JP2002368983A - Picture processor and picture processing method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To solve the problem that the number of samples constituting the feature value space of an observation picture is not sufficient in a region where the density of the picture is low and therefore the detection precision of access control information remarkably drops. SOLUTION: The Mahalanobis distance MD1 between the feature value space (the set of feature value vectors) of a reference picture and a reference vector is calculated (S1). M×N pixel blocks are cut from a picture signal as the observation picture (S3 to S7). The number of samples which can be used for constituting the feature value space of the observation picture is compared with a prescribed value (S10). When the number of samples < the prescribed value is realized, the area of the observation picture is extended to a plurality of blocks (S12). The Mahalanobis distance MD2 between the feature value space of the observation picture and the reference vector is calculated (S14). The feature value space of the observation picture is moved so that the relation of MD1 and MD2 is controlled in accordance with an access control signal (S15 to S17). Then, the observation picture is reconstituted based on feature value space after movement (S18).

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

[0001] 1. Field of the Invention [0002] The present invention relates to an image processing apparatus and method, for example, to image processing for embedding a control signal in an image.

[0002]

2. Description of the Related Art The image quality of digital image forming apparatuses such as printers and copiers has been remarkably improved, and in recent years, high quality printed matter can be easily obtained. That is, anyone can obtain printed matter of desired image quality by performing image processing using a high-performance scanner, printer, copier, and computer. As a result, problems such as unauthorized copying of printed documents (hereinafter simply referred to as “printed materials”) occur,
There is an access control function that prevents or suppresses unauthorized use of printed matter due to unauthorized copying.

[0003] This function is generally implemented by embedding access control information in a printed matter invisibly, and by embedding a bitmap pattern (glyph code, DD code, etc.) corresponding to the access control information in a margin of the printed matter. ,
Alternatively, it is realized by embedding scrambled encryption in a document image. Methods for invisiblely embedding access control information include a type for controlling the space between English words, a type for rotating characters, and a type for enlarging / reducing characters.

FIG. 1 is a view for explaining a method of embedding access control information by controlling the space between English words. As shown below, two consecutive spaces p and s represent one bit.

FIG. 2 is a diagram for explaining a method of embedding access control information by controlling the rotation angle θ of a character.

FIG. 3 is a view for explaining a method of embedding access control information by controlling enlargement / reduction of characters. For example, 2701 indicates the original size, and 2702 indicates the reduced size.

[0007]

When the access control information is embedded by the method shown in FIGS. 1 to 3, in order to accurately detect the access control information, the space width, the rotation angle, the enlargement / If you use a low-resolution scanner that needs to read the reduced size and is inexpensive,
Access control information cannot be detected accurately.

[0008] Small and complicated characters (for example, kanji)
It is difficult to accurately detect the access control information from the printed matter mainly by the user.

In addition, embedding of access information causes a sense of incongruity in characters and spaces and changes in the density of characters, resulting in subjective deterioration of image quality.

There is also a method of embedding access control information by allocating an embedding area per unit information (1 bit) with a fixed area. In other words, the method of embedding 1 bit of access control information in the area of M × N pixels, but in the area where the density of the image such as characters, tables, graphs, etc. is low, the number of samples is not enough to compose the feature space of the observed image As a result, the detection accuracy of the access control information may be extremely reduced.

FIG. 4 is a diagram showing an example of a document image including an area having a small number of samples for forming a feature amount space of an observation image. Areas 1901, 1902, 1903, and 1904 are areas having a small number of samples.

An object of the present invention is to solve the above-mentioned problems individually or collectively, and it is an object of the present invention to provide information embedding capable of accurately detecting embedded information.

It is another object of the present invention to prevent an image from being unnatural before and after embedding information.

It is another object of the present invention to provide information embedding in which the detection accuracy does not decrease due to the density of an image of a document.

[0015]

The present invention has the following configuration as one means for achieving the above object.

An image processing apparatus according to the present invention is an image processing apparatus for embedding a predetermined control signal in an image signal output to a printing apparatus, and cuts out an M × N pixel block from the image signal as an observation image. Expansion means for calculating the number of samples for extracting the feature amount, and when the calculated number of samples is smaller than a predetermined value, expanding the observation image to an adjacent block; Extracting means for extracting as a set of vectors, and generating means for generating, from a reference image sufficiently larger than the M × N pixel block, a set of second feature vectors having the same dimension as the first feature vector. Calculating a Mahalanobis distance (MD1) of the reference image based on the reference vector having the same number of dimensions as the first feature vector and the second feature vector, Calculating means for calculating a Mahalanobis distance (MD2) of the observed image based on the first feature amount vector, and the MD in accordance with the predetermined control signal.
Control means for controlling the value of the first feature amount vector, so as to control the relationship between 1 and MD2, and reconstructing the observation image based on a set of the first feature amount vectors whose values are controlled. Reconstructing means.

An image processing apparatus for extracting a control signal embedded in an image signal, comprising:
Expanding means for extracting a pixel block as an observation image, calculating the number of samples for extracting the feature amount thereof, and expanding the observation image to an adjacent block if the calculated number of samples is smaller than a predetermined value; and Extraction means for extracting the feature quantity of the first feature quantity vector as a set of first feature quantity vectors, and a second feature quantity having the same dimension as the first feature quantity vector from a reference image sufficiently larger than the M × N pixel block. Generating means for generating a set of vectors, calculating the Mahalanobis distance (MD1) of the reference image based on the reference vector having the same dimension number as the first feature vector and the second feature vector, the reference vector Calculating means for calculating the Mahalanobis distance (MD2) of the observed image based on the first feature amount vector; and the image signal based on the similarity between the MD1 and MD2. It characterized by having a determining means for embedded control signals.

An image processing method according to the present invention is an image processing method for embedding a predetermined control signal in an image signal output to a printing apparatus, wherein an M × N pixel block is cut out from the image signal as an observation image. Calculate the number of samples for extracting the feature amount, and if the calculated number of samples is smaller than a predetermined value, extend the observed image to an adjacent block, and set the feature amount of the observed image to a set of first feature amount vectors. Extracted from the reference image sufficiently larger than the M × N pixel block, to generate a set of second feature amount vector of the same dimension number as the first feature amount vector, the first feature amount vector and A Mahalanobis distance (MD1) of the reference image is calculated based on the reference vector having the same dimension number and the second feature amount vector, and based on the reference vector and the first feature amount vector. Calculating the Mahalanobis distance (MD2) of the observation image, and calculating the M value according to the predetermined control signal.
Controlling the relationship between D1 and MD2, controlling the value of the first feature amount vector, and reconstructing the observation image based on a set of the first feature amount vectors whose values are controlled. And

An image processing method for extracting a control signal embedded in an image signal, comprising the steps of:
The pixel block is cut out as an observation image, the number of samples for extracting the feature amount is calculated, and if the calculated number of samples is smaller than a predetermined value, the observation image is extended to an adjacent block, and the feature amount of the observation image is calculated. Is extracted as a first feature amount vector set, and from the reference image sufficiently larger than the M × N pixel block, a second feature amount vector set having the same dimension number as the first feature amount vector is generated. ,
Calculate the Mahalanobis distance (MD1) of the reference image based on the reference vector having the same dimension number as the first feature vector and the second feature vector, based on the reference vector and the first feature vector. A Mahalanobis distance (MD2) of the observed image is calculated, and a control signal embedded in the image signal is determined based on the similarity between the MD1 and MD2.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, an image processing apparatus according to an embodiment of the present invention will be described in detail with reference to the drawings.

In the embodiment, an example in which the present invention is applied to a monochrome laser beam printer (hereinafter, simply referred to as a "printer") will be described. The document image is a binary black-and-white image, and a low-cost scanner is used for an image reading apparatus that reads printed matter.

FIG. 5 is a block diagram showing an example of conversion of document image data in the printer.

First, if the document image data is binary image data 10
It is passed as 1 to the printer driver for the printer. The printer driver sequentially converts the binary image data 101 into device binary data 102, multi-valued K image data 103, and binary K image data 104 that match the characteristics of the printer. The binary K image data 104 is passed to a printer engine, and a high-resolution document image is printed on recording paper. In addition,
The document image data represents a document, a table, a graph, and the like.

Next, the feature space and the reference vector of the observation image and the reference image according to the present embodiment will be described.

FIG. 6 is a diagram showing an example of an observation image. Figure 6
(a) is an observed image of M × N pixels corresponding to a part of the binary K image data 104, and FIG. 6 (b) is a contour image obtained by extracting the contour part of the observed image and thinning it into one pixel. Shown respectively.

FIG. 7 is a view for explaining a direction index index for extracting the feature amount of the target pixel when the center pixel of the 3 × 3 pixel block is set as the target pixel. FIG. 7 shows an example in which there is an effective pixel having a pixel value> 0 in the t1, t2, t3, and t8 directions with respect to the target pixel, and there is no effective pixel in the t4, t5, t6, and t7 directions.

FIG. 8 is a diagram showing the values of the feature values t1 to t8 of the direction index of the target pixel shown in FIG. That is, a feature value of “1” indicates that there is an effective pixel in that direction, and a feature value of “0” indicates that there is no effective pixel in that direction. When the feature amount of the direction index of the target pixel Pij shown in FIG. 8 is represented by an eight-dimensional vector, the following expression is obtained. Hij = (1, 1, 1, 0, 0, 0, 0, 1)

The feature space of the observed image is a set of feature vectors of the direction index in each pixel of the contour image shown in FIG. Note that the dimension number of the feature space of the observed image is “8”.

FIG. 9 is a diagram showing the characteristic amount of the direction index of the reference vector. The characteristic amount in the t1 and t5 directions is '1'. Therefore, the reference vector is represented by the following equation. DIST = (1, 0, 0, 0, 1, 0, 0, 0)

The feature space of the reference image is represented by S shown in FIG.
It is a set of feature amount vectors of the direction index of each pixel in the contour image shown in FIG. 10B of the reference image of × T pixels. Note that the number of dimensions of the feature space of the reference image is “8”. S × T is set to a value sufficiently larger than M × N.

[Embedding of access control information]
A procedure for embedding a bit sequence of access control information shown in FIG. 11 in a document image will be described. The access control information is for preventing data conversion, copying, etc. due to unauthorized reading of the printed matter, such as restricting access to the image of the printed matter, preventing tampering, searching for the original, and the like. Has functions.

FIGS. 12 and 13 are flowcharts showing an example of a procedure for embedding access control information in a document image.

First, the Mahalanobis distance between the feature space (a set of feature vectors) of the reference image and the reference vector
(Mahalanobis's generaliezd distance) MD1 is calculated by the following equation (S1). D ² = (x-μ) 'Σ ^-1 (x-μ) where x is the reference vector μ is the mean vector of the feature space of the reference image Σ ^-1 is the inverse matrix of the variance-covariance matrix D ² is MD1

Then, the binary K image data 104 (see FIG. 5)
Is read (S2), and the document image is divided into M × N blocks as shown in FIG. In FIG. 14, reference numeral 2101 denotes a block, 2102 denotes a row number index, and 2103 denotes a column number index. Therefore, the index numbers are an M × N matrix. An M × N-dimensional index vector is generated from the index matrix (S3). At this time, the element number of the index vector becomes the index number of each block.

Next, a random number of R _{M × N} is generated from R ₁ based on the seed information determined in advance or input by the user (S 4). The generated random number corresponds to the element number of the index vector. That is, each random value is an index number of each block. Then, the bit sequence of the access control information shown in FIG. 11 is associated with a random number sequence (block index number) (S5). At this time, when the association reaches the end of the bit sequence, the next association returns to the head of the bit sequence.

FIG. 15 is a diagram showing an example of the correspondence between the bit sequence 1701 and the random number sequence 1702. Although details will be described later, the bit sequence 17
The number of 01 and the number of the random number sequence 1702 do not always match. That is, a plurality of random numbers (in other words, blocks) may correspond to one bit of the access control information.

In this case, a plurality of blocks correspond to one bit, and the number of bit sequences 1401 does not always match the number of 1402, which is a random number sequence and also a block index number. In the example of FIG. 15, three blocks correspond to one bit.

Next, “1” is set to the variable X and “0” is set to the variable k (S6), and the index number of the block corresponding to the X-th (R _X ) of the random number sequence 1702 and the bit of the access control information are set. Is acquired (S7), and an image of a block corresponding to the acquired block index number is acquired (S8).

Next, the number of samples that can be used to compose the feature space of the acquired image is calculated (S9), and the calculated number of samples is compared with a predetermined value (S10).
If the predetermined value increments the variable k (S11), blocks corresponding to the region of the observation image to the random number R _X, and the adjacent
The k-th block is set as the observation image area (S1
2) Return to step S8 to acquire images of those blocks. For example, the first time (k = 1) in step S11 is R _X and R _X +1
, The second (k = 2) is R _X , R _X +1 and R _X
Three blocks corresponding to +2, ... become observation images.

If the result of comparison between the calculated number of samples and the predetermined value is smaller than the predetermined number of samples, the variable k is cleared to zero (S1).
3) Using the obtained block image as an observation image, a Mahalanobis distance (MD2) between a feature space (a set of feature vectors) of the observed image and a reference vector is calculated by the following equation (S14). D ² = (x-μ) 'Σ ^-1 (x-μ) where x is the reference vector μ is the mean vector of the feature space of the observed image Σ ^-1 is the inverse matrix of the variance-covariance matrix D ² is MD2

If it is determined in step S15 that the bit of the access control information is '0', the entire feature space of the observed image is moved so that MD1> MD2 + defMD (S16).
If the bit is “1”, the entire feature space of the observed image is moved so that MD1> MD2 + defMD (S17). DefMD is a preset value. However, steps S16 and S17
Is performed in the direction of increasing the value of each element of the feature amount vector of the observed image, and is not performed in the direction of decreasing the value of each element.

Next, the observed image is reconstructed based on the feature space of the moved observed image (S18), the variable X is incremented (S19), and the determination of X> M × N is made (S20). , Step
The processing from S7 to S19 is repeated up to a random number RM _{× N.}

The binary K image data obtained in this way is passed to the printer engine, and the image is printed on the recording paper.

[Extraction of Access Control Information] FIG. 16 and FIG.
17 is a flowchart showing a procedure for extracting access control information from a printed material.

First, the Mahalanobis distance between the feature space (a set of feature vectors) of the reference image and the reference vector
MD1 is calculated by the following equation (S21). However, according to the following equation. D ² = (x-μ) 'Σ ^-1 (x-μ) where x is the reference vector μ is the mean vector of the feature space of the reference image Σ ^-1 is the inverse matrix of the variance-covariance matrix D ² is MD1

Next, the printed matter is read in an 8-bit gray scale mode by a scanner (S22), a contour image is extracted and a contour image thinned to one pixel is generated (S23), and the contour image is obtained by a registration technique. The size is converted to the size when the access control information is added (S24).

Then, as shown in FIG.
It is divided into × N blocks, and an M × N-dimensional index vector is generated from the M × N matrix index matrix (S2
Five). Then, a random number of R _{M × N} is generated from R ₁ based on predetermined seed information or seed information input by the user.
(S26). The generated random number corresponds to the element number of the index vector. That is, each random value is an index number of each block.

Next, "1" is set to the variable X (S27), the index number of the block corresponding to the X-th (R _X ) of the random number sequence 2 is obtained (S28), and the index number of the obtained block is set as the index number. The contour image of the corresponding block is obtained (S29).

Next, the number of samples that can be used to compose the feature space of the acquired image is calculated (S30), and the calculated number of samples is compared with a predetermined value (S31), and the number of samples is smaller than the predetermined number. if the value increments the variable k (S32), blocks corresponding to regions to the random number R _X observed images, and sets the block to the k-th adjacent to the region of the observation image (S3
3) Return to step S29 to acquire images of those blocks. For example, the first time (k = 1) in step S33 is R _X and R _X
2 blocks corresponding to +1, the second (k = 2) is R _X , R _X +1 and
Three blocks corresponding to R _X +2, ... are the observation images.

If the result of the comparison between the calculated number of samples and the predetermined value is the number of samples <the predetermined value, the variable k is cleared to zero (S3
4) The Mahalanobis distance (MD2) between the feature space (a set of feature vectors) of the observed image and the reference vector is calculated by the following equation using the acquired contour image of the block as the observed image (S35). D ² = (x-μ) 'Σ ^-1 (x-μ) where x is the reference vector μ is the mean vector of the feature space of the observed image Σ ^-1 is the inverse matrix of the variance-covariance matrix D ² is MD2

The similarity g of the Mahalanobis distances MD1 and MD2 thus obtained is calculated by the following equation (S36). g = MD1-MD2

Then, the similarity g> defMD is determined (S3
7) If g> defMD, the embedded bit is determined to be '0' (S38), otherwise the embedded bit is determined to be '1' (S39). DefMD is a preset value.

Next, the variable X is incremented (S40), and by determining that X> M × N (S41), steps S28 to S40 are performed.
Is repeated up to the random number R _{M × N.}

The bit sequence extracted in this way is subjected to a majority decision to reproduce the bit length at the time of embedding.

According to the present embodiment, even when a low-cost scanner with a low resolution is used, the access control information can be accurately detected.

Further, there is no occurrence of a sense of incongruity due to embedding of access information or a subjective deterioration in image quality due to a change in character density.

Further, for example, an area 1901 having a small number of samples for forming a feature amount space of an observation image shown in FIG.
In 1902, 1903 and 1904, since the access control information is embedded including the adjacent area, it is possible to prevent the detection accuracy from being lowered due to the density of the image of the document.

[0058]

[Other Embodiments] Even if the present invention is applied to a system including a plurality of devices (for example, a host computer, an interface device, a reader, a printer, etc.), an apparatus (for example, a copying machine) Machine, facsimile machine, etc.).

An object of the present invention is to supply a storage medium (or a recording medium) in which a program code of software for realizing the functions of the above-described embodiments is recorded to a system or an apparatus, and a computer (a computer) of the system or the apparatus It is needless to say that the present invention can also be achieved by a CPU or an MPU) reading and executing the program code stored in the storage medium. In this case, the program code itself read from the storage medium implements the functions of the above-described embodiment, and the storage medium storing the program code constitutes the present invention. Also,
When the computer executes the readout program code, not only the functions of the above-described embodiments are realized, but also the operating system (OS) running on the computer based on the instructions of the program code.
It goes without saying that a case where the functions of the above-described embodiments are implemented by performing some or all of the actual processing, and the processing performs the functions of the above-described embodiments.

Further, after the program code read from the storage medium is written into the memory provided in the function expansion card inserted into the computer or the function expansion unit connected to the computer, the program code is read based on the instruction of the program code. Needless to say, the CPU included in the function expansion card or the function expansion unit performs part or all of the actual processing, and the processing realizes the functions of the above-described embodiments.

When the present invention is applied to the storage medium, the storage medium stores program codes corresponding to the flowcharts described above.

[0062]

As described above, according to the present invention,
Information embedding capable of accurately detecting embedded information can be provided.

Further, it is possible to prevent the image from being uncomfortable before and after embedding information.

Further, it is possible to provide information embedding that does not lower the detection accuracy due to the density of the image of the document.

[Brief description of the drawings]

FIG. 1 illustrates a method of embedding access control information by controlling the space between English words,

FIG. 2 is a view for explaining a method of embedding access control information by controlling a rotation angle θ of a character;

FIG. 3 is a view for explaining a method of embedding access control information by controlling enlargement / reduction of characters;

FIG. 4 is a diagram showing an example of a document image including an area having a small number of samples for forming a feature amount space of an observation image;

FIG. 5 is a block diagram showing a conversion example of document image data in the printer.

FIG. 6 is a diagram showing an example of an observation image.

FIG. 7 is a view for explaining a direction index index for extracting a feature amount of a target pixel when a center pixel of a 3 × 3 pixel block is set as a target pixel;

FIG. 8 shows the feature values t1 to t of the direction index of the target pixel shown in FIG.
Diagram showing the value of 8,

FIG. 9 is a diagram showing a feature amount of a direction index of a reference vector;

FIG. 10 is a diagram illustrating a set of feature amount vectors of the direction index of each pixel in a contour image of a reference image of S × T pixels;

FIG. 11 is a diagram showing an example of a bit sequence of access control information;

FIG. 12 is a flowchart showing an example of a procedure for embedding access control information in a document image,

FIG. 13 is a flowchart showing an example of a procedure for embedding access control information in a document image;

FIG. 14 is a diagram illustrating block division of an image.

FIG. 15 is a diagram showing an example of correspondence between a bit sequence and a random number sequence;

FIG. 16 is a flowchart showing a procedure for extracting access control information from a printed material,

FIG. 17 is a flowchart illustrating a procedure for extracting access control information from a printed material.

 ──────────────────────────────────────────────────の Continued on the front page F term (reference) 2C061 AP01 AP04 AQ06 CL10 2C087 AA09 AA18 AB05 AC08 BA14 BB10 BD06 BD24 5B057 AA11 BA02 CA08 CA12 CA16 CB18 CC01 CE08 CH08 5C076 AA14 BA06

Claims (12)

[Claims] 1. An image processing apparatus for embedding a predetermined control signal in an image signal output to a printing apparatus, comprising: extracting an M × N pixel block from the image signal as an observation image; Expansion means for calculating the number of samples and, if the calculated number of samples is smaller than a predetermined value, extending the observation image to adjacent blocks; and extracting the feature amount of the observation image as a first set of feature amount vectors. Means, from a reference image sufficiently larger than the M × N pixel block,
Generating means for generating a set of second feature value vectors having the same dimension as the first feature value vector; and a reference vector having the same dimension as the first feature value vector and the second feature value vector. Calculating the Mahalanobis distance (MD1) of the reference image based on the reference vector and the Mahalanobis distance (MD2) of the observed image based on the first feature amount vector, and Control means for controlling the value of the first feature amount vector so as to control the relationship between the MD1 and MD2, and re-creating the observation image based on a set of the first feature amount vectors whose values are controlled. An image processing apparatus, comprising: reconstructing means for configuring. 2. The image processing apparatus according to claim 1, wherein the image signal represents at least one of a document, a table, and a graph. 3. The control signal according to claim 1, wherein the predetermined control signal is for preventing use of an image printed from the image signal due to unauthorized reading.
An image processing apparatus according to claim 1. 4. The apparatus according to claim 3, wherein the predetermined control signal has at least one function of restricting access to the image or preventing tampering, and searching for a document. Image processing device. 5. The extraction means generates a contour image in which a contour portion extracted from the observation image is thinned to one pixel, and deletes an inverted image of the contour image from the observation image to form a thin line in one pixel. Contour image generating means for generating a contour image, and feature value extracting means for setting a pixel of interest at a predetermined pixel interval and extracting the feature vector indicating the directionality of the contour image with respect to the pixel of interest. 5. The image processing device according to claim 1, comprising: 6. The apparatus according to claim 1, wherein the control means controls a magnitude relationship or a ratio of the MD1 and MD2.
The image processing device according to any one of claims 1 to 5. 7. An image processing device for extracting a control signal embedded in an image signal, wherein an M × N pixel block is cut out from the image signal as an observation image, and the number of samples for extracting the feature amount is calculated. And if the calculated number of samples is smaller than a predetermined value, extending means for extending the observed image to an adjacent block; extracting means for extracting a feature amount of the observed image as a first set of feature amount vectors; From a reference image that is sufficiently larger than the M × N pixel block,
Generating means for generating a set of second feature value vectors having the same dimension as the first feature value vector; and a reference vector having the same dimension as the first feature value vector and the second feature value vector. Calculating means for calculating the Mahalanobis distance (MD1) of the reference image based on the reference vector and the first feature amount vector, and calculating the Mahalanobis distance (MD2) of the observation image based on the reference vector and the first feature amount vector; and a similarity between the MD1 and MD2. Determining means for determining a control signal embedded in the image signal based on the image signal. 8. The image processing apparatus according to claim 7, wherein the similarity is represented as a difference between the MD1 and the MD2. 9. An image processing method for embedding a predetermined control signal in an image signal output to a printing apparatus, the method comprising: extracting an M × N pixel block from the image signal as an observation image; The number of samples is calculated, and if the calculated number of samples is smaller than a predetermined value, the observed image is extended to an adjacent block, and the feature amount of the observed image is extracted as a first set of feature amount vectors. From a reference image that is sufficiently larger than the N pixel block,
Generate a set of second feature vectors having the same dimension as the first feature vector, and refer to the reference based on the reference vector and the second feature vector having the same dimension as the first feature vector. Calculate the Mahalanobis distance of the image (MD1), calculate the Mahalanobis distance of the observed image based on the reference vector and the first feature amount vector (MD2), and calculate the MD1 and MD2 according to the predetermined control signal. Controlling the value of the first feature vector so as to control the relationship, and reconstructing the observation image based on a set of the first feature vectors whose values are controlled. Method. 10. An image processing method for extracting a control signal embedded in an image signal, wherein an M × N pixel block is cut out from the image signal as an observation image, and the number of samples for extracting the feature amount is calculated. If the calculated number of samples is smaller than a predetermined value, the observation image is extended to an adjacent block, and the feature amount of the observation image is extracted as a first set of feature amount vectors. From a sufficiently large reference image,
Generate a set of second feature vectors having the same dimension as the first feature vector, and refer to the reference based on the reference vector and the second feature vector having the same dimension as the first feature vector. Calculate the Mahalanobis distance (MD1) of the image, calculate the Mahalanobis distance (MD2) of the observed image based on the reference vector and the first feature amount vector, and calculate the image signal based on the similarity between the MD1 and MD2. Determining an embedded control signal in the image processing method. 11. A program for controlling an image processing apparatus to execute the image processing according to claim 9 or 10. 12. A recording medium on which the program according to claim 11 is recorded.