced. If there is, it should be replaced. This method can greatly reduce the amount of image data compared with the conventional methods of combining data into images using the modulo masking method and the density pattern method, and can be compared with the multi-stage quantization method. There is a feature that image quality is less deteriorated.

(Example) Hereinafter, the present invention will be described in detail based on the illustrated example, but before that, a systematic dither method will be briefly described to facilitate understanding of the present invention.

The systematic dither method is a method of displaying an intermediate gradation of an image, and binary display is particularly often used. More specifically, as shown in FIG. 2A, the entire screen of the original image 1 composed of a matrix of dot arrays is divided into a partial area Fn composed of n × n dots.
And a predetermined n × n dither matrix Dn corresponding to the partial area Fn is prepared as shown in FIG.

That is, the n × n dither matrix Dn is Dn = [d _pq ⁽ⁿ⁾ ] where d _pq ⁽ⁿ⁾ is an element of p rows and q columns, and a threshold in which an integer from 0 to n ₂ −1 appears only once. It shall indicate the sequence number of the value.

At this time, the luminance Iuv of the pixel on the u-th row and the v-th column on the display screen is a threshold value T = d _pq ⁽ⁿ⁾ in the dither matrix Dn defined as p = u mod n q = v mod n (D _pq ⁽ⁿ⁾ +1/2) R / n ² is output, and if Iuv ≧ T, 1 is output if Iuv <t. Here, R is the luminance range of the multi-valued image.

If each pixel is very small, the dither image 2 represented in this way is recognized as an intermediate density according to the ratio of black and white to human eyes, and the matrix element n ² +1 gradation can be expressed.

The density gradation expression by the dither method can be applied not only to a white / black image but also to a color image. For a color, the density expression may be performed for each of the three primary colors as described above.

Furthermore, in the dither method, the means for encoding an image is “0”.
It is applied not only to the binary digital signal of "1" but also to multi-level digital signals, and research is progressing to obtain an image close to the original image with as little data amount as possible.

In general, the number of pseudo gradations Leq that can be represented by an L-value dither matrix of size n × n is Leq = (L−1) n ² +1 (1), and in the case of a binarized dither with L = 2, Leq = N ² -1.

Also, the size of n is determined in consideration of the resolution and visual characteristics of a normally used display device. In this case, if the size of n is selected so that the resolution of the display device is high and the visual integration effect can be used. The gradation is determined only by the ratio of black and white in each matrix of the generated dither image, and is independent of the arrangement order. That is, there is a predetermined degree of freedom in selecting a dither image when representing the same gradation.

The present invention utilizes the degree of freedom of a dither image based on the visual integration effect as described above when creating a dither image, and replaces pixels in the image according to data to be mixed.

That is, by paying attention to a combination of two cells in the dither matrix in which the threshold value difference is a predetermined value, a combination of pixels in the dither image corresponding to the combination is set according to data to be mixed. The data is combined with the image signal.

In this embodiment, the case of the binary dither method is exemplified for simplicity of explanation.

FIG. 2 is a block diagram for explaining a technique for realizing a method for synthesizing data into an image according to the present invention shown in FIG. 1;

In FIG. 1, reference numeral 1 denotes an original image. The partial area Fn is taken out, the density of each pixel is read out by a quantization block 3, and a numerical value indicating the gradation order of each pixel is displayed in a gradation comparison block 4. And outputs it to the dithering block 5.

Reference numeral 6 denotes an encoding block, which is a dictionary 7 storing data of characters, symbols, codes, and the like which may be used in advance.
Having. When document data such as characters to be mixed is input, the encoding block 6 generates a bit sequence of "0" and "1" in accordance with a code representing data in the dictionary corresponding to the data, for example, JIS code or ASCII code. A numerical sequence Ki which is coded as shown in FIG. 2 or is further encrypted is generated and input to the dithering block 5.

In the dithering block 5, the image signal input from the gradation comparison block 4 and the encoding block 6 will be described in detail later.
It combines both of the input document data.

Next, the threshold value difference of the dither matrix and the output dither image based on the difference will be described.

FIG. 2A shows an example in which each pixel of the original image is quantized,
(B) is an example of a 4 × 4 dither matrix, which is a spiral dither matrix. As shown in FIG. (C), each cell has a code x _i (i = 0, 1,..., 15). Attached.
(D) is a dither image obtained as a result of comparison between the original image Fn and the dither matrix Dn, and each pixel has y _i (i = 0, 1,..., 15) as shown in (e). Is given.

 Now, the threshold difference k is defined as follows.

k = x _j −x _i (k = 1, 2, 3,..., 15) where x _j is the j-th threshold, x _i is the i-th threshold, and x _i
<X _j

For example the _{k = x j -x i = x} 3 -x 6 = 9-1 = 8 in spiral dither matrix shown in FIG. 2 (d) when i = 6, j = 3. That is, the threshold difference in this case is 8. Thus, a pair of x _i and x _j having a threshold difference of k is defined as (x _i ,
x _j ) _k , then (1,9) _{8 in} this case.

Similarly, for k = 8, (0,8) ₈ , (2,10) ₈ , (3,1
1) ₈ , (4,12) ₈ , (5,13) ₈ , (6,14) ₈ , (7,15) ₈ and this set is represented by S ₈ = 8 (0,8), (1, 9), ……, (7,15)｝. There is a similar set Sk for k = 1 to 15.

Further, a binary code output by a set of thresholds (x _i , x _j ) _k is _represented by (y _i , y _j ) _k . At this time, the output (y _i , y _j ) _k
Is uniquely determined by the input signal, that is, the original image 1 and the threshold difference k, and the combination is (0,0), (0,1), (1,
0) and (1,1).

For example, when k = 8, i = 6 and j = 3 as before, FIG. 2 (d)
Is (y _i , y _j ) _k = (y ₆ , y ₈ ) ₈ = (1,0), and when k = 8, i = 1, j = 12, (y _i , y _{j) k} = (y _1, a _{y 12) 8 = (0,0)} .

Next, one direction in which characters are combined with the above-described dither image and a method for restoring the dither image will be described.

First, a dither matrix Dn to be used and a threshold difference k are determined. As described above, the binary output (y _i , y _j ) _k corresponding to (x _i , x _j ) _k based on the dither matrix Dn and the threshold difference k is (0,0), (0,1) , (1,0), (1,1) 4
One of the streets.

Even if the two dots of the output (y _i , y _j ) _{k in} the dither image are exchanged, the density of one dither image is the same as a whole. At this time, although the arrangement of the dither image does not change even if two dots are exchanged for (0,0) and (1,1), (0,0)
In (1) and (1,0), the arrangement of dither images changes when two cells are exchanged. Therefore, data synthesis is performed by focusing on the two sets (0, 1) and (1, 0).

However, when determining the threshold difference k, as will be described in detail later, the number of sets of (x _i , x _j ) _k that generates more (1, 0) patterns in the dither image depends on k. Note the differences.

First, find out what (y _i , y _j ) _k is, and if (y _i , y _j ) _k
= (0,0) or (y _i , y _j ) _k = (1,1), no data is combined with this set. Otherwise, combine the data.

For example, the bit to be synthesized is 0 and (y _i , y _j )
_{If k} = (0, 1), output as it is, and (y _i , y _j ) _k = (1,
If (0), the cells are exchanged and (y _i , y _j ) _k = (0,1) is output.

On the other hand, the bit to be synthesized is 1 and (y _i , y _j )
_{If k} = (0,1), replace the dots and (y _i , y _j ) _k =
(1,0) is output, and if (y _i , y _j ) _k = (1,0), it is output as it is.

When the bit to be synthesized by the above procedure is 0, (0,1),
When 1, data is mixed as (1,0).

In order to restore the document data synthesized by this method, a process reverse to the above synthesis procedure is performed. That is, the pattern corresponding to (x _i , x _j ) _k in the partial region of the dither image is (0,
If (0), (1,1), skip, and (y _i , y _j ) _k = (0,1)
If b _r = 0, then output b _r = 0, and if (y _i , y _j ) _k = (1, 0), b _r
= 1 is output. Document data can be decoded from this bit sequence.

Next, the amount of data that can be synthesized by the above method will be examined.

For example, when one dither matrix is composed of 4 × 4 pixels, the number of sets of threshold differences existing in one dither matrix is 8 or less when 0 <k ≦ 8. Of these, (0,1) and (1,0) that can be used for data synthesis are considered to be about half of the total.

Therefore, since data of 8/2 = 4 bits can be combined in one dither matrix, the size of the entire image is 256 × 256.
If it is an image Thus, if the character is represented by 8 bits, about 2 kbytes can be combined.

On the other hand, the data synthesizing method using the density pattern method is capable of synthesizing approximately 70 kbytes of characters in an image of 256 × 256 pixels, and a larger amount of data can be mixed than in the present invention. However, in the density pattern method, the data amount required to represent one pixel is 16 times larger than that of the systematic dither method if one dither matrix is 4 × 4, and the original data amount is extremely large.

Therefore, the present invention has a small amount of document data that can be synthesized as compared with the density pattern method, but requires much less original data amount, and is extremely convenient as a data processing means.

 The following procedure can be summarized as follows.

(Synthesis procedure)

STEP 1: A partial area corresponding to the dither matrix is extracted, and a set of thresholds (x _i , x _j ) _k and a set of output values (y
_i , y _j ) Make _k .

STEP 2: If (y _i , y _j ) _k = (1,0) or (y _i , y _j ) _k = (0,1), the data is synthesized as follows. If br = 0, (y _i , y _j ) _k = (0,1) If br = 1, then (y _i , y _j ) _k = (1,0) STEP is performed for all sets in the matrix. Step 2 is executed, and STEP1 and STEP2 are repeated for the entire image.

[Decoding means]

STEP 1: A partial area corresponding to the dither matrix is extracted, and a set of thresholds (x _i , x _j ) _k and a set of output values (y
_i , y _j ) Make _k .

STEP 2: If (y _i , y _j ) _k = (1,0) or (y _i , y _j ) _k = (0,1), the data is decoded as follows. That is, if (y _i , y _j ) _k = (0,1), br = 0 is extracted as synthesized data.

If (y _i , y _j ) _k = (1,0), br = 1 is taken out as synthesized data.

STEP 2 is executed for all sets in the matrix, and STEP 1 and STEP 2 are repeated for the entire image.

As described above, according to the present invention, a large amount of information can be synthesized while compressing data. However, it is clear from the procedure that the dither image is disturbed by combining the document data with the dither image. Therefore, it is necessary to restore the image in order to obtain good image quality when outputting the image.

By the way, the output of the dither image before combining characters (y _i , y
_j ) Comparing the number of (1,0) and (0,1) in _k , (1,0) is the majority and the (0,1) pattern is extremely small. Although this ratio varies depending on the type of the dither matrix, it depends on the threshold difference k, and as k is increased, the probability of becoming (1, 0) becomes extremely high.

For example, Girl, Moon, Aeria of the standard image database SIDBA
For each pixel of l (256 × 256 pixels 8 bits), k = 1,2,
When the ratio of (0,1) to the (0,1) and (1,0) in which (y _i , y _j ) _k in 15 is _obtained , as shown in FIG. Become. It can be seen that the appearance ratio of the (0,1) pattern is extremely small even when any of the dither matrices of (a) Bayer type, (b) spiral type, and (c) halftone type is used. In particular, if the coordinate positions of i and j in an arbitrary (x _i , x _j ) _k ∈ Sk are close to each other, (y _i , y _j ) _k = (1,0) ).

For example, k = 8,12 in Bayer type and k in halftone type
= 12 has a feature that the appearance ratio of the (0,1) pattern is almost zero.

Therefore, using this feature, (y _i , y _j ) _k =
Convert all parts (0,1) to (1,0). By this processing, an image very close to the dither image before the synthesis of the character information can be restored.

Further, the present invention can be applied to encryption communication means, and FIG. 4 is a block diagram showing an embodiment in that case.

In the figure, reference numeral 8 denotes an original image signal, which is a pixel signal extracted therefrom and a signal obtained by encrypting data 9 to be mixed with a scrambler 11 using an encryption key 10 by a decryption circuit 12 by the method described above. The images are synthesized, a dither image 13 is generated, and transmitted to desired users 14 and 14. Upon receiving this, the user separates the original image signal 16 and the encrypted signal by the decryption circuit 15 based on a predetermined rule, and further deciphers the encrypted signal by the descrambler 17 using the predetermined encryption key 18. The data 19 is extracted.

Since each block in this system, for example, the scrambler 11, the descrambler 17, or the encryption method can be easily realized by using an existing technique, detailed description will be omitted.

When various experiments were performed based on the method described above, extremely good results were obtained.

For example, images of the standard image database SIDBA (256 x 25
(6 pixels x 8 bits), and when it is performed on a binary image with n = 4, as a result of converting to ASCII code and synthesizing characters of about 2 kbytes, nothing using the conventional organized dither method There was almost no difference from the reproduced image when not synthesized.

Experiments have also confirmed that image data generated by the method of the present invention can be restored to an image substantially equivalent to an image obtained by a conventional systematic dither method by changing its pixel arrangement.

Further, when data is mixed into an image according to the present invention, if a long input of bit 0 occurs as a data string,
The output (y _i , y _j ) _k at this time is (0, 1), and
The image distortion increases. At this time, an operation such as inverting and inputting one bit at a time immediately before combining the data may be added, and this has also been confirmed by experiments.

This method is not limited to the case described above, and may be widely used when a data string having a certain regularity may have an undesired effect on a reproduced image.

In this way, there are various ways to use means for confidentially transmitting or storing other data in an image. For example, a system for synthesizing personal information of a person with face photograph data of the person and managing the data collectively is constructed. If this is the case, only a person with a matching encryption key that is recognized as a face photograph and can be known to a third party can know the confidential data, and a system that ensures the confidentiality can be obtained.

Also, irrespective of the presence or absence of the encryption means, if the present invention is applied to facsimile or other image signal transmission that is used everyday, data related to the image can be transmitted together with the image, so that the transmission processing can be unified. I can do it. Further, it is obvious that the data to be mixed is not limited to simple characters, but may be any signal such as an audio signal or an image transmission signal.

Furthermore, although it has been described that (0,1) is used when combining bit 0 and (1,0) is used when combining bit 1, the procedure is the same even if this is reversed.

In the embodiment described above, for simplicity of explanation, 2
Although the description has been given of the case of a value image, the present invention is not limited to this, and can be applied to a multi-value image or a color image, and the configuration of an apparatus and a system at the time of implementation may be variously modified. Good.

(Effects of the Invention) As described above, the present invention mixes desired data in a dither image processing step as a gray scale display means of an image, so that the presence of the mixed data appears on an image signal and a reproduced image. Difficult and without loss of image quality,
Further, it has a remarkable effect in providing a means for synthesizing an extremely large amount of information in an image while compressing the data amount.

Also, according to the present invention, the image signal and other data can be handled by the same means, so that the information medium can be greatly simplified. Further, the presence or absence of mixed data is difficult to appear on the screen. It is also effective in confidentially storing and transmitting data.

[Brief description of the drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a diagram for explaining a general dither method, and FIG. 3 is (y _i , y _j ) _k = (0,1) in a dither image. FIG. 4 is a block diagram showing an embodiment in which the present invention is applied to cryptographic communication means. 1 ... Original image, 2 ... Dither image, 3 ... Quantization block, 4 ... Tone comparison block, 5 ... Dithering block, 6 ... Coding block, 7 ... Dictionary.

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-63-214067 (JP, A) IEICE D-IIII Vol. J72- D- ▲ II ▼ No. 6 pp. 880-886 June 1989 (58) Field surveyed (Int. Cl. ⁶ , DB name) H04N 1/38-1/393 G09C 5/00

Claims (5)

(57) [Claims] When quantizing density information of an original image using a dither method to represent a pseudo gradation, the density information corresponds to two cells in which a difference between threshold values assigned to each cell of a dither matrix becomes a predetermined value. A method of combining data with an image, characterized in that the arrangement of two pixels of a dither image to be determined is determined based on other data to be mixed. 2. The image according to claim 1, wherein, when separating the data synthesized by the method, other data mixed by the arrangement of two pixels in the dither image is specified. How to combine data into 3. The value of two pixels in the dither image is represented by y _i , y
_j , if y _i = y _j, no data is synthesized, otherwise, if the data to be mixed is 1, y _i > y _j , if 0, y _i <y _j or mixed If the data to be set is 1, y _i <y _j , 0
3. The data synthesizing method according to claim 1, wherein the pixel replacement is performed as necessary so that y _i > y _j is satisfied, and the pixel arrangement is determined. 4. When separating data synthesized by the above method, if y _i = y _j, no data is synthesized, otherwise, if y _i > y _j , the mixed data is 1, The characteristic is that mixed data is specified as 0 if y _i <y _j , or mixed data is specified as 1 if y _i <y _j and 0 if y _i > y _j. 4. A method for synthesizing data to an image according to claim 3. 5. The method according to claim 1, further comprising, after specifying the mixed data, replacing pixels as necessary so that y _i > y _j so as to improve image quality, and determining a pixel arrangement. 5. A method for synthesizing data to an image according to claim 4, wherein the method is characterized in that: