JPH10145612A - Single color conversion system for color image and image processing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To extract a contrast image from a color image so that a difference between different color areas is most remarkable on the contrast image. SOLUTION: In the case of compositing RGB components in a color image at a prescribed ratio to extract a monochromatic contrast image F, a conversion vector is used to decide a ratio of the RGB components of the color image to emphasize a contrast difference between areas of each color to be identified included in the color image. A vector maximizing the dispersion in the contrast image F is adopted for the conversion vector. The conversion vector is obtained by calculating a mean value of each color component of the color image, obtaining a co-variance matrix of the color image from the mean value of each color component and color picture element data configuring the color image and calculating a proper vector corresponding to a proper value whose absolute value is maximum among proper values of the co-variance matrix.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color image sensor for detecting an edge to be measured from a color image obtained by imaging an object to be measured by an image pickup means such as a color CCD camera and recognizing a necessary area. The present invention relates to a color image monochrome conversion method and an image processing method for extracting a contrast image from an image.

[0002]

2. Description of the Related Art When extracting a desired contour or shape from a monochrome image of an object to be measured or an object to be recognized, a multi-tone monochrome image is usually extracted at a predetermined threshold level.
The edge image is extracted by converting into a value. On the other hand, in contrast to such a black and white image, a color image contains information on both surface physical properties such as the light reflectance of the object and the like, and therefore utilizes differences in brightness and colors of image signals. Therefore, there is an advantage that it is possible to detect a boundary line between regions having different light reflectances and colors (a boundary line between an object and a background region) and a region of a specific color. As a result, it is possible to measure the size and shape, and to recognize the presence / absence and the position.

Generally, a color image is an image of three colors of red (R), green (G), and blue (B), that is, R,
G and B are composed of three two-dimensional array data. For measurement processing and recognition processing, this color image is converted to a single color to obtain a contrast image as one two-dimensional array data, and the size and shape of the object are determined from the two-dimensional distribution of the density of the contrast image. It detects the presence or absence of the object. Therefore, the contrast image must clearly express the difference in brightness and the difference in color in the color image.

Conventionally, there are the following two methods for extracting a contrast image from a color image. (1) Method of Converting Luminance of Color Image to Contrast Image The first method is a method of converting a luminance signal Y of the NTSC color system into a contrast image F.
Is

[0005]

## EQU1 ## F = 0.30R + 0.59G + 0.11B

Is converted as follows. This contrast image F is equivalent to a black-and-white image obtained by receiving a color television broadcast on a black-and-white television. However, since this conversion is a conversion in which the ratio of each color component is fixed, if the luminances happen to be the same even if the colors are different, it may not be possible to distinguish the regions having different colors on the contrast image. For example, as shown in FIG. 8A, the target image S includes two regions A and B to be recognized, and these regions A and B
Is expressed as a ratio of (R, G, B) and is (150, 0, 0) and (0, 76, 0), respectively, along the horizontal scanning line H on the target image S. The image data of each color is as shown in R, G, B in FIG. 8B, and the contrast images Fa, Fb of the regions A, B are

[0007]

## EQU2 ## Fa = 0.30 × 150 = 45 Fb = 0.59 × 76 ≒ 45

Therefore, as shown in FIG. 1B, the regions A and B cannot be distinguished on the contrast image F.

(2) Method of Making Image of Specific Color a Contrast Image In the second method, when an area having a specific color is detected, the image of that color is separated from the color image, and a contrast image F How to Assuming that the color is red, for example, and the color vector (mixing ratio of RGB) is (1, 0, 0), the contrast image F is

[0010]

## EQU3 ## F = 1 · R + 0 · G + 0 · B = R

Is converted as follows. This method uses (1)
It can be said that the method is more flexible than the method of (1) because the color to be detected can be specified. However, since the color of the area adjacent to the area of the color to be detected is not taken into consideration, it may be impossible to distinguish areas having different colors as in the method (1). For example, the area A in FIG. 8A is red (150, 0,
0), when the area B is yellow (150, 150, 0), the image data of each color along the horizontal scanning line H is as shown by R, G, B in FIG. In this case, if an attempt is made to detect a red area, the contrast image F becomes equal to R as shown in FIG. 3C, so that the areas A and B cannot be distinguished.

[0012]

As described above, in the conventional single-color conversion method for a color image, regardless of the specific color arrangement in the target color image, the contrast image is always extracted by the same calculation formula, Since the color of the region to be extracted is extracted as a contrast image without considering the color of the region or the background region, there is a problem that regions having different colors may not be distinguished on the extracted contrast image.

The present invention has been made in view of such a point, and a color image extracting apparatus capable of extracting a contrast image from a color image so that a difference between regions having different colors appears most conspicuously on the contrast image. An object of the present invention is to provide a monochromatic conversion method. Another object of the present invention is to provide an image processing method that enables highly accurate measurement and recognition processing of an object from a color image using such a color image single-color conversion method. .

[0014]

According to a first aspect of the present invention, there is provided a color image single-color conversion system for combining color components of a color image at a predetermined ratio to extract a single-color contrast image. Conversion vector calculation means for calculating a conversion vector for determining a ratio of each color component of the color image so as to emphasize a contrast difference between regions of each color to be identified included in the image, and conversion vector calculation means for calculating the conversion vector. Contrast image conversion means for converting the color image into a contrast image using a conversion vector.

According to the present invention, a color vector is converted from a color image by using a conversion vector for determining a ratio of each color component of the color image so as to emphasize a contrast difference between regions of each color to be identified included in the color image. Extract a contrast image. For this reason, unlike the conventional method of specifying a fixed operation or a specific color, the difference in color between adjacent areas is emphasized in the contrast image,
It is possible to reliably distinguish between areas of different colors.

As the conversion vector, for example, a vector that maximizes the variance of the contrast image can be used. Preferably, the average value of each color component of the color image is calculated, and the average value of each color component is calculated. A variance-covariance matrix of the color image is obtained from each color pixel data constituting the color image, and an eigenvector corresponding to the largest eigenvalue of the eigenvalues of the variance-covariance matrix is obtained. According to this method, there is no need for prior information (such as identification of a color to be identified in an area) relating to a target color image, so that there is an advantage that an operation for generating a contrast image becomes extremely simple.

Further, when the representative color of each color region to be identified included in the color image is given as advance information, the conversion vector is used to convert the representative color between each color region to be identified into each color component. Each color component may be determined so as to be weighted in accordance with the magnitude of the contrast for each color component. As described above, when the representative color of each color region to be identified is given as the advance information, the amount of calculation processing for calculating the conversion vector can be extremely reduced.

As a method of calculating a representative color of each color to be identified, for example, the following two methods can be considered. The first method includes an area designating unit for designating a partial area from each color area to be identified of the color image, and calculates an average value of the colors of the partial areas respectively designated by the area designating unit. It is calculated as a representative color of each area. This method requires an operation of designating an area, but has the advantage that the amount of calculation for calculating a representative color can be reduced.

In a second method, an average value of each color component of the color image is calculated, and the variance-covariance of the color image is calculated from the average value of each color component and each color pixel data constituting the color image. A matrix is obtained, and among the eigenvalues of the variance-covariance matrix, the eigenvalue corresponding to the largest eigenvalue is calculated, and the average value of each color component of the color image is set as a boundary in the direction of the eigenvector on the three-dimensional color space. Each of the color pixel data is divided into groups, and the average color of the color pixel data belonging to each group is set as a representative color of each of the regions. This method requires a larger amount of calculation than the first method, but if a color image is obtained so as to include each color to be identified, the operation of specifying the area as in the first method becomes unnecessary. Becomes easy.

These methods include an image processing method for detecting a boundary of an object to be measured from a color image obtained by imaging the object to be measured and measuring its size and shape, and an image processing method for imaging an object to be recognized. The present invention can be applied to obtaining a contrast image for detecting a boundary in an image processing method or the like for detecting a boundary of a recognition target from the obtained color image and recognizing the recognition target. According to these image processing methods, it is possible to measure and recognize a target object with high accuracy from a color image containing more information than a black and white image.

[0021]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram for explaining the principle of a color image monochromatic conversion system according to an embodiment of the present invention. FIG. 1A shows a target image S composed of a color image.
Two areas A and B to be identified are included. Here, two methods for extracting the contrast image F from the image data along the horizontal scanning line H in the target image S so that the difference between the regions A and B appears most remarkably will be exemplified.

(1) Method of Extracting Contrast Image from Color Image (Part 1) Coordinate values of each pixel included in target image S are ij and pixel ij
Cij = (Rij,
Gij, Bij) ^T , and the conversion vector is V = (Vr, Vg,
Vb) Assuming that ^T , each pixel data Fij of the contrast image F can be obtained as in the following Expression 4.

[0023]

[Number 4] ^{Fij = Cij T V = VrRij +} VgGij + VbBij

In order to maximize the contrast on the contrast image F, the variance Ver of the contrast image F may be maximized. The variance Ver of the contrast image F is obtained as in the following Expression 5.

[0025]

(Equation 5)

Here, N is the number of pixels constituting the target image S. Further, F (with bars) is the average pixel level of the contrast image F, and when C (with bars) is the average pixel level of the color image C, it can be expressed as follows.

[0027]

(Equation 6)

From Equations 4 and 6, Equation 5 becomes Equation 7.

[0029]

(Equation 7)

Here, A is a variance-covariance matrix of the color image C, and is represented by the following equation (8).

[0031]

(Equation 8)

Since the transform vector V that maximizes Ver in Equation 7 becomes the eigenvector V corresponding to the largest eigenvalue λ _MAX among the eigenvalues λ of the variance-covariance matrix A, a well-known iterative method is used. You can ask for this.

That is, first, after obtaining the variance-dispersion matrix A, for example, (1,1,1) ^T is given as an initial value of V, and a conversion vector obtained by repeating the operation of the following equation 9 a plurality of times is obtained. V. The number of calculations is appropriately set in consideration of the processing speed and accuracy.

[0034]

V = AV

However, in the course of this calculation, V becomes Vr ² +
It must be normalized so that Vg ² + Vb ² = 1. As a result, the target contrast image F can be obtained by performing the operation of Expression 4 using the obtained conversion vector V.

For example, in the example of FIG. 1A, it is assumed that the ratio of the area A to the area B is 1: 8.
As shown in (b), the case where the average color pixel data Cij of the area A is (150, 0, 0) ^T and the average color pixel data Cij of the area B is (0, 76, 0) ^T Considering that, the average value of each color is

[0037]

(Equation 10)

The variance-covariance matrix A
Is as shown in Expression 11.

[0039]

[Equation 11]

In this case, the conversion vector V converges by one operation in the iterative method, and the obtained conversion vector V is V = (0.89
2, -0.452, 0) ^T. Therefore, the contrast image F
Let F = 0.892R−0.452G, the contrast image F along the horizontal scanning line H of the target image S in FIG.
(B), it is possible to obtain a contrast image F in which the contrast difference between the adjacent areas A and B is emphasized. One feature of this method is that some coefficients of the conversion vector have negative values in order to enhance contrast.

Further, as shown in FIG. 1C, the average color pixel data Cij of the area A is (150, 0, 0) ^T ,
The average color pixel data Cij of the area B is (150,
150, 0) ^T , when the same calculation is performed as above, a conversion vector V = (0, −1, 0) ^T is obtained, and the contrast image F becomes F = −G.
As shown in FIG. 1C, the contrast difference between the adjacent areas A and B is emphasized.

As described above, by calculating the conversion vector V that can obtain the maximum value of the variance of the contrast image F, the target image S is given only by providing the target image S without any other prior information. It is possible to extract a contrast image F that emphasizes a contrast difference between regions of each color in the image. This method is suitable for the auto mode because the operation is simple.

(2) Method for Extracting a Contrast Image from a Color Image (Part 2) The above-mentioned method (Part 1) requires no prior information.
Although the arithmetic processing is slightly complicated, when the information of the color of the region to be identified included in the target image S is given in advance as the advance information, the method described below is advantageous in that it requires a small amount of arithmetic. It is. The prior information required here is a region to be identified, for example, representative colors (Ra, Ga, Ba), (Rb, G) of the colors of the regions A and B in FIG.
b, Bb), and an average value or a predicted value of the color can be used as the representative color. Conversion vector V = (V
r, Vg, Vb) ^T can be obtained as in the following Expression 12.

[0044]

(Equation 12)

As a result, the operation of Equation 4 is performed using the obtained conversion vector V, and a target contrast image F can be obtained. This method has two regions A,
In B, each coefficient of the conversion vector V is determined so that a color component having a small contrast difference has a small weight and a color component having a large contrast difference has a large weight. Because it is small, it is suitable for, for example, a repeat mode in measurement processing of the same kind of object. Even when the conversion vector in each case of FIGS. 1B and 1C is calculated by this method, the V
^{= (0.892, -0.452,0) T,} V = (0, -1,0) T is required.

In this method, the representative colors of the areas A and B need to be given in advance as information. Therefore, two methods for obtaining the representative colors of the areas A and B will be described next. (2-1) Method of Determining Representative Colors of Regions A and B (Part 1) As shown in FIG. 2, an operator uses a man-machine interface to extract partial regions Sa, Specify Sb. The representative colors (Ra, Ga, Ba) and (Rb, Gb, Bb) of the areas A and B are calculated by the following equation (13). Here, Na and Nb are the numbers of pixels in the partial areas Sa and Sb, respectively.

[0047]

(Equation 13)

(2-2) Method of Obtaining Representative Colors of Regions A and B (Part 2) The method of extracting a contrast image from a color image (Part 1) is used to obtain the average values R, G, B (with bars) of the color image. ) And a conversion vector V are calculated. Next, Equation 14
, The representative color (Ra, Ga, Ba) of the area A is calculated. Na is the number of pixels satisfying the condition Lij ≧ 0.

[0049]

[Equation 14]

Next, the representative color (Rb, Gb, Bb) of the area B is calculated as shown in Expression 15. Nb is the number of pixels that satisfy the condition of Lij <0.

[0051]

(Equation 15)

The meaning of this processing will be described as follows. That is, an RGB space that represents a color image as shown in FIG. 3 is defined. When the average value (R, G, B) (with a bar) of the color image obtained by Expression 8 is plotted in the RGB space, the result is as shown in the figure. When a plane SV that passes this average value and has a transformation vector V = (Vr, Vg, Vb) as a normal vector is arranged in the RGB space, the pixel data of the target image S is divided into two groups with the plane SV as a boundary. a and b. That is, the pixel located on the side of the plane SV in the direction indicated by the conversion vector V is expressed by the following equation (1).
The pixel satisfying the condition of Lij ≧ 0 in 4 is a group a. Also, the conversion vector V of the plane SV
The pixel located on the opposite side to the direction indicated by is a pixel that satisfies the condition of Lij <0 in Expression 15, and is referred to as a group b. As described above, each pixel data included in the target image S is projected onto the axis of the transformation vector V, and the obtained 2
The average value of the two distributions is the representative color (R
a, Ga, Ba) and (Rb, Gb, Bb).

In the above description, it is assumed that not only chromaticity information contained in a color image but also lightness information is used. For this reason, it is also possible to distinguish between regions having the same color but different lightness. If it is desired to remove the lightness information from the color image and compare only the chromaticity information, each color pixel data Cij is normalized as shown in the following Expression 16 in the preceding stage of each processing, and the pixel of the chromaticity image is obtained. Find data cij,
The above-described processing may be performed using the pixel data cij of the chromaticity image as color pixel data.

[0054]

## EQU16 ## c ij = (Rij, Gij, Bij) ^T / (Rij + Gij + Bi
j) = (rij, gij, bij) ^T

By this conversion, the chromaticity image c becomes r + g
+ B = 1, but in the actual processing, r + g + b = 12
A predetermined value, such as 6, can be used. Performing such a conversion has the following advantages. In general, a color image to be subjected to image measurement or image recognition rarely has the same R, G, and B values even if the color is the same due to the shape and surface physical properties of the object and the influence of uneven illumination. However, it is generally known that, in a region of the same color, variation in chromaticity is relatively small even though variation in brightness is large. That is, it can be said that the chromaticity information of the color image has a relatively strong property against a change in light amount and a change in illumination. Therefore, more stable image processing can be performed by performing processing mainly using chromaticity information.

Next, an example in which the above method is specifically applied to a non-contact image measurement system will be described. FIG. 4 is a perspective view showing the entire configuration of the non-contact image measurement system. The system includes a non-contact image measurement type coordinate measuring machine 1, a computer system 2 which controls the driving of the coordinate measuring machine 1 and executes necessary data processing, and a printer 3 which prints out measurement results. It consists of.

The CMM 1 is configured as follows. That is, a measurement table 13 on which a work 12 is placed is mounted on the gantry 11.
3 is driven in the Y-axis direction by a Y-axis drive mechanism (not shown). Support arms 14 and 15 extending upward are fixed to the center of both side edges of the gantry 11.
An X-axis guide 16 is fixed so as to connect both upper ends of the X-ray guides 4 and 15. An imaging unit 17 is supported by the X-axis guide 16. The imaging unit 17 is driven along the X-axis guide 16 by an X-axis driving mechanism (not shown). A color CCD is provided at the lower end of the imaging unit 17.
A camera 18 is mounted so as to face the measurement table 13. Further, in addition to a lighting device (not shown) and a focusing mechanism (not shown), a color CC
A Z-axis drive mechanism for moving the position of the D camera 18 in the Z-axis direction is built in.

The computer system 2 includes a computer main body 21, a keyboard 22, a joystick box 23, a mouse 24, and a CRT display 25. The computer body 21 is, for example, as shown in FIG.
It is configured as shown in FIG. That is, color image information input from the color CCD camera 18 is stored in a color image memory 32 via an interface (hereinafter, referred to as an I / F) 31. The color image information stored in the color image memory 32 is displayed on the CRT display 25 via the display control unit 33. On the other hand, operator commands from the keyboard 22 and the mouse 24 are transmitted to the CPU 35 via the I / F 34. The CPU 35 extracts a contrast image from the target image in the designated area in accordance with a command from the operator or a program stored in the program memory 36, detects edges from the extracted contrast image, and measures the size and the size of the work 12 A series of processes such as measuring a shape are sequentially executed. That is, the CPU 35 and the predetermined program constitute the conversion vector calculating means and the contrast image extracting means of the present invention. Further, the work memory 37 provides a work area for various processes in the CPU 35.

The measuring system includes, for example, an auto mode and a manual mode, and in each mode, executes any one of a single mode, a teaching mode, and a repeat mode. The CPU 35
In the auto mode, a contrast image is automatically extracted from a target image. In the manual mode, a contrast image is extracted based on a partial area specified by operating the keyboard 22 and the mouse 24 by the operator. Also, C
In each of these modes, the PU 35 performs only one measurement from a given target image in the single mode, and in the teaching mode, obtains prior information (representative color of an area to be identified) necessary for repeated measurement. The processing sequence and the measurement program are stored in the program memory 36.
In the repeat mode, a process for continuously measuring the same kind of work 12 is executed according to the advance information or the program recorded in the program memory 36 in the teaching mode.

FIG. 6 is a flowchart showing a contrast image extracting process in the auto mode. CPU35
First, the measurement mode is determined (S1). If the measurement mode is the single mode, the contrast image F is extracted by using the above-described method (part 1) of extracting the contrast image from the color image. I do. That is, the average values R, G, and B (with bars) of the color pixel data of the target image S are calculated based on the aforementioned equation 8 (S2), and the variance-covariance matrix A of the color image is calculated based on the equation 8 (S3). Then, an eigenvector corresponding to the largest eigenvalue of the matrix A is obtained as a conversion vector V by the iterative method shown in Expression 9 (S4), and a contrast image F is calculated using the obtained conversion vector V (S4).
5).

When the measurement mode is the teaching mode (S1), an eigenvector corresponding to the largest eigenvalue of the absolute value of the matrix A is calculated by the same method as in the single mode described above (S6 to S8). The representative color (Ra, Ga, Ba) of the area A is obtained based on the equation (14) by using the method of obtaining the representative color of B (No. 2) (S
9), the representative color (Rb, Gb, B
b) is determined (S10). Then, the representative colors of the obtained areas A and B are recorded in the program memory 36 (S11).

If the measurement mode is the repeat mode (S1), the representative color of the areas A and B recorded in advance in the teaching mode is read (S12), and a contrast image is extracted from the color image (the method thereof). Using the method 2), the conversion vector V is calculated based on the equation 12 (S13), and the contrast image F is calculated using the obtained conversion vector V (S14).

On the other hand, in the manual mode, processing as shown in FIG. 7 is executed. First, when the measurement mode is the single mode (S21), the operator operates to
The partial areas Sa and Sb are specified (S22). Subsequently, the average color of the region Sa is calculated as the representative color of the region A (S2
3) The average color of the area Sb is calculated as the representative color of the area B (S24). Then, the conversion vector V is calculated based on Equation 12 (S25), and the contrast image F is calculated using the obtained conversion vector V (S26).

When the measurement mode is the teaching mode (S21), the partial areas Sa and Sb are designated by the operator as in the single mode (S2).
7) The average color of the area Sa is calculated as the representative color of the area A (S23), and the average color of the area Sb is calculated as the representative color of the area B (S24). Then, the obtained representative colors of the areas A and B are recorded in the program memory 36 (S
30).

In the repeat mode, the representative colors of the areas A and B recorded in advance in the teaching mode are read (S
31), and calculates a conversion vector V based on Equation 12 (S
32), a contrast image F is calculated using the obtained conversion vector V (S33).

As described above, by extracting a contrast image by a calculation method suitable for various measurement modes, it is possible to extract a contrast image F in which the contrast difference of each color region to be identified is enhanced from the color image. The size and shape of the work 12 can be accurately measured.

In the above embodiment, an example in which the present invention is applied to a non-contact image measurement system has been described. However, the present invention relates to any image processing system such as an image recognition device for extracting necessary information from a contrast image. It is needless to say that the present invention is applicable.

[0068]

As described above, according to the present invention, the conversion vector for determining the ratio of each color component of the color image so as to emphasize the contrast difference between the regions of each color to be identified included in the color image. Extract the contrast image from the color image using. For this reason, unlike the conventional method of specifying a fixed operation or specifying a specific color, the difference in color between adjacent areas is emphasized in the contrast image, and the area between different colors is reliably determined. This has the effect of enabling distinction.

[Brief description of the drawings]

FIG. 1 is a diagram for explaining a monochrome conversion method of a color image according to an embodiment of the present invention.

FIG. 2 is a diagram for explaining a first method for obtaining a representative value of two areas in the same method.

FIG. 3 is a diagram for explaining a second method for obtaining a representative value of two regions in the same method.

FIG. 4 is a perspective view of a non-contact image measurement system to which the method is applied.

FIG. 5 is a block diagram showing a configuration of a computer system in the system.

FIG. 6 is a flowchart showing a procedure for extracting a contrast image in an auto mode in the system.

FIG. 7 is a flowchart showing a procedure for extracting a contrast image in a manual mode in the system.

FIG. 8 is a diagram for explaining a conventional color image monochromatic conversion method.

[Explanation of symbols]

DESCRIPTION OF SYMBOLS 1 ... Coordinate measuring machine, 2 ... Computer system, 3 ... Printer, 11 ... Stand, 12 ... Work, 13 ... Measurement table, 14, 15 ... Support arm, 16 ... X-axis guide, 17
... Imaging unit, 18 ... Color CCD camera, 21 ... Computer body, 22 ... Keyboard, 23 ... Joystick box, 24 ... Mouse, 25 ... CRT display, 31,34 ... Interface, 32 ... Color image memory, 33 ... Display control unit , 35 CPU, 36 program memory, 37 work memory.

Claims (10)

[Claims] 1. A color image monochromatic conversion method for combining color components of a color image at a predetermined ratio to extract a monochromatic contrast image, wherein a contrast difference between regions of each color to be identified included in the color image is determined. Conversion vector calculation means for calculating a conversion vector for determining a ratio of each color component of the color image so as to emphasize; and contrast for converting the color image into a contrast image using the conversion vector calculated by the conversion vector calculation means. A color conversion method for a color image, comprising: an image conversion unit. 2. The color image monochromatic conversion method according to claim 1, wherein said conversion vector calculation means calculates a conversion vector that maximizes the variance of said contrast image. 3. The conversion vector calculation means calculates an average value of each color component of the color image, and calculates a variance of the color image based on the average value of each color component and each color pixel data constituting the color image. 3. A color image monochromatic conversion method according to claim 2, wherein a variance matrix is obtained, and an eigenvector corresponding to a largest eigenvalue among eigenvalues of the variance-covariance matrix is calculated as the conversion vector. . 4. The conversion vector calculation means, based on a representative color of each color region to be identified included in the color image given as prior information, converts the representative color for each color component between these color regions. 2. A color image single-color conversion method according to claim 1, wherein the conversion vector is determined so that a weight corresponding to the magnitude of the contrast is given to each color component as compared with. 5. An area designating means for designating a partial area from each color area to be identified in the color image, and calculating an average value of the colors of the partial areas designated by the area designating means in each of the areas. 5. A representative color calculating means for calculating as a representative color of the image data, wherein a representative color of each area calculated by the representative color calculating means is supplied to the conversion vector calculating means. A single color conversion method for color images. 6. An average value of each color component of the color image is calculated, and a variance-covariance matrix of the color image is obtained from the average value of each color component and each color pixel data constituting the color image. Calculating an eigenvector corresponding to the largest eigenvalue of the eigenvalues of the covariance matrix,
The color pixel data is grouped in the direction of the eigenvector on the three-dimensional color space with the average value of each color component of the color image as a boundary, and the average color of the color pixel data belonging to each group is represented by the representative color of each region. 5. The color image according to claim 4, further comprising: a representative color calculating unit that calculates a color as a color, wherein a representative color of each area calculated by the representative color calculating unit is supplied to the conversion vector calculating unit. Single color conversion method. 7. The color image monochromatic conversion method according to claim 1, wherein the color image includes chromaticity information and lightness information. 8. The color image monochromatic conversion method according to claim 1, wherein the color image is a chromaticity image from which lightness information has been removed. 9. A contrast image is extracted from a color image obtained by imaging an object to be measured by a color image monochromatic conversion method according to any one of claims 1 to 8, and the extracted contrast image is extracted. An image processing method, wherein a size or a shape of the measured object is measured by detecting a boundary of the measured object with respect to a contrast image. 10. A contrast image is extracted from a color image obtained by imaging an object to be recognized by a color image monochromatic conversion method according to any one of claims 1 to 8, and the extracted contrast image is extracted. An image processing method, wherein the recognition target is recognized by detecting a boundary of the recognition target in a contrast image.