JPH0744709A - Color discrimination device - Google Patents

Abstract

PURPOSE:To provide a color discrimination device which discriminates colors and reduces the colors when the point distributions of picture elements in a color space are linear ones straddling several clusters and when the discrimination of the number of the colors is difficult. CONSTITUTION:The linear point distributions (101-103) in the color space are detected by Hough-transforming, the point distributions on the color space, which are obtained by mapping the picture elements of a color picture in a specified area on the color space, so as to obtain the number of the colors. The main components of the point distributions on the color space is analyzed and the first main component (104) and the second main component (105) are used. Thus, the number of the colors in the color picture area is obtained. The point distributions on the color space, which are obtained by mapping the respective picture elements of the color picture area where the colors are to be reduced on the color space, are Hough-transformed. Thus, the linear point distributions (101-103) on the color space are detected, and the colors are reduced by using the colors for the number of the clusters obtained by clustering the linear point distributions.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color discriminating apparatus for obtaining the number of colors in a specific area of a color image.

[0002]

2. Description of the Related Art Conventionally, when obtaining the number of colors in a specific area of a color image, a point distribution obtained by mapping each pixel in the area onto a certain color space is clustered to obtain the number of clusters. I was looking for the number of colors.

As such a clustering method, for example, as disclosed in Japanese Unexamined Patent Publication No. 4-61558, color images are mapped into a color space, and clustering is performed using principal component analysis and distance comparison. Things are known. FIG. 19 shows an example of clustering. This figure shows the result of mapping color images in which white, black, red, and blue exist on the RGB color space and clustering the distribution. The black dots in the figure indicate the points where each pixel of the color image is mapped in the RGB color space, and the area surrounded by the circle forms one cluster. In this example, 4
Since it is divided into two clusters, the number of colors in the color image is “4”.

Further, it is possible to reduce the number of colors by replacing the color in each of the clustered areas with one color. That is, for example, when the pixel value represented by each point in the cluster as shown in FIG. 19 is replaced with the center of gravity of the distribution or the average color, it becomes as shown in FIG. By this operation, a color image having only four colors in FIG. 20 is converted.

[0005]

Various methods have been proposed as a method for clustering a point distribution obtained by mapping each pixel in a region on a certain color space, but there is a general-purpose method. At present, the clustering method is selected according to the state of the point distribution. Clustering may be difficult depending on the point distribution. As an example in which clustering is difficult, for example, the point distribution may be distributed over several clusters as shown in FIG.

For example, the reason why the distributions of the white color 301 and the red color 303 are connected to each other is that when a color image is read, the boundary between the white color 401 of the background and the red color 402 written on the color of the color sensor 403 as shown in FIG. Pixel 4
No. 05 reads, and a color in which these two colors are mixed is generated, or both the light reflected by the white color 401 of the background and the light reflected by the red color 402 written on the ground as shown in FIG. At the same time, it is presumed that this occurs because the pixel 406 of the color sensor 403 reads.

In addition to the above, as a cause of the distribution of the white color 301 and the black color 302 being connected to each other, it is considered that shadows are generated due to the bending and wrinkles of the base, and the white color becomes blackish when these are read. To be

For these reasons, the distributions are connected as shown in FIG. 5, making clustering difficult. If you make a mistake in clustering, the number of colors extracted will be different,
It also affects color reduction. For example, in FIG. 5, the color should originally be determined to be four colors, but there is a risk that it will be determined to be one color as indicated by the broken line due to clustering, and if the number of colors is reduced, one color image will be obtained.

Therefore, according to the present invention, the point distribution of pixels in the color space is a linear distribution that extends over several clusters, and when it is difficult to determine the number of colors, the linear point distribution of the color space An object of the present invention is to provide a color discriminating apparatus capable of discriminating the number of colors in a specific region of a color image and reducing the number of colors by detecting a position and analyzing a principal component of a point distribution on a straight line.

[0010]

The color discriminating apparatus of the present invention comprises:
A color image input means for inputting a color image, an area designating means for designating an image area for obtaining the number of colors for the color image, and an image area designated by the area designating means are inputted by the color image input means. Cutting means for cutting out from the color image, mapping means for mapping each pixel of the image area cut out by the cutting means on a color space, and point distribution of the pixels mapped by the mapping means on the color space It is provided with a color number discriminating means for discriminating the number of colors and a color number outputting means for outputting the color number discriminated by the color number discriminating means.

Further, the color discriminating apparatus of the present invention comprises a color image input means for inputting a color image, and an area designating means for designating an image area for obtaining the number of colors for the color image.
Cutting out means for cutting out the image area designated by the area designating means from the color image input by the color image inputting means, and mapping for mapping each pixel of the image area cut out by the cutting out means onto a color space. Means and each point on the color space of the pixel mapped by this mapping means is Hough (H
a conversion means for performing a conversion, a color number discrimination means for discriminating the number of colors by detecting the maximum point of this distribution from the distribution of each point on the color space Hough-transformed by this conversion means, and this color The color number output means outputs the number of colors determined by the number determination means.

Further, the color discriminating apparatus of the present invention comprises a color image input means for inputting a color image, and an area designating means for designating an image area for obtaining the number of colors for the color image,
Cutting out means for cutting out the image area designated by the area designating means from the color image input by the color image inputting means, and mapping for mapping each pixel of the image area cut out by the cutting out means onto a color space. Means, a principal component analysis means for principal component analysis of the point distribution in the color space of the pixels mapped by this mapping means, and a value of the first principal component obtained by this principal component analysis means is preset. A first principal component comparing means for obtaining the number of colors by comparing it with a value, and a first principal component comparing means for obtaining the number of colors by comparing the value of the second principal component obtained by the principal component analyzing means with a preset predetermined value. It comprises two principal component comparing means, a first principal component comparing means, and a color number outputting means for outputting the number of colors obtained by the second principal component comparing means.

Further, the color discriminating apparatus of the present invention comprises a color image input means for inputting a color image, and an area designating means for designating an image area for obtaining the number of colors for the color image.
Cutting out means for cutting out the image area designated by the area designating means from the color image input by the color image inputting means, and mapping for mapping each pixel of the image area cut out by the cutting out means onto a color space. Means, a color reducing means for reducing the color of the image area mapped on the color space by the mapping means, and an image output means for outputting a color image of the image area reduced by the color reducing means. It is equipped with.

Further, the color discriminating apparatus of the present invention comprises a color image input means for inputting a color image, and an area designating means for designating an image area for obtaining the number of colors for the color image,
Cutting out means for cutting out the image area designated by the area designating means from the color image input by the color image inputting means, and mapping for mapping each pixel of the image area cut out by the cutting out means onto a color space. Means and each point on the color space of the pixel mapped by this mapping means is Hough (H
from the distribution of each point on the color space Hough-transformed by this conversion means, the maximum point detection means for detecting the maximum point of this distribution, and the maximum point detection means. A linear component detecting means for detecting the position of the linear point distribution on the color space from the maximum point, and a linear component dividing means for dividing the linear point distribution of the color space detected by the linear component detecting means. , Clustering means for clustering the point distribution on the color space including the linear point distribution of the color space divided by the linear component dividing means, and one color in each cluster clustered by this clustering means. Data replacing means for reducing the number of colors by replacing with a color, and image output means for outputting a color image of the image area reduced in color by the data replacing means It is provided.

[0015]

[Function] An image area for which the number of colors is desired to be set is set, and the point distribution on the color space obtained by mapping each pixel of the specified image area on the color space is subjected to Hough transform to obtain a linear shape of the color space. By detecting the position of the point distribution of, the number of colors in the designated image area can be obtained.

Further, by performing a principal component analysis on the point distribution of each pixel of the designated image area mapped onto the color space and using the first principal component and the second principal component thereof, The number of colors of can be calculated.

Further, by designating an image area to be reduced in color,
By Hough transforming the point distribution on the color space obtained by mapping each pixel of the specified image area on the color space,
It is possible to reduce the number of colors in which an image is displayed using the number of clusters obtained by detecting the position of a linear point distribution in the color space and clustering the linear point distribution.

[0018]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings. FIG. 2 shows the configuration of a color discriminating apparatus according to an embodiment of the present invention. In the figure, a CPU 201 which controls the entire control is connected to a storage device 20 which is a working memory of the CPU 201 via a CPU bus 202.
3. A main storage unit 204 in which a control program for controlling the color discrimination apparatus and the like are stored, and a data input device 205 such as a keyboard for inputting data or instructions from the outside are connected. The main storage unit 204 has a program storage area and a temporary storage area. A program is stored in the program storage area in advance, and the temporary storage area is used as a temporary storage area of the CPU 501.

The CPU 201 has a CPU bus 20.
An input interface device 206 and an output interface device 207 are connected via the interface 2. An image input device 208 is connected to the input interface device 206. When the image input device 208 reads a color image prepared in advance, this color image passes through the input interface device 206 and the CPU bus 202. Then, it is sent to the storage device 203 and stored there.

An image output device 209 is connected to the output interface device 207, and the color image stored in the storage device 203 passes through the CPU bus 202 and the output interface device 207 to the image output device 209. It is sent and output to any device.

Next, the flow of processing of this apparatus having the above-mentioned structure will be described with reference to the flow chart shown in FIG. When the image input request command is output from the CPU 201, the image input device 208 reads a color image prepared in advance and stores the color image in the storage device 203 (step 601).

The image area for which the number of colors is desired to be determined is input and specified from the data input device 205 to cut out the specified image area from the color image stored in the storage device 203 (step 602). Each pixel of the area is mapped on the color space (step 603). That is, assuming that the RGB three-dimensional color space is used as the color space and each pixel value of the color image is given by the red component, the green component and the blue component, this pixel is mapped to the RGB three-dimensional color space. The values may be mapped to corresponding positions on the RGB three-dimensional color space composed of the red (R) component, the green (G) component and the blue (B) component. When all the pixels in the designated image area are mapped on the color space by such a method, the pixels in the designated area are represented by a point distribution in the color space.

Next, in the color number discrimination process (step 604), the color number is obtained from the point distribution on the color space obtained in step 603, and this color number is output to an arbitrary device (step 605).

The color number discrimination process (step 604) is further divided into the processes shown in the flowchart of FIG. The process here is based on the point distribution in the color space shown in FIG.
This is to obtain the number of linear components (101 to 103) as shown in (a). The point distribution on the color space of the pixel obtained in step 603 is obtained by Hough transform.
It is mapped to another space (step 701). This space is called the Hough transform space here. The Hough transform is a method for detecting a figure such as a straight line from an image. Here, the most widely used straight line detection method will be described (Duda and Hart's method: Hideyuki Tamura, Computer Image Processing Introduction: Research Institute Publishing 1985. Excerpt).

A straight line is expressed by ρ = x cos θ + y sin θ (1), and as a parameter for describing the straight line,
(Ρ, θ) is used. Here, ρ is the length of the perpendicular line drawn from the origin, and θ is the angle between the perpendicular line and the x axis. If this straight line passes through the point (x ₀ , y ₀ ) on the image, the relationship of ρ = x ₀ cos θ + y ₀ sin θ (2) holds. This relationship is a sine curve on the parameter space ρ-θ. That is, one point in the xy space corresponds to one locus in the ρ-θ space, and conversely, the locus of ρ-θ expressed by the equation (2) is (x ₀ , y in the xy space.
It means that all straight lines passing through ₀ ) are represented. Therefore, when a point on a straight line in the xy space is mapped to the ρ-θ space, the locus in the ρ-θ space created from these points intersects at one point.

Here, a specific example of the Hough transform process will be described based on the point distribution in the color space as shown in FIG. First, the distribution in the RGB three-dimensional color space is RG, RB, G
Replace with any two-dimensional space of B. As a replacement method, for example, each point on the RGB three-dimensional color space is R
Map to the point seen from each surface of G, RB, GB, and R
The surface with the widest distribution among G, RB, and GB surfaces,
Alternatively, a surface in which all straight line components are clearly visible is extracted by visual inspection. For example, in order to replace the distribution in the RGB three-dimensional color space shown in FIG. 5 with the two-dimensional space, the data may be mapped onto the RB plane where the straight line component clearly appears. When mapped to the RG plane, white 301 to red 30
When a straight line extending to 3 does not appear and is mapped to the GB plane, white 3
This is because the straight line extending from 01 to blue 304 does not appear. Mapping the distribution of FIG. 5 on the RB plane results in the distribution as shown in FIG.

Next, each point mapped in the RB space of FIG. 1A is mapped in the ρ-θ space (Hough transform space) by the equation (1). In the following, description will be given with reference to FIG. 6 showing a state in which the distribution of FIG. 1A is approximated by a straight line by a process such as thinning. Each point in FIG. 6 (A, B, C, D, E, F, G)
FIG. 7 shows a state in which is mapped on the ρ-θ space (Hough transform space).

In the RB color space as shown in FIG. 6, three arbitrary points on the straight line 801 connecting the white distribution region and the red distribution region are A, B and C, and on the straight line 802 connecting the white distribution region and the black distribution region. , A, D, E, and arbitrary three points on a straight line 803 connecting the white distribution region and the blue distribution region are A,
F and G.

When the Hough transform is applied to the point A (x ₀ , y ₀ ),
In FIG. 6, assuming that the length of the perpendicular line drawn from the origin to the straight line 803 passing through the point A is ρ and its angle is θ, the combination of these two variables is shown in FIG. It is represented by a single sine curve (807) as shown. This is similarly performed for the points B to G, and the sine curve of each point is obtained on the ρ-θ space. If there is a point that is linearly distributed in the color space, the sine curves corresponding to that point intersect at one point in the ρ-θ space, so points A, B, and
Each sine curve of C is the point α in FIG. 7, the point A in FIG.
The sine curves of D and E intersect at a point β in FIG. 7, and the sine curves of points A, F, and G in FIG. 6 intersect at a point γ of FIG. That is, if a linear point distribution exists in the color space, ρ−θ
In the space, the intersections where the sine curves corresponding to the linearly distributed points intersect with each other are concentrated at one point.

Since the Hough transform space has been obtained as described above, the peak points where the intersections of the sine curves are concentrated are detected in the Hough transform space (step 702). Figure 6
Since the distribution has been approximated by a straight line in the above description, it is sufficient to simply detect the peak in the ρ-θ space.
There is some variation, not a perfect straight line. Therefore, for example, the following peak detection method is used. To detect peaks, just find the distribution peaks,
For example, a process like the flowchart shown in FIG. 8 can be considered.

First, the distribution state in the two-dimensional space is searched in the horizontal direction, it is determined whether the left and right distribution states are smaller than the distribution state at the current position, and processing for finding the horizontal peaks is performed (step 901). ). Next, when a horizontal mountain is found, a vertical search is performed to determine whether the vertical distribution state is smaller than the current position distribution state, and the vertical mountain is found (step 902). . In this way, when horizontal and vertical peaks are found, those points are output as peak points (step 903). Although it has been described here that there is one peak point, it does not matter how many peak points there are. In this case, all peak points are extracted and output.

Finally, in step 702, the number of peaks detected by the above processing is counted to determine the number of colors (step 703). The number counted here indicates the number of straight lines in the color space, and the number of colors to be obtained is, in the case of a color image in which colors exist on a uniform background,
This is the number obtained by adding "1" to the number of this straight line.

The determination of the number of colors in a specific area of a color image is 1
When it is sufficient to determine whether there are two colors, three colors or more, the color number determination process (step 604) can be performed by the process shown in the flowchart of FIG. First, in the principal component processing (step 1001), the principal component analysis of the distribution in the color space is performed. Principal component analysis is a transformation in which, when multivariate measurement values are given, the correlation between the variables is eliminated by linear transformation, and the characteristics of the measurement target are described with fewer variables.

As a concrete example, the distribution in the color space is shown in FIG.
As shown in FIG. 0, the direction component having the largest spread of the distribution 1601 is the first principal component 1602, the direction component orthogonal thereto is the second principal component 1603, and the direction component orthogonal thereto is the third principal component 1604. , ...
This is an analysis method that attempts to express a large amount of data with these several types of components.

Next, it is determined whether the first principal component is larger than the threshold value θ ₁ (step 1002). If the first principal component is smaller than the threshold θ ₁ , the number of colors in this image area is determined to be “1” (step 1003). If the first principal component is larger than the threshold θ ₁ , it is further determined whether the second principal component is larger than the threshold θ ₂ (step 1004). Here, when the second principal component is smaller than the threshold value θ ₂ , it is determined that the number of colors of this image region is “2” (step 1005). If the second principal component is larger than the threshold value θ ₂ , it is determined that the number of colors is “3” or more (step 1006).

FIG. 1 shows a specific example of color number discrimination by principal component analysis.
1, 12, and 13 are shown. FIG. 11 shows a distribution when the first principal component is small.
Shows the color distribution of. FIG. 12 shows a distribution in which the first principal component is large, but the second principal component is small, and this shows a color distribution of two colors in an image in which one color is present in addition to the background. FIG. 13 shows a distribution when both the first principal component and the second principal component are large, and this shows a color distribution of the number of colors “3” in an image in which two colors are present in addition to the background, for example.

Further, when the point distribution on the color space is linear, a specific example of the number of colors determination by the principal component analysis is shown in FIG. 1 (b).
Shown in. This figure is a diagram in which the distribution in the RGB three-dimensional space is replaced with a two-dimensional space of any of RG, RB, and GB as in FIG. 1A. First principal component axis 104 and second
The principal component axis 105 is as shown in the figure, and since the values of both the first principal component and the second principal component are large, the number of colors is determined to be “3” or more. Threshold theta _1, theta ₂ is, for example, when the resolution of the pixel values of the color image is that it 256 gradations, theta ₁ is "128", theta ₂ becomes about "20".

By thus determining the number of colors in the specific area of the color image, for example, a red alphabet 1101, a flower 1102, a person 1103, and a black character are displayed on a uniform background as shown in FIG. From the color image in which the date 1104 is written, each area is identified as a red single color 1105, a multicolor (A) 1106, a multicolor (B) 1107, and a black single color 1108, as shown in FIG. 14B. can do.

As described above, the number of colors in the specific area of the color image is obtained by detecting the position of the linear distribution on the color space or analyzing the components of the linear distribution. After the linear distribution in the color space is detected by the same processing as the discrimination method, the linear distribution is divided into two, whereby the distribution in the color space is clustered and the color reduction can be realized. An example thereof will be described with reference to the flowchart shown in FIG.

A color image is input from the image input device 208 and stored in the storage device 203 (step 1201). Next, after the image area to be reduced in color is input and specified from the data input device 205, the specified image area is cut out from the color image stored in the storage device 203 (step 1202), and then the specified area is specified. Each pixel of the image area is mapped on the color space (step 1203). Here, an RGB three-dimensional color space will be used as the color space for description. If each pixel value of a color image is given by a red component, a green component and a blue component, this pixel value is composed of a red component, a green component and a blue component in order to map it in an RGB three-dimensional color space. It suffices to map to a corresponding position in the RGB three-dimensional color space. By performing this operation for all the pixels in the designated image area, the color of the color image is reduced from the obtained color space distribution (step 1204). The color-reduced color image is output to the image output device 209 (step 1205).

Next, color reduction processing (step 1204)
Is further divided into the processes shown in the flowchart of FIG. First, color number discrimination processing (step 1301)
Then, the same processing as steps 701 to 703 in FIG. 4 described above is performed. Then, the position of the linear distribution in the color space is obtained from the peak points detected in the Hough transform space (step 1302). That is, for example, the peak point α (8
06) is a straight line 803 on the color space of FIG.
Is required as.

Next, the linear distribution obtained in step 1302 is divided into two (step 1303). More specifically, for example, the linear distribution 101 in the color space in FIG. 1A is actually the distribution curve 140 in FIG.
Suppose it was 1. To divide this distribution into two,
For example, there is a modal method. The modal method is shown in FIG.
If there is a data distribution as shown in (b), two mountains 1
It is a method of bisecting at a valley 1403 between 404 and 1405.

After dividing each linear distribution in the color space into two, clustering of the distribution in the color space is performed (step 1304). As an example of the color space distribution, when the distribution shown in FIG. 5 is clustered by detecting the position of the linear distribution on the color space or by analyzing the distribution on the straight line,
As shown in FIG. White 1501 and red 1502
It can be seen that the linear distribution connecting the distributions of 1 and 2 is divided at the division point 1505 in step 1303, and white and red are separated. Similarly, white 1501 and black 1503
Is divided at a division point 1506, and white 1501 and blue 1504 are divided at a division point 1507. Finally, as shown in FIG. 18, four areas (1501, 1
502, 1503, 1504).

Then, the pixel value in each cluster clustered in step 1304 is replaced with, for example, the average color in the cluster, whereby the color reduction can be realized (step 1305). That is, when the pixel values in each cluster clustered as shown in FIG. 18 are replaced with the average color in the cluster to reduce the number of colors,
Reduced to 4 colors.

As described above, the point distribution of each pixel in the designated image area mapped on the color space is Hough transformed, or
By performing the principal component analysis, the number of colors in the designated image area can be obtained and clustering can be performed. Furthermore, by knowing the number of colors, the area division of the color image,
It can also be used for area identification. In addition, when the number of display colors on the display is limited when displaying a color image due to clustering, the number of colors of the obtained cluster is used to reduce the number of colors used to display the image, and the image is divided into regions. It is possible to know the number of divided areas when performing the above-mentioned step, and it can be used for data compression as the minimum required number of colors when compressing color image data.

In the above embodiment, the RGB three-dimensional space is replaced with the RB two-dimensional space, but it is also possible to carry out Hough transformation in the RGB three-dimensional space and extract straight lines. Also, not only the RGB color space but also the HV
It is also possible to analyze using a C color space, a CMY color space, an L ^* a ^* b ^* color space, and the like.

[0047]

As described above in detail, according to the present invention, when it is difficult to determine the number of colors such that the data distribution is distributed over several clusters, a linear distribution on the color space is obtained. It is possible to provide a color discriminating apparatus capable of discriminating the number of colors in a specific region of a color image and reducing the number of colors by detecting the position of or the principal component analysis of the linear distribution.

[Brief description of drawings]

FIG. 1 is a diagram illustrating a color number determination method of a color determination device according to an embodiment of the present invention.

FIG. 2 is a block diagram schematically showing a configuration of a color number determination device according to an embodiment of the present invention.

FIG. 3 is a flowchart illustrating a color number determination method.

FIG. 4 is a flowchart illustrating a color number determination process by detecting a linear distribution in a color space.

FIG. 5 is a diagram showing an example of a color distribution when clustering on a color space is difficult by a conventional method.

FIG. 6 is a diagram illustrating a detection process of a linear distribution in a color space by Hough transform.

FIG. 7 is a diagram illustrating a detection process of a linear distribution in a color space by Hough transform.

FIG. 8 is a flowchart illustrating peak point detection processing.

FIG. 9 is a flowchart illustrating a color number determination process based on principal component analysis.

FIG. 10 is a diagram illustrating principal component analysis.

FIG. 11 is a diagram showing an example of a distribution in which the number of colors is discriminated (the number of colors is 1) by the color number discrimination processing by the principal component analysis.

FIG. 12 is a diagram showing an example of a distribution in which the number of colors is discriminated (the number of colors is 2) by the color number discrimination processing by the principal component analysis.

FIG. 13 is a diagram showing an example of a distribution in which the number of colors is discriminated (the number of colors is 3) by the number-of-colors discrimination processing by the principal component analysis.

FIG. 14 is a diagram showing an example of area identification of a color image.

FIG. 15 is a flowchart illustrating color reduction by clustering.

16 is a flowchart for explaining the color reduction processing of FIG.

FIG. 17 is a diagram illustrating a method of dividing a linear distribution.

FIG. 18 is a diagram showing an example of clustering a linear point distribution.

FIG. 19 is a diagram showing an example of color distribution clustered in a color space.

FIG. 20 is a diagram showing an example of color reduction.

FIG. 21 is a diagram illustrating a cause of connection of color distributions.

[Explanation of symbols]

101 to 103 ... Linear point distribution, 104 ... First principal component axis, 105 ... Second principal component axis, 201 ... CPU, 202 ...
CPU bus, 203 ... Storage device, 204 ... Main storage device,
205 ... Data input device, 206 ... Input interface device, 207 ... Output interface device, 208 ... Image input device, 209 ... Image output device.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Office reference number FI Technical display location H04N 1/60 1/46 4226-5C H04N 1/40 D 4226-5C 1/46 Z

Claims (5)

[Claims] 1. A color discriminating apparatus for obtaining the number of colors in a specific area of a color image, and a color image input means for inputting the color image, and an area designation for specifying an image area for obtaining the number of colors for the color image. Means, cropping means for cropping the image area designated by the area designating means from the color image input by the color image input means, and each pixel of the image area sliced by the cropping means on a color space. Mapping means for mapping, color number discriminating means for discriminating the color number from the point distribution on the color space of the pixels imaged by the mapping means, and color number for outputting the color number discriminated by the color number discriminating means. A color discrimination device comprising: an output unit. 2. A color discriminating apparatus for obtaining the number of colors in a specific area of a color image, and a color image input means for inputting the color image, and an area designation for specifying an image area for obtaining the number of colors for the color image. Means, cropping means for cropping the image area designated by the area designating means from the color image input by the color image input means, and each pixel of the image area sliced by the cropping means on a color space. From the mapping means for mapping, the transforming means for Hough transforming each point in the color space of the pixel mapped by this mapping means, and the distribution of each point in the color space Hough transformed by this transforming means A color number discriminating means for discriminating the color number by detecting the maximum point of this distribution, and a color number outputting means for outputting the color number discriminated by the color number discriminating means. Color discrimination and wherein the. 3. A color discriminating apparatus for obtaining the number of colors in a specific area of a color image, and a color image input means for inputting the color image, and an area designation for specifying an image area for obtaining the number of colors for the color image. Means, clipping means for clipping the image area designated by the area designating means from the color image input by the color image input means, and each pixel of the image area clipped by the clipping means in the color space. Mapping means for mapping, principal component analysis means for principal component analysis of the point distribution in the color space of the pixels mapped by this mapping means, and the value of the first principal component obtained by this principal component analysis means is set in advance. A first principal component comparing unit that obtains the number of colors by comparing it with a predetermined value, and a second preset principal component value obtained by the principal component analyzing unit. A second principal component comparing means for obtaining the number of colors by comparing; and a first and second principal component comparing means, and a color number outputting means for outputting the number of colors obtained by the second principal component comparing means. A color discrimination device characterized by. 4. A color discriminating apparatus for obtaining the number of colors in a specific area of a color image, and a color image input means for inputting the color image, and an area designation for specifying an image area for obtaining the number of colors for the color image. Means, cropping means for cropping the image area designated by the area designating means from the color image input by the color image input means, and each pixel of the image area sliced by the cropping means on a color space. Mapping means for mapping, decolorizing means for decolorizing the image area mapped on the color space by the mapping means, and image for outputting a color image of the image area decolorized by the decolorizing means A color discrimination device comprising: an output unit. 5. A color discriminating apparatus for obtaining the number of colors in a specific area of a color image, and a color image input means for inputting the color image, and an area designation for specifying an image area for obtaining the number of colors for the color image. Means, cropping means for cropping the image area designated by the area designating means from the color image input by the color image input means, and each pixel of the image area sliced by the cropping means on a color space. From the mapping means for mapping, the transforming means for Hough transforming each point in the color space of the pixel mapped by this mapping means, and the distribution of each point in the color space Hough transformed by this transforming means , A maximum point detecting means for detecting the maximum point of this distribution, and a straight line forming means for detecting the position of the linear point distribution on the color space from the maximum point detected by the maximum point detecting means. Detecting means, linear component dividing means for dividing the linear point distribution of the color space detected by the linear component detecting means, and linear point distribution of the color space divided by the linear component dividing means Clustering means for clustering point distributions in the color space, data replacement means for reducing the color by replacing the colors in each cluster clustered by this clustering means with one color, and this data replacement means for reducing the number of colors An image output unit for outputting a color image of a converted image region, and a color discrimination device comprising: