JP3306964B2 - Limited color determining method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method for generating a natural image using a limited number of representative colors when a digitized natural image is displayed on a display device in a limited number of colors. The present invention relates to a method and an apparatus for determining a limited color for display.

[0002]

2. Description of the Related Art Generally, a color image used in a color display device is represented by RGB color components. Each color component is digitized by allocating an 8-bit memory to each color component. Thus, the full color display for displaying a color image which is the digitized, color 24 bits for each pixel, that is, simultaneously capable of displaying 2 ²⁴ colors can be displayed natural images seamlessly is there.

On the other hand, CD-ROM, CD-I, CAD
(Computer Aided Design), CAI (Computer Assist
ed Instruction), a personal computer, an X-Window System, etc., in a world where images are handled, a color display device that can simultaneously display only a limited number of the above-mentioned 224 colors is often used. Since the capacity per pixel of the memory is usually set to 8 bits or less,
The number of colors that can be displayed simultaneously is up to 2 ⁸ = 256. With a limited number of colors such as 256 colors or less, it is difficult to display a natural image without deterioration, but there are many demands to make this possible.

[0004] Further, from the relation of the data capacity of the display device and the like, there are many demands to reduce the number of colors of the natural image so that the natural image is not deteriorated, thereby reducing the data amount of the image data. For such a purpose, a method of displaying a natural image with a limited number of colors while suppressing visual deterioration has been conventionally used for a color map display that selects and displays a small number of display colors from a large number of colors. It has been known.

As the above method, for example, the colors R, G,
When represented by the values of the three primary colors of B, a space represented by a three-dimensional orthogonal coordinate system having coordinates of the values of R, G, and B as a color space, and included in the color image data in this color space. The color distribution of each pixel is analyzed. Then, display colors are selected such that the display colors are large from the color regions in the color space where many colors included in the data are distributed and are small from the color regions with a small distribution, and these display colors are registered in the color map. . Then, a display color closest to the color of each pixel of the color image data is assigned to each pixel as a representative color, and a color code registered as a representative color in the color map is written at the position of each pixel in the image memory to perform approximate display. There is something to do.

As one of the methods for selecting a representative color as described above, Otsu's discriminant analysis (IEICE (D). J63-D, 4, pp. 349-3)
56. Showa 55-4.). Further, the color space is not limited to one that expresses a color by a combination of three primary colors of RGB, but also L ^* u ^* v ^* and L ^{* in} consideration of human color difference ^.
a ^* b ^* uniform color space (recommended by the CIE (International Commission on Illumination)) and the like, and the discriminant analysis method can be performed using these color spaces. In addition, a method using the visual characteristics of humans and further applying a dithering method (Journal of the Institute of Image Electronics Engineers of Japan, Vol. 18, No. 5, (1989) p. 293-301; Tajima, Ikeda) and the like have been proposed.

[0007]

Since the above-described discriminant analysis method simply determines a display color based on the goodness of the color distribution, the discriminant analysis method is created by determining a display color by the discriminant analysis method in the RGB color space. The image did not become a natural image for the human eye without discomfort. For example, when a small number of representative colors are assigned to a portion where the color change of the image data is gradual, false contours may occur. There is a problem in that when the change in the gradation is represented by a small number of representative colors, false contours are noticeable.

The representative color selected by the above-described discriminant analysis is determined by the distribution of each pixel in a state where the color of each pixel of the image data is arranged in a color space. However, it is difficult to reproduce a natural image without discomfort.

For example, in image data of a landscape photograph,
When the sky and the sea are imaged, the sky and the sea each contain the largest number of pixels having a bluish color, and each pixel of this image data is arranged in a color space based on that color. In this case, the pixels of the sky and the pixels of the sea are arranged at positions close to each other in the color space, and form one cluster. As a result, there is a possibility that substantially similar representative colors may be assigned to the sky and the sea, and in that case, there is a problem that the color difference between the sky and the sea disappears.

[0010] This problem is not limited to a landscape photograph of the sky and the sea, but an attempt is made to approximate and display image data in which different objects having similar colors are drawn using a small number of representative colors. This is a problem that arises in some cases.

In addition, as a conventional method, a method in which the state of an original image, for example, a color distribution tendency is extremely different from the original image state, is not taken into consideration, such as the method based on the above-described discriminant analysis method. There are a non-adaptive method of creating an image by selecting a limited display color, and an adaptive method of selecting a limited color so as to be suitable for statistical properties such as the color distribution of the original image.

In the above non-adaptive method, unlike the case of using the above-described discriminant analysis method, the image is not always approximated to the original image when viewed from the human eyes, and there is a problem in image quality. In addition, in the adaptive method, it is necessary to check the color distribution and the like of each image data, and to change the statistical processing method according to the state of the color distribution. There are problems that require human judgment and that the processing time is long.

The non-adaptive method that can maintain a certain level of image quality for general image data does not particularly select image data and does not require human judgment. Although natural images serving as image data have more advanced statistical characteristics such as color distribution depending on their contents, the superiority of the applied method has not changed in terms of image quality.

In particular, in order to obtain a small number of limited colors of 50 colors or less, it is essential to consider a human viewpoint, and the non-adaptive method has not been successful yet. An adaptive method that can input and judge a person's judgment on the way is still superior.

The present invention has been made in view of the above problems, and focuses on the fact that human vision is particularly sensitive to a change in luminance. A limited color determination method that can solve the problems of the non-adaptive method and the adaptive method from the viewpoint that the color cohesion as an image, that is, the continuity of colors on the image is important, and It is intended to provide a device.

[0016]

According to a first aspect of the present invention, there is provided a limited color determining method for converting color image data represented by a coordinate system of three primary colors into a coordinate system of a uniform color space, and converting the converted coordinates. A representative color is obtained by performing a discriminant analysis on a frequency distribution of each pixel of the image data in the system, and the image data is displayed approximately by a limited representative color. From the distribution of the color components of each pixel arranged in the image array, the goodness of the unity of the color components of each pixel on the image array is quantified, and the color of each pixel on the quantified image array The multi-dimensional space is provided by associating the unity of the components with each color component of each pixel and newly setting a coordinate axis indicating the unity of the color components in the color space, in the coordinate system of the multi-dimensional space. The above image data Obtains a frequency distribution of the pixel, the determination of the representative colors by performing discriminant analysis on the frequency distribution was means for solving the above-described problem.

The limited color determining method according to the second aspect of the present invention converts color image data expressed in a coordinate system of three primary colors into a coordinate system in a uniform color space, and converts the image data in the converted coordinate system into a uniform color space. The representative color is obtained by performing a discriminant analysis on the frequency distribution of each pixel of the data, and is used for approximating the image data with a limited representative color, and is arranged on the image of the image data. From the distribution of the color components of each pixel on the image array, the image is divided into regions that can be divided based on the color change between pixels, and it is checked which region of the image each pixel belongs to. The goodness of the color components of each pixel therein is quantified, and the goodness of the quantified color components is associated with each color component of each pixel, and the goodness of the color components is newly collected in the color space. Set the coordinate axis indicating Means for obtaining a frequency distribution of each pixel of the image data in a coordinate system of the multidimensional space, and obtaining a representative color by performing a discriminant analysis on the frequency distribution. And

According to a third aspect of the present invention, there is provided a limited color determining apparatus for converting color image data represented by a coordinate system of three primary colors into a coordinate system of a uniform color space, and converting the image data in the converted coordinate system to a uniform color space. The representative color is obtained by performing a discriminant analysis on the frequency distribution of each pixel of the data, and is used for approximating the image data with a limited representative color, and is arranged on the image of the image data. Numerical means for quantifying the goodness of the color components of each pixel from the distribution of the color components on the image array of each pixel, and the goodness of the color components quantified by the numerical means for each pixel. A four-dimensional space is provided by newly setting a coordinate axis indicating the goodness of the color components in the uniform color space in association with the color components,
A frequency distribution calculating means for obtaining a frequency distribution of each pixel of the image data in the coordinate system of the four-dimensional space; and a frequency distribution of each pixel of the image data in the four-dimensional space obtained by the frequency distribution calculating means. And a discriminant analysis operation means for obtaining a representative color by performing discriminant analysis by using the discriminant analysis.

The limited color determining apparatus according to a fourth aspect of the present invention converts the color image data represented by the coordinate system of the three primary colors into a coordinate system of a uniform color space, and converts the image data in the converted coordinate system. The representative color is obtained by performing a discriminant analysis on the frequency distribution of each pixel of the data, and is used for approximating the image data with a limited representative color, and is arranged on the image of the image data. An image dividing unit that divides the image into regions that can be divided based on a color change between the pixels based on a distribution of color components on an image arrangement position of each pixel, and each pixel of the image data is transmitted to the image dividing unit. Pixel belonging area calculating means for determining which of the divided areas belongs to, and numerical value means for quantifying the goodness of the color components of each pixel from the distribution of the color components of each pixel in the area. , The digitizing means A four-dimensional space is provided by associating the united goodness of the color components more digitized with each color component of each pixel and newly setting a coordinate axis indicating the goodness of the unity of the color components in the uniform color space, A frequency distribution calculating means for obtaining a frequency distribution of each pixel of the image data in the coordinate system of the four-dimensional space; and a frequency distribution of each pixel of the image data in the four-dimensional space obtained by the frequency distribution calculating means. And a discriminant analysis operation means for obtaining a representative color by performing discriminant analysis by using the discriminant analysis.

[0020]

According to the structure of the first and third aspects, for example, the uniform color space represented by three coordinate axes of L ^* , u ^* , v ^* or L ^* , a ^* , b ^* is used for image arrangement. A new coordinate axis indicating the cohesion of the color components of each pixel is newly added to provide a four-dimensional space. Here, the goodness of the unity of the color components of each pixel on the image array will be described. In general, objects drawn on an image are represented by relatively similar colors. For example, the sea and sky contain many blue-based colors, and apples contain many red-based colors. Then, in a portion where the sea and the sky are drawn on the image array, pixels having a bluish color are present collectively, and this is expressed as goodness of unity.

The goodness of the unity of the color components of each pixel on the image array is determined by, for example, how many pixels of similar colors are present together, and how many pixels of similar colors are within an arbitrary range. It can be converted into a numerical value by measuring whether or not it exists.

Then, each pixel of the image data is developed in the four-dimensional space, a frequency distribution of the pixels in the four-dimensional space is obtained, and a representative color is obtained from the obtained frequency distribution by a discriminant analysis method. Conventionally, discriminant analysis can be performed in a state in which a part of the information on the relationship between the position on the image array and the color of each pixel, which is lost when the discriminant analysis is performed on the uniform color space, is conventionally taken in. .

That is, in the uniform color space, pixels having colors similar to each other are arranged at positions close to each other to form a set of one pixel = cluster, but in the four-dimensional space, Pixels having colors similar to each other are not always arranged at close positions, and if the unity of the numerically represented color components is different, they belong to different clusters. Such a color may belong to a divided area based on different threshold values for discriminant analysis.

Therefore, in an image in which two or more subjects having similar colors such as the sea and the sky are drawn, when the image is displayed with only a few representative colors, the sea and the sky are different. Likely to be painted in color.

In the image data, pixels included in one subject portion are likely to have similar colors and exhibit similar coherence, and are included in the portion representing the subject. The pixels are densely packed in the four-dimensional space, and when selecting a relatively large number of representative colors, many colors are selected from the densely packed pixels by the discriminant analysis method. That is, many colors are arranged in a portion where similar colors are present collectively on the image (a portion where the color changes gradually), and the occurrence of a false contour can be prevented.

According to the second and fourth aspects of the present invention, in order to quantify the goodness of the color components of the respective pixels on the image array, the respective pixels previously arranged on the image of the image data are used. From the distribution of the color components on the image array,
The image is divided into regions that can be divided based on a color change between pixels.

For example, in a subject drawn in an image, the color components of pixels adjacent to each other on the image array greatly change in the outline, and the image can be divided by this change. Further, an image can be divided by obtaining a spatial frequency distribution using a frequency transform such as a discrete cosine transform used in JPEG or the like.

As described above, the image is divided into regions in advance.
By calculating the ratio of pixels of similar colors in the divided areas, it is possible to easily digitize the goodness of unity of color components. As in the first and third aspects of the present invention, a part of the information on the relationship between the positions of the pixels on the image and the colors can be taken into the discriminant analysis using the uniform color space.

[0029]

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a schematic diagram of a limited color determining apparatus using the limited color determining method of the present invention. FIG. 2 is a flowchart showing an algorithm of the limited color determining method of the present invention. 3 to 9 are views for explaining the limited color determining method of the present invention.

A limited color determining apparatus for performing image processing using the limited color determination according to the present embodiment will be described below with reference to FIG. The limited color determination device includes an image storage unit 1 for storing image data, an upper 4-bit image storage unit 2 for storing upper 4 bits of the image data stored in the image storage unit 1, and an upper 4-bit image storage unit. Top 4 stored in the means
A mosaic image calculating means 3 for dividing the bit image data into partial cells of a size determined according to the image size, and a mosaic image calculating means 3 for dividing each mosaic image into partial cells into cells. A mosaic cell color determining means 4 for finding a color component which is superior to the mosaic cell, a mosaic cell color storing means 5 for storing the mosaic cell color calculated by the mosaic cell color determining means 4, A continuity value calculating means 6 for determining the continuity of appearance of the same mosaic cell color in each of the four directions of vertical, horizontal, diagonally left and diagonally right, and quantifying the continuity as a continuity value; Continuity value storage means 7 for storing the continuity value of each color component digitized by the continuity value calculation means 6
And

Further, the above-described limited color determining apparatus converts the RGB data of the color image into L according to the following equations (1) to (7).
color conversion means 8 for performing color conversion for each pixel into ^* a ^* b ^* data;
Each pixel is developed based on the L ^* a ^* b ^* and the continuity value in a four-dimensional space spanned by L ^* , a ^* , b ^* and the continuity value of the color component, and each pixel in the four-dimensional space is expanded. Set of =
Cluster storage means 9 for storing clusters; and principal axis calculation means for calculating a coordinate axis at which variance is maximum among coordinate axes indicating continuity values of L ^* , a ^* , b ^* and color components in each of the clusters = main axis. 10, a spindle storage unit 11 for storing the spindle obtained by the spindle calculation unit 10, and a frequency distribution of the pixels included in the cluster for each cluster stored in the cluster storage unit 9 in each of the four-dimensional spaces. The apparatus includes a histogram calculating means 12 for each coordinate axis and a histogram storing means 13 for storing the histogram calculated by the histogram calculating means 12.

The limited color determination device includes an average / variance calculating means 14 for calculating an average and a variance from the histogram stored in the histogram storage means 13, and an average / variance calculated by the average / variance calculating means 14. , A weighted variance calculating means 16 for obtaining a weighted variance based on the variance stored in the mean / variance storage means 15, and a histogram stored in the main axis stored in the main axis storage means 11. A dividing threshold value determining means 17 for obtaining a dividing threshold value by discriminant analysis based on the histogram stored in the storing means 13; a dividing threshold value storing means 18 for storing the dividing threshold value obtained by the dividing threshold value determining means 17; The clusters are sorted in the descending order of the variance by the division threshold on the main axis stored in the division threshold storage means 18 for two minutes. Cluster dividing means 19, a cluster sorting means 20 for sorting the clusters divided by the cluster dividing means 19 according to their weighted variance values, and the operations of the means 2 to 20 are obtained by the cluster sorting means 20. A cluster division control unit 21 is provided which repeats these processes until there are no more clusters to be divided into clusters having the largest weighted variance value or the number of clusters reaches a specified number of colors.

Further, the above-mentioned limited color determining device is provided with L ^* ,
a ^* , b ^* , and emphasis means 22 for changing the rank of each coordinate axis in accordance with the variance ratio of each coordinate axis in the four-dimensional space spanned by the continuity values of the color components, and the emphasis means 22 The above means 2 to 2
An emphasis coefficient calculating means 23 for normalizing the value of the frequency distribution obtained by performing the operation 21; an emphasis coefficient storing means 24 for storing the emphasis coefficient obtained by the emphasis coefficient calculating means 23 for each pixel; A histogram updating means for multiplying the histogram stored in the histogram storage means 13 by the normalized enhancement coefficient read from the coefficient storage means 24, adding the result to the histogram, and rewriting the obtained new histogram with the histogram. 25
And

The limited color determination device reads the limited color obtained by performing the operations of the above means 2 to 25 from the averaging / dispersion storage means 15 and performs the inverse conversion of the color conversion means 8 by the color conversion means 26. A representative color allocating unit 27 that replaces the color of each pixel with the limited color using the cluster number of the limited color from the cluster storage unit 9; and a limited color pallet calculating unit 28 that sorts the limited colors by distance in a color space. And a color correction unit 29 that enables partial color correction on a screen in which a color image is approximately displayed with limited colors obtained by performing the operations of the above units 2 to 28.

The image storage means 1 has an image memory 1a and stores (R, G,
B) A value (each color component is 8 bits) is stored. The upper 4-bit image storage unit 2 stores the upper 4 bits of each pixel (RGB) value of the image data stored in the image storage unit 1.

The mosaic image calculation means 3 is the image storage means 1
1 for each of the vertical and horizontal directions corresponding to the size of the image stored in
An arbitrary size of about / 30 is calculated, and the image data stored in the high-order 4-bit image storage means 2 is divided by a rectangular area of the obtained size = cell. Mosaic cell color determination means 4
Is a color component present in the cell for each cell of the image data divided into cells by the mosaic image calculation unit 3 (the value of the upper 4 bits stored in the upper 4 bit image storage unit 2). Is determined as a mosaic cell color.

If one or more color components exist in the cell, the color component is determined to be a mosaic cell color.
If there is no color component having more than / 4, the average value of the color components in the mosaic cell is set as the mosaic cell color. The mosaic cell color storage unit 5 stores the mosaic cell color determined by the mosaic cell color determination unit 4 as mosaic image data.

The continuity value calculating means 6 is the same as the above-mentioned cells in the four directions of vertical, horizontal, diagonal left and diagonal right in each cell on the mosaic image data stored in the mosaic cell color storage means 5. The number of times the cell color continues continuously is determined as a continuity value. The continuity value storage means 7 assigns the continuity value obtained for each cell to the pixels included in each cell, and stores the continuity value in the same array as the pixels on the image data.

The color conversion means 8 converts the (R, G, B) values of each pixel of the image data into (L ^* , a ^* , b ^* ) values using a conversion formula described later. The cluster storage means 9 stores L ^* ,
Each pixel is developed in a four-dimensional space spanned by a ^* and b ^* and the continuity values of the color components, and a set of pixels in the four-dimensional space = cluster is stored. Here, the cluster indicates a divided part of the four-dimensional space when the four-dimensional space is divided using discriminant analysis as described later, and a cluster number is assigned to each divided part. For each cluster number, the position on the coordinates of the divided portion of the four-dimensional space and the pixels included in the cluster are stored.

The main axis calculating means 10 calculates, for each cluster, L ^* , a ^* , b ^* and, among the coordinate axes indicating the continuity values of the color components, the coordinate axis at which the variance of the pixel distribution with respect to each coordinate axis is equal to the main axis = main axis Is calculated. The spindle storage means 11
The spindle calculated by the spindle calculating means 10 is stored.

The histogram calculating means 12 obtains a frequency distribution of each pixel belonging to each cluster based on a class number to be described later and creates a histogram.
To memorize. The average / variance calculating means 14 calculates the average and variance of the pixel distribution for each coordinate axis in each cluster, and stores the average and variance in the average / variance storage means 15.

The weighted variance calculating means 16 obtains a weighted variance by multiplying the variance by the relative frequency of the number of pixels included in each cluster with respect to the total number of pixels. The division threshold value determining means 17 performs Otsu's discriminant analysis (IEICE (D) .J63-D, 4, p
p.349-356. Showa 55-4.) to determine the division threshold as described later. The division threshold storage unit 18 stores the division threshold calculated by the division threshold determination unit 17.

The cluster dividing means 19 reads the main axis of the cluster having the largest weighted variance obtained by sorting each cluster by the cluster sorting means 20 described later from the main axis storing means 11 and stores the cluster in the division threshold storing means 18. Is divided into two by the main axis division threshold value read from.

The cluster sorting means 20 sorts each cluster according to the weighted variance value of each cluster obtained by the weighted variance calculating means 16. The cluster division control unit 21 controls the repetition of these processes until the total number of clusters divided by the cluster division unit 19 reaches the specified number of colors or there are no more clusters to be divided. .

The emphasizing means 22 calculates the ratio of the variances obtained by the average / variance calculating means 14 for each coordinate axis of L ^* , a ^* , b ^* and the continuity value, and calculates the class number of each axis by the ratio. 1/8 times, 1/4 times,
1 / 2x, 1x, 2x, 4x, 8x, 16x, 32x,
64 times, and the frequency distribution of the pixels on each axis is determined by the above rank. The emphasis coefficient calculation means 23 normalizes the histogram value of the frequency distribution obtained by the emphasis means 22, and uses the normalized value as an emphasis coefficient.

The emphasis coefficient storing means 24 divides the emphasis coefficient obtained by the emphasis coefficient calculating means 23 by the number of pixels included in each class, and uses the result as the coefficient of each pixel included in the class, It is stored for each pixel. The histogram updating means 25 associates the enhancement coefficient read from the enhancement coefficient storage means 24 with the class number of the histogram read from the histogram storage means 13, multiplies the histogram value, and adds the result to the histogram value. The result is stored in the histogram storage unit 13 as a new histogram value.

The color conversion means 26 performs the inverse conversion of the color conversion means 8 and converts the value of (L ^* , a ^* , b ^* ) of the obtained representative color into the value of (R, G, B). . The representative color assigning means 27
After the above division is completed, the average color of each cluster is read from the average / dispersion storage unit 15 as a limited color, and the average color is adopted as a representative color of each cluster, and is included in each cluster in the original image data. The average color of the cluster is assigned to the pixel to be replaced, and the pixel is replaced with a value converted by the color conversion unit 26.

That is, the representative color allocating means 27 reads out the cluster number and the pixels included in the cluster of each cluster number from the cluster storage means 9 and calculates the average color of the cluster of each cluster number in the cluster included in the cluster of the cluster number. The color of each pixel is replaced with the value converted by the color conversion means 26. The limited color color palette calculation means 28
The obtained representative colors are sorted on a uniform color space (L ^* , a ^* , b ^* ) for distance.

The color correcting means 29 corrects the display image by enclosing the color of the display image with a rectangle in accordance with an instruction from the mouse, and focusing on the portion of the color to be changed in the rectangle and clicking the mouse. The place and the color to be chosen are to be selected. When the selected color is corrected and changed by the color correcting means 29, there are two methods. One is to change the color selected as described above to another color registered in the color map 30d. It is exchanged for the representative color. That is, after the color to be changed is selected, the color correcting means 29 displays the representative color registered in the color map 30d as a limited color palette screen.

On the displayed limited color color palette screen, the limited color palette calculation means 28 sorts the representative colors according to the distance in the uniform color space, that is, the color difference, and displays the colors having similar color differences adjacently. The representative color to be changed, which is selected as described above, is displayed in a blinking manner, and a color close to the representative color to be changed can be selected.

Then, for example, a representative color similar to the representative color to be changed is searched from the color palette screen, and the representative color is designated on the color palette screen, thereby replacing the representative color to be changed with the designated representative color. It is supposed to. That is, the representative color at the location of the display image designated by the mouse can be replaced with another representative color. In this case, only the color of the place specified by the mouse on the display image first is replaced with one of the representative colors that have been determined.

Another method of modifying and changing the color is to change the representative color registered in the color map 30d.
After selecting the representative color to be changed as described above, the color pallet screen is displayed, and the selected color blinks. A 24-bit full color color palette is displayed together with the limited color palette screen. In the case of this correction method, the color selected on the representative color palette is replaced with the color designated by the mouse on the full color palette.

That is, by performing this operation, for the selected representative color, the data itself of the color map 30d is changed, and on the display screen, the data other than the position designated by the mouse is selected. All the portions using the same representative color as the representative color are changed in color.

The limited color determination device having the above-described configuration displays full-color image data in a limited display color on the color map display device 30 by using the limited color determination method. In this embodiment, the color map display device 30 can also display the image data stored in the image storage unit 1 in full color, and compare the full color image with the limited color image. Thus, the color can be corrected by the color correcting means 29. Note that reference numeral 30c denotes the image memory 1a,
It is a D / A converter for displaying the image data of 30a on the display 30b.

Next, a limited color determining method using the above limited color determining apparatus will be described with reference to the flowchart of FIG. 2 and FIGS. 3 to 9. (A) Measurement of the degree of color change of image data to be displayed In the conventional method, the degree of color change of image data,
When measuring whether the color of each pixel in the image and the color of the pixel in the vicinity of the pixel are significantly different or substantially the same, the image represented by each of RGB 8 bits is used. RGB data is converted to a uniform color space such as L ^* a ^* b ^* or the like, and measurement is performed using the magnitude of the color difference.

However, the degree of color change of image data is
Since the color changes continuously on the image, not on the color space, it is necessary to consider the continuity on the image. Continuity in a color space does not necessarily indicate continuity in an image. Therefore, in the present embodiment, the color continuity of each color component on an image is checked, and the color continuity is reflected on a color space, thereby specifying a gradual change in color in image data to be displayed. In addition, in the color continuity on the image, when similar colors are continuously present on the image, it indicates that pixels of similar colors are collectively present on the image, and Is an index indicating the goodness of unity of the color components.

The color continuity on an image is calculated by calculating the color difference between each pixel and its neighboring pixels. The smaller the color difference, the higher the continuity, and the larger the color difference, the less the continuity. In the embodiment, the continuity is measured as follows.
First, by storing the value of the upper 4 bits of each pixel of the image data to be displayed in the limited color stored in the image storage unit 1 in the upper 4 bit image storage unit, the image data represented by the upper 4 bits is stored. Is created (step S1).

Next, the mosaic image calculation means 6
The image data is divided on a mesh by cells having a size of 1/20 to 1/40 of the image size to create a mosaic cell image (step S2). Then, as described above, the value of the color component occupying 1/4 or more in the cell (the average value when there are a plurality of 1/4 or more color components) by the mosaic cell color determination means 4 as described above
Is the mosaic cell color.

The method of selecting the mosaic cell color is not limited to the above method. For example, for each cell, the values of the upper 4 bits of the color components (RGB values) included in the cell are compared, and the principal component is determined. Principal components and principal component vectors are obtained by analysis. When the contribution ratio of the first principal component is 80% or more, the first principal component vector is obtained. When the contribution ratio is less than 80%, the first and second principal component vectors are obtained. And if the contribution ratios of the first and second principal components are less than 80%,
A composite vector of the first, second, and third principal component vectors may be obtained and adopted as the mosaic cell color of the cell (step S3). Then, a mosaic image in which the mosaic cell colors are assigned to the cells as described above is created, and stored in the mosaic cell color storage unit 5.

The continuity value calculating means 6 calculates the representative value of each cell thus obtained as shown in FIG.
Four (8) radiations of vertical (upper / lower), horizontal (left / right), diagonal left (diagonally upper left / lower left lower), diagonal right (diagonally upper right / lower right lower) The continuity is examined by comparing the representative values of the cells arranged continuously in the direction and checking whether the same value (mosaic cell color of the upper 4 bits) appears continuously.

As shown in FIG. 4, a specific method of calculating the continuity is to first compare the representative value of the cell of interest 41a with the representative value of the cells 41 while sequentially moving in the above-mentioned direction. Is counted. When the same value no longer appears, the counting in that direction ends. The above operation is performed in the above eight directions, and the total count of the number of appearances of the same value is defined as the continuity value of the cell of interest 41a. The above operation is performed with all the cells 41 of the image data as the cell of interest 41a, and the continuity values of all the cells are measured. In FIG. 4, the continuity value of each of the two cells of interest 41a is 3. The continuity value calculated for each cell is
The continuity value of the pixels included in each cell is used.

When selecting a representative color, which will be described later,
^It is assumed that a four-dimensional space is obtained by adding the coordinate axis of the continuity value to the ^* a ^* b ^* space, and the coordinate axis of the continuity value is a minimum value to a maximum value of the continuity value of the pixels included in the mosaic cell image data. Is created so as to have a class number to be described later (step S4).

In the above continuous value measuring method, the image data is temporarily converted into a mosaic image, and the color of each mosaic cell of the mosaic image is set to the color of the pixel that predominates in the mosaic cell, Although the continuity value of each cell is measured throughout, the image data may be divided into a plurality of regions, and the continuity value may be measured in each of the divided regions.

As a method of dividing the image data into a plurality of areas, for example, the outline of the object is extracted by emphasizing the edges of the object in the image data, and the image data is divided by the outline. A known method of dividing a digital image, such as a method of dividing a region by obtaining a spatial frequency distribution using a discrete cosine transform used in JPEG or the like, can be used.

Further, the numerical value of the goodness of the unity of the color components of each pixel on the image array is not necessarily limited to the measurement of the continuity value described above. In the above, a method of calculating a ratio of pixels of similar colors may be used.

(B) Conversion to Uniform Color Space In order to select a representative color, first, the color distribution of image data is arranged in a color space. At this time, the uniform color space is set in consideration of human visual characteristics. use. As a uniform color space, the above L ^* u ^*
Use v ^* or L ^* a ^* b ^* . Therefore, the value of each pixel of the image data composed of RGB data is represented by L ^* u ^* v ^* or
Convert to L ^* a ^* b ^* . In the present embodiment, as an example, the details of the conversion by the color conversion means 8 will be described using L ^* a ^* b ^* .

Conversion from RGB data to L ^* a ^* b ^* is performed by the following equation via the XYZ system recommended by CIE 1931. X = 0.61R + 0.17G + 0.20B, (1) Y = 0.36R + 0.59G + 0.11B, (2) Z = 0.066G + 1.12B (3)

When Y / Y _n > 0.008856, L ^* = 116 (Y / Y _n ) ^1/3 −16 (4) When Y / Y _n ≦ 0.008856, L ^* = 903.3 ( Y / Y _n ) (5) X / X _n > 0.008856, Y / Y _n > 0.00885
6, in the case of _{Z / Z n> 0.008856, a} * = 500 {(X / X n) 1/3 - (Y / Y n) 1/3}, (6) b * = 200 {(Y / Y _n ) ^1/3 − (Z / Z _n ) ^1/3 ｝ (7)

0.0088 for X / X _n , Y / Y _n and Z / Z _n
When there are 56 or less, the above equations (6) and (7)
(X / X _n ) ^1/3 , (Y / Y _n ) ^1/3 and (Z / Z _n ) ^1/3 terms are 7.787 (X / X _n ) +16/116 7.787 ( Y / Y _n ) +16/116 7.787 (Z / Z _n ) +16/116. Here, the chromaticity of R, G, B and white is N
It conforms to the TSC television standard. The RGB data of the NTSC television standard is converted into L ^* a ^* b ^* data recommended by the CIE according to the above conversion formula (step 5).

(C) Frequency distribution of pixels on the L ^* , a ^* , b ^* axis and the continuity axis The value of each pixel is obtained by converting the above-mentioned RGB into L ^* a ^* b ^* , an 8-bit integer. It will deviate from the value.
Therefore, to obtain the frequency distribution on the L ^* a ^* b ^* axis,
Determine the number of classes of each coordinate axis, and L ^{* according} to the class of each coordinate axis ^.
It is necessary to divide the a ^* b ^* space into clusters and obtain the number of pixels included in the clusters.

In order to divide the space spanned by L ^* a ^* b ^* and the continuity axis into the required number of colors, and to select the average color of each color region as a representative color, Watanabe's algorithm (Transactions of the Institute of Electronics, Information and Communication Engineers) Journal DVol.J70-DNo.4pp.720-726 1987.4) was modified and used.

The main improvement is that RG is used as the original data.
Instead of using the B upper 5 bits, a value obtained by converting the RGB 8-bit value into L ^* a ^* b ^* was used. This allows
The 8-bit precision of the data can be used as it is. In addition, the initial space to be divided is not an RGB three-dimensional space but a four-dimensional space spanned by four L ^* a ^* b ^* axes and a continuity axis. Thereby, the color continuity on the image can be reflected in the selection of the representative color.

The rank of each coordinate axis of the space is represented by L ^* a ^* b
^* 256 × 64 × 64 × in order of 4 axes of axis and continuity axis
As shown at 64, the rank of the coordinate axis indicating the luminance component L ^* is quadrupled with respect to the rank of the other coordinate axes, so that the luminance component L ^* is emphasized. As described in the report by Kira et al. (IEICE Technical Report, IE83-92, "Adaptive Natural Image Display with Limited Number of Representative Colors", Kira, Inoue, Fukui)
This is because human visual characteristics are considered to be about four times more sensitive to luminance components than chromaticity components.

More specifically, the input image data usually has RGB values arranged in pixel order. Therefore,
Considering an RGB space having coordinates of RGB values, RGB
The space is a three-dimensional space in which R, G, and B are coordinate axes, and each coordinate axis has an 8-bit value, that is, a value of 0 to 255. Therefore, a histogram of each color component is obtained by associating the number of pixels having the RGB value with a point in space. That is, histogram (r, g, b) = the total number of pixels of RGB (r, g, b).

Next, in the case of the histogram of the uniform color space L ^* a ^* b ^* , the value (r, r,
g, b) is converted to (L ^* , a ^* , b ^* ) by the above equation. That is, histogram (r, g, b) = L ^* a ^* b
^* (L ^* a ^* b ^* ) is the total number of pixels. However, since the value of L ^* a ^* b ^* is obtained by non-linear conversion, it deviates from an integer value of 0 to 255, and it is necessary to provide an appropriate class number to emphasize the luminance component.

In this embodiment, as described above, the luminance, that is, L ^*
In order to emphasize the components and prevent the data volume from increasing by the conversion to L ^* a ^* b ^* , it is necessary to determine the class interval of each L ^* , a ^* , b ^* . As the data capacity, since each value of RGB is set to 8 bits, the maximum is set to 8 bits. Therefore, the emphasized luminance L
^* Is assigned to 0 to 255. The a ^* and b ^* components take values from 0 to 63.

Here, the actual calculation of the class interval is performed for each L ^*
When a ^* b ^* is 8-bit or less data, it can be obtained by the following equation (1).

(Equation 1)

L ^* _interval , a ^* _interval , b
^* _interval indicates the class interval of L ^* , a ^* , b ^* . L ^* _max ,
a ^* _max and b ^* _max indicate the maximum values of L ^* , a ^* , and b ^* , respectively.
L ^* _min , a ^* _min , and b ^* _min indicate the minimum values of L ^* , a ^* , and b ^* , respectively. Further, m and n are changed by the ratio between the luminance component (L ^* ) and the chromaticity components (a ^* , b ^* ). In the present embodiment, L ^* : a ^* , b ^{* is changed} to 4: In order to set it to 1, the luminance component m is set to 1 and the chromaticity component n is set to 4.

Regarding the continuity value axis, a ^* , b ^*
In the same manner as the class interval of, the following equation 2 is used.

(Equation 2) C indicates the continuity value. C _interval indicates a class interval of C. C _max indicates the maximum value of C. C _min indicates the minimum value of C.

Then, the space spanned by the L ^* a ^* b ^* and continuity axis determined as described above is
The data is divided four-dimensionally according to the class of each coordinate axis, and is used as a basic unit when creating a histogram as described later. Then, the cluster storage unit 9 stores the four-dimensional space divided according to the rank of each coordinate axis as a cluster image as described above (step S6). The cluster storage unit 9 stores clusters that are sequentially divided as described later, with a cluster number attached thereto, and also stores pixels included in each cluster. Then, the histogram calculation unit 12 obtains a frequency distribution for each cluster in each coordinate axis of the image data (step S7).

Referring to FIGS. 5 and 6, the effect of selecting a representative color by discriminant analysis by making the rank of the coordinate axis indicating luminance larger than the rank of the coordinate axis indicating chromaticity or continuity value will be described. A brief description will be given. Note that in FIGS. 5 and 6, L ^*
A coordinate axis indicating luminance in a four-dimensional space spanned by a ^* b ^* and a continuity axis and one of coordinate axes indicating chromaticity are selected and described in a two-dimensional coordinate system.

First, when the y-axis is luminance (L ^* ), the x-axis is chromaticity (a ^* ), and each rank is 8,
It is assumed that the color space is divided by the discriminant analysis as shown in FIG. On the other hand, it is assumed that the color space is divided with the rank on the x axis being 4, and the rank on the x axis is set to 4, so that the rank on the y axis is 2 of the rank on the x axis.
In the case where the number is doubled, as shown in FIG. 6, the portion that could be divided in FIG. 5 on the x axis cannot be divided, and
The number of divisions of luminance increases.

That is, when the average color in the divided color space is set as the representative color, many colors having the same hue but different luminances are selected as the representative colors, and the colors of different hues are reduced, so that the luminance is emphasized. The selected representative color is selected.
Therefore, as described above, the rank of each coordinate axis of the space is represented by L ^* a ^*
b ^* axis and continuity axis: 256 × 64 × 64
By making the rank of the coordinate axis indicating the luminance component L ^* four times as large as the rank of the other coordinate axes, such as × 64, a representative color with enhanced luminance is selected.

(D) Enhancement of Luminance, Chromaticity, or Color Continuity Value In the present embodiment, in the space division method, the sampling of the luminance component is emphasized, so that a portion where the change of the luminance component is gradual is emphasized. That is, paying attention to the fact that the division with the luminance direction as the main axis is the main division, and further giving sampling in which the luminance component is emphasized with respect to the chromaticity component, conversely, the portion and the color whose luminance change is gradual can be reduced. A method for identifying has been invented.

Specifically, L ^* a ^* b ^* (or L ^* u)
^* v ^* ) and a four-dimensional space spanned by the continuity axis,
Emphasis sampling is performed 16 times in the luminance direction, such as 256 × 16 × 16 × 16, and the number of limited colors is specified by a smaller number, for example, 100 colors. Ask for. As a result, 100 colors in which the luminance component is enhanced 16 times in this case are selected. These are the top 100 colors that emphasize the color of the portion where the luminance changes slowly.

Then, as described above, the emphasis means 22 calculates the ratio of each coordinate axis to the variance in each coordinate axis direction of the four-dimensional space defined by the uniform color coordinate axis and the continuity axis obtained by the average / variance calculation means 14. Corresponding to the ratio, the ratio of the class number of the coordinate axis indicating the luminance to the class number of the coordinate axis indicating the continuity value and the chromaticity is 1
/ 8x, 1 / 4x, 1 / 2x, 1x, 2x, 4x, 8
It was made possible to select from any of x1, x16, x32, and x64.

The emphasizing means 22 can independently change the rank of the continuity axis by the number of representative colors to be limited. If the number of colors is relatively large, such as 100 colors or more, the number of classes on the color continuity axis is reduced and the division is weakened, so that the color change in the uniform color space can be gradually reduced. False contours are reliably suppressed by increasing the number of representative colors to be selected. Further, when the number of colors such as 32 or less is small, the number of classes on the color continuity axis is increased to emphasize the division, thereby enabling color selection that emphasizes the unity of hues.

Then, the emphasis coefficient calculating means 23 divides the space spanned by the uniform color coordinate axis and the continuity axis into clusters according to the rank determined as described above,
A frequency distribution is obtained in the same manner as in the above-described histogram calculation means 12, and normalization is performed with the frequency peak in this frequency distribution being set to 1 to obtain a coefficient for each class. Then, this coefficient is divided by the number of pixels included in each class to obtain a coefficient for each pixel. Then, the coefficient for each pixel is stored in the enhancement coefficient storage unit 2.
4 is stored (step S8).

(E) Change of histogram Using the coefficients stored in the emphasis coefficient storage means 24,
The histogram stored in the histogram storage unit 13 is updated with weight correction (step S9). Specifically, first, for each density value (for example, 0 to 255), the emphasis coefficients stored in the emphasis coefficient storage unit 24 of the pixels included in the density value are totaled, as shown in FIG. The coefficient 44 for each value is obtained.

Next, the histogram 43 stored in the histogram storage means 13 as shown in FIG.
Is multiplied for each density value by a coefficient 44 for each density value, and the obtained result is added to the histogram 43 for each density value, whereby a weighted and corrected histogram 45 shown in FIG. 9 can be obtained. Is used to select a representative color by a discriminant analysis described later.

(F) Selection of representative color by discriminant analysis The selection of a representative color by the discriminant analysis method of the present embodiment is basically the same as that of the above-mentioned prior art, and a threshold value is determined by using a histogram for each color component. Is obtained, and the color space is sequentially divided into a number of clusters at this threshold value. In this embodiment, a four-dimensional space in which coordinate axes indicating the continuity are further set in the color space is used. Discriminant analysis is performed using the weighted histogram.

Here, first, in the four-dimensional space or the cluster divided as described later from the histogram, the average and variance of each coordinate axis of the cluster (firstly, the four-dimensional space) is obtained by the average / variance calculating means 14. (Step S10). Further, as described above, the main axis of the coordinate axes of each cluster is determined by the main axis calculation means 10. Further, the weighted variance of each cluster is obtained by the weighted variance calculating means 16 and the cluster sorting means 2
A cluster having the largest weighted variance is obtained from 0, and division is performed on the main axis of the cluster having the largest weighted variance by a division threshold determined as follows.

[0093] Here will be described a method for determining the threshold value T _i by the discrimination analysis of dividing threshold determining means 17. When the number of pixel areas S N, the number of pixels in the density values i and n _i, the occurrence probability q _i is given by the following equation. q _i = n _i / N Then, with the density value k, the section [a, b] is divided into two sections C ₁
[A, k] and C ₂ [k + 1, b] are assumed, and the pixel occurrence probabilities ω ₁ , ω ₂ , the average value μ ₁ , μ ₂ , and the variance σ ₁ ² , σ ₂ ² a weighted variance sum sigma _w ² given by: when the determined density values k so as to minimize. σ _w ² = ω ₁ ω ₂ (μ ₁ −μ ₂ )

The k that maximizes σ _B ² (that is, the threshold value T _{i to be obtained} ) can be obtained sequentially as shown in the following equation (3).

(Equation 3) Here, μ _T is the total average and is shown in the following Expression 4.

(Equation 4)

[0095] As described above, it is possible to determine the threshold value T _i of each color component and the continuity values (step S1
1). Then, the four-dimensional space is first divided into two regions by the obtained threshold value T _i , and the main axis calculation means 10 determines the color component having the largest variance σ ² among the respective color components. or continuity value and spindle, at the position of the principal axis corresponding to the threshold T _i of the color components or continuity value, the cluster dividing unit 19 divides the four-dimensional space by a curved surface which is perpendicular to the main axis into two clusters (step S1
2).

Next, in the newly divided cluster, the average and variance are obtained by the average and variance calculating means 14.
The weighted variance is obtained by the weighted variance calculating means 16,
The main axis calculation means 10 finds the main axis, the cluster sorting means 20 compares the weighted variances among all the clusters in the four-dimensional space, and sorts the clusters in descending order of the weighted variances, thereby giving the largest weighted variance. A cluster is obtained, and this cluster is set as a region to be divided next (step S13). Then, in the obtained cluster,
The threshold value Ti is determined as if the color space was divided, and the cluster is further divided into two clusters.

As described above, of all the divided clusters, the cluster having the largest weighted variance is obtained.
The process of dividing the obtained cluster into two as described above is repeated. Then, when the number of divided clusters becomes equal to the number of representative colors to be displayed, the division ends (step S14). Next, the representative color allocating unit 27 sets the average value of the density values for each color component in each cluster as the representative color (step S15), and the color conversion unit 26 performs the inverse conversion of the color conversion unit 8 on the representative color. The representative color is R
The GB value is set (step S16).

Then, the representative color is registered as a color code in the color map 30d of the color map display device 30, and the color distribution in the color space of the color image data to be displayed is associated with each of the clusters. A color similar to the original color of the pixel (the average color of the cluster to which the pixel belongs) is assigned to the pixel as a color code from the color map 30d. In this manner, the image data of which color is assigned to each pixel and whose color is limited is displayed on the display 30b of the color map display device 30 having the color map 30d in which the color code is registered.

(G) Determining the representative color by repeating the above operation In the limited color determining method of the present embodiment, the representative color selected as described above is repeatedly selected for the selected representative color. The colors are gradually reduced. That is, an iterative control unit (not shown) is provided, and the image represented by the representative color selected as described above is selected as the original image data again by the above-described operation to select the representative color. It is also possible to reduce the number of representative colors finally required.

Each time a representative color is repeatedly selected, the continuity value calculating means 6 changes the coordinate axis of the continuity value every time, and the emphasis means 22 and the emphasis coefficient operation means 23 are selected.
Is calculated by the histogram updating means 25.
Updates the histogram, weights the gentle part of the color change, and emphasizes the luminance or chromaticity or continuity value. Therefore, the image data is
With regard to the statistical properties of chromaticity distribution, luminance distribution and continuity value distribution, representative colors will be narrowed down while being adjusted, and suitable for the statistical properties of image data as well as the adaptive method. The representative color will be selected.

The limited color image data can be stored in a storage device or the like. The image once processed can be instantaneously displayed on a display. Since the memory capacity is reduced as compared with the image data of the original natural image, an apparatus having a small memory capacity can be easily handled.

With the above-described method and apparatus for determining a limited color according to the present embodiment, a natural image that should be originally displayed on a full-color display device can be displayed on a color map display device that can simultaneously display only a small number of colors. Can be displayed without. That is, when the representative color is determined by the discriminant analysis, by making the class number on the coordinate axis indicating luminance different from the class number on the coordinate axis indicating chromaticity, it is possible to correspond to the distribution of luminance and chromaticity of the image. Thus, it is possible to select a representative color whose luminance is emphasized or a representative color whose chromaticity is emphasized. In particular, by selecting a representative color whose luminance is emphasized, it is possible to select a representative color corresponding to a human visual characteristic which is sensitive to the luminance, and to suppress the occurrence of a false contour.

Further, the limited color determining apparatus of this embodiment specifies a portion where the color changes gradually, based on the color continuity of the upper 4-bit image of the image data, and obtains this continuity for each pixel. By providing a four-dimensional space in addition to the three-dimensional space of the color space as a coordinate axis, when the number of colors is relatively large, such as 100 colors or more, the division on the color continuity axis is weakened, so that the color By increasing the number of representative colors selected from a portion having a gradual change, false contours can be surely suppressed. Further, when the number of colors such as 32 colors or less is small, the division on the color continuity axis is emphasized to enable the color selection with the hue unit being emphasized, and the color faithful to the hue distribution of the original image. You can make a choice.

Note that the entirety of these processes can be configured as a program on a general-purpose computer and executed. Further, the memory capacity per pixel of the original image data does not necessarily have to be 24 bits, and the upper bits used for specifying the color change need not be the upper 4 bits. The number of high-order bits to be taken may be changed depending on the memory capacity per pixel of the original image data.

To reduce the number of colors according to the present invention,
Processing time is required, but when displaying the resulting compressed image, complex processing is not required, so compressed images with limited colors can be displayed instantaneously, regardless of differences in machines and systems. It is. Of course, the image data of which the number of colors has been reduced according to the present invention may be subjected to the recommended image compression using JPEG or the like, and the compression rate is more effectively increased. For these reasons, the limited color determination method and apparatus of the present invention can be expected to be used in many industrial fields.

[0106]

According to the method and the apparatus for determining a limited color according to the first and third aspects, the goodness of the color components of each pixel on the image array is converted into a numerical value, and the color components are newly added to the uniform color space. By setting a coordinate axis indicating the cohesiveness of the multi-dimensional space, a discriminant analysis is performed on the frequency distribution of each pixel of the image data expanded in this multi-dimensional space, and a limited representative color is obtained. It is possible to select a representative color corresponding to the relationship between the position of each pixel on the image and the color, that is, the hue distribution of the original image.

Therefore, unlike the case where each pixel of the image data is simply developed on a uniform color space and discrimination analysis is performed on the frequency distribution of the pixels on the uniform color space to determine representative colors, extremely few representative colors are determined. Even if the image data is approximately displayed, the subject of a similar color drawn on the image data can be drawn in a different color.

Therefore, even for natural images such as CD-ROMs and CD-Is that need to reduce the number of colors in particular,
By identifying a portion where a sense of discomfort is conspicuous and obtaining a limited number of specified colors with less discomfort from the portion, it is possible to deal with the problem effectively, which is very effective in compressing image data.

Further, according to the method and apparatus for determining a limited color according to the second and fourth aspects, the first and the fourth aspects are set forth.
The same effects as those of the described limited color determination method and apparatus can be obtained, and the image is divided in advance based on the color change between pixels, thereby quantifying the goodness of the color components of each pixel on the image array. In doing so, many methods can be used.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an outline of a limited color determining apparatus according to an embodiment.

FIG. 2 is a flowchart illustrating an algorithm according to a limited color determination method of the embodiment.

FIG. 3 is a diagram for explaining the above-described limited color determining method.

FIG. 4 is a view for explaining the above-described limited color determining method.

FIG. 5 is a drawing for explaining the above-described limited color determining method.

FIG. 6 is a diagram for explaining the above-described limited color determining method.

FIG. 7 is a diagram for explaining the above-described limited color determining method.

FIG. 8 is a diagram for explaining the above-described limited color determining method.

FIG. 9 is a diagram for explaining the above-described limited color determining method.

[Explanation of symbols]

 2 Upper 4-bit image storage means (Numerization means) 3 Mosaic image calculation means (Numerization means) 4 Mosaic cell color determination means (Numerization means) 5 Mosaic cell color storage means (Numerization means) 6 Continuity value calculation means (Numerization means) 7 Continuity value storage means (Numerization means) 9 Cluster storage means (Frequency distribution operation means) 10 Spindle operation means (Discriminant analysis operation means) 11 Spindle storage means (Distinction analysis operation means) 12 Histogram operation means (Frequency distribution calculation means) 13 histogram storage means (frequency distribution calculation means) 14 average / variance calculation means (discrimination analysis calculation means) 15 average / variance storage means (discrimination analysis calculation means) 16 weighted distribution calculation means (discrimination analysis calculation) Means) 17 division threshold determination means (discrimination analysis operation means) 18 division threshold storage means (discrimination analysis operation means) 19 cluster division means ( Another analysis computing means) 20 cluster sorting means (discriminant analysis operations unit) 21 division control means (discriminant analysis operations unit)

──────────────────────────────────────────────────続 き Continued on the front page (58) Fields surveyed (Int.Cl. ⁷ , DB name) G06T 1/00 510 G06T 5/00 100 G06T 11/00 110 G06T 11/60 120 G09G 5/02 G09G 5 / 06 JICST file (JOIS)

Claims (4)

(57) [Claims] 1. A color image data expressed by a coordinate system of three primary colors is converted to a coordinate system of a uniform color space, and discriminant analysis is performed on a frequency distribution of each pixel of the image data in the converted coordinate system. In the limited color determination method for obtaining a representative color and approximating the image data with a limited representative color, the distribution of the color components of each pixel arranged on the image of the image data on the image array From the above, the goodness of the unity of the color components of each pixel on the image array is quantified, and the goodness of the unity of the color components of each pixel on the quantified image array is associated with each color component of each pixel. A multi-dimensional space is provided by newly setting coordinate axes indicating the goodness of the color components in the color space, and the frequency distribution of each pixel of the image data in the coordinate system of the multi-dimensional space is obtained. Against And determining a representative color by performing a discriminant analysis. 2. Converting color image data expressed by a coordinate system of three primary colors into a coordinate system of a uniform color space, and performing discriminant analysis on a frequency distribution of each pixel of the image data in the converted coordinate system. In the limited color determination method for obtaining a representative color and approximating the image data with a limited representative color, the distribution of the color components of each pixel arranged on the image of the image data on the image array From the above, the image is divided into regions that can be divided based on the color change between pixels, and to which region of the image each pixel belongs, for each region, the goodness of the unity of the color components of each pixel inside it Is converted into a numerical value, and the numerical value of the unity of the color components is associated with each color component of each pixel, and a new coordinate axis indicating the unity of the color components is set in the color space. Provided, this multi-order Obtains a frequency distribution of the pixels of the image data in the coordinate system of the space, limited color determination method characterized by determining the representative color by performing discriminant analysis on said frequency distribution. 3. Converting color image data expressed by a coordinate system of three primary colors into a coordinate system of a uniform color space, and performing discriminant analysis on a frequency distribution of each pixel of the image data in the converted coordinate system. A limited color determination device for obtaining a representative color by the above method and approximating the image data with a limited representative color, wherein a distribution of a color component on an image array of each pixel arranged on the image of the image data is provided. Numericalizing means for quantifying the goodness of the color components of each pixel from, and the uniformity of the color components quantified by the quantifying means in association with the color components of each pixel in the uniform color space A four-dimensional space is provided by newly setting a coordinate axis indicating the goodness of color component cohesion; a frequency distribution calculating means for obtaining a frequency distribution of each pixel of the image data in a coordinate system of the four-dimensional space; A discriminant analysis calculating means for obtaining a representative color by performing a discriminant analysis on a frequency distribution of each pixel of the image data in the four-dimensional space obtained by the cloth calculating means. Color determination device. 4. Converting color image data expressed in a coordinate system of three primary colors into a coordinate system of a uniform color space, and performing discriminant analysis on a frequency distribution of each pixel of the image data in the converted coordinate system. In the limited color determination device for obtaining the representative color by the above and approximating the image data with the limited representative color, the color component on the image array position of each pixel arranged on the image of the image data is determined. An image dividing unit that divides the image from the distribution into regions that can be divided based on a color change between the pixels, and to which of the regions divided by the image dividing unit each pixel of the image data belongs Pixel-affected region calculating means, and numerical value means for quantifying the goodness of unity of the color components of each pixel from the distribution of the color components of each pixel in the region; color components quantified by the numerical value means Nomatoma The four-dimensional space is provided by associating the goodness of each pixel with each color component of each pixel and newly setting a coordinate axis indicating the goodness of the unity of the color components in the uniform color space. Frequency distribution calculating means for obtaining a frequency distribution of each pixel of the image data; and representative color by performing discriminant analysis on the frequency distribution of each pixel of the image data in the four-dimensional space obtained by the frequency distribution calculating means. And a discriminant analysis calculating means for determining the color difference.