JPH04190466A - Limited color representation device for color image - Google Patents

Abstract

PURPOSE:To obtain a proper color tone by dividing a color space for representing the color image into a finite number of areas, selecting representative colors of the respective areas and calculating color space distances between the representative colors and representative colors in a table where similar colors are registered, and assigning the representative color with the shortest distance to an original picture element. CONSTITUTION:The image is read first, a detecting circuit 13 extracts high-order 6 bits from 8-bit data R, G, and B, and linear color values are obtained from a prepared three-dimensional/linear conversion table 11. The color values are divided by a histogram generating means 14 and an area dividing means 15 into (10<8>) areas which is in inverse proportion to the amount of the image data. A means 16 selects representative colors according to the mean values of the respective areas, reconverts the data into 8-bit constitution, and the color space distances D between the representative colors are calculated from an equation I; and color codes which are shorter than a certain value are displayed in a representative color relation table 21. Then assigned colors are determined according to the shortest distances between temporary representative colors generated with the 6-bit data as they are and the color codes in the relation table.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to a limited color expression device for a color image that expresses a digitized natural color image using a limited number of representative colors.

[Prior Art] A color image is usually represented by three components: R (red), G (green), and B (blue), and digitized image data is often composed of 8 bits for each component. When displaying such digital color images on a display device,
As shown in FIG. 4, the three components of R, G, and B in one pixel are each represented by 8 bits in the image memory 1, and these components are converted into analogs by D/A converters 2a, 2b, and 2c, respectively. In the full-color display device that supplies display 3, one pixel is represented by 24 bits, so it is possible to display 224 (←16.77 million) colors at the same time, but each pixel requires 24 bits of memory. The required capacity of the image memory 1 becomes enormous.

On the other hand, color display devices used in personal computers, workstations, etc. are provided with, for example, an 8-bit image memory 4 for each pixel, as shown in FIG. R, G for actually displaying this color code using a conversion table using color map
, B values (8 bits each), and then convert each component into D/A
Conversion i6a, 6b, 6c respectively convert analog to C
The configuration is such that the data is supplied to the RT display 7.

In this device, the output from color map 5 is 24 pixels per pixel.
Since it is expressed in bits, it is possible to select a color from 16.77 million colors, or the capacity for one pixel of the image memory 4 is 8.
Since it is a bit, the number of colors that can be displayed simultaneously is 256. However, since this device requires only a small amount of image memory, it is desirable to be able to display natural images with this limited number of colors without causing any discomfort.

For this reason, in the past, a histogram was created to examine the color distribution of the original image, and color areas with a large distribution (color areas of the image with a large number of pixels and a high degree of dispersion) were For color areas with a small distribution, the color space is divided sparsely to obtain the same number of color areas as the limited number of colors, then a representative color is selected within each area as the expression color, and then applied to each original pixel. It is known that representative colors are assigned and limited colors are displayed.

The following two methods are used for assigning representative colors.

One method is to calculate the Euclidean distance between each pixel color in the original image and all representative colors, and assign the representative color with the shortest distance to the corresponding pixel in the limited color image. In this method, the color of each pixel or the representative color of the area to which it belongs is directly assigned to the corresponding image of the limited color image.

[Problems to be Solved by the Invention] However, with the former method of calculating the Euclidean distance between the original pixel and all representative colors and assigning a representative color, a limited color image with high image quality can be obtained; If the original image size is, for example, 512 x 512 pixels and the number of representative colors is 256, calculations must be performed 512 x 512 x 256 times, resulting in an enormous amount of calculation, making it difficult to display limited colors at high speed.

In addition, in the latter method of assigning the original pixel to the representative color of the area to which it belongs, if the correspondence between the representative color and the area is created when dividing the color space into regions, the assignment can be done by simply converting the table, which is fast. However, depending on how the area is divided, it may not be possible to allocate an appropriate representative color.

Especially when dealing with images in which similar colors change smoothly, such as the skin in portraits (hereinafter referred to as gradation images), or when performing segmentation processing after expanding the color space into a one-dimensional space. There is a high possibility that multiple similar representative colors will be selected, and since multiple small regions are concentrated in a small space, the probability that the color will be the representative color of the region to which it belongs or the ideal representative color is low. In addition, in such a gradation image, unless ideal representative colors are used, visual deterioration in image quality becomes noticeable.

Therefore, the present invention reduces the amount of calculations required when assigning representative colors, allows high-speed assignment of representative colors, and always assigns suitable representative colors to original pixels, especially for gradations. The present invention aims to provide a limited color expression device for color images that can allocate ideal representative colors to sexual images.

[Means for Solving the Problems] The present invention provides a color distribution creation means for creating a color distribution in the color space of a digitized natural color image, and a method based on the color distribution created by the color distribution creation means. A region dividing means for dividing a color space into a limited number of regions, a representative color selection means for selecting a representative color of each region divided by the region dividing means, and each representative color selected by the representative color selection means. A representative color related table that registers similar colors between colors, the color of each original pixel, the representative color of the area to which the original pixel color belongs, and the representative color registered in the related table of the representative color of the area to which it belongs. The layout for calculating the three-dimensional Euclidean distance between the two is provided with a distance calculation means and a representative color allocation means for allocating the representative color with the shortest distance to the original pixel based on the calculation result by this distance calculation means. .

[Operation] In the present invention having such a configuration, a color distribution in the color space of a digitized natural color image is created, and the color space is divided into a limited number of regions based on this color distribution. Select a representative color for each area. Further, similar colors between the selected representative colors are registered in the representative color related table.

Then, the three-dimensional Euclidean distance for allocation is calculated between the color of each original pixel, the representative color of the area to which the original pixel color belongs, and the representative color registered in the association table of the representative colors of the area to which it belongs. , the representative color with the shortest distance is assigned to the original pixel based on this calculation result.

[Example] Hereinafter, one embodiment of the present invention will be described with reference to the drawings.

There are usually 3 color spaces such as R, G, B, Lx, am, and bx.
It is a dimensional space, for example R, G.

The space of B is expanded into a one-dimensional space in advance. That is, the color space of the upper 5 bits of R, G, and B, each represented by 8 bits, is developed one-dimensionally.

In this case, the total combination of R, G, and B is 215-3-2
, 768 pairs exist. Therefore, numbers from 0 to 32,767 are assigned to the combinations of RlG and B to create a conversion table for the correspondence between the one-dimensional space and the three-dimensional space, that is, the three-dimensional/one-dimensional color space conversion table 11. put. The Peano curve is used as a method of expansion into a one-dimensional space.

Specifically, a natural color image is read by a color image scanner 12 and output as digital image data of 8 bits each for R, G, and B.

First, the first reading is performed at high speed with the resolution reduced and digital image data is output.

Then, the upper 6 bits detection circuit 13 validates only the upper 6 bits of each color, and the 3D/1D color space conversion table 11 outputs the color value in the 1D space.

Then, the distribution of color values of all pixels of the image is created on a 1□-dimensional space by the histogram creation means 14 as a color distribution creation means.

Next, the entire color space is divided into a limited number of regions, for example, 256 regions, by the region dividing means 15 based on the distribution of the histogram. This area is divided into small areas based on the frequency and variance of the histogram, and is divided into areas where the frequency is high and the variance is large.

The region division work up to this point has been performed on a one-dimensional space.

Next, the average value within each area is calculated by the representative color selection means 16, and the representative color of that area is selected.

The average value at this time is determined by converting the color value in the one-dimensional space into R, G,
The value is an 8-bit value for each B color, and the average value is calculated for each color.

Then, the representative color code and the RlGSB value of the representative color are registered in the color lookup table 18 as shown in FIG. That is, each color code number "0" to r255
The values of RlG and H in the three-dimensional color space are assigned to J.

19 is representative color distance calculation means, and this representative color distance calculation means 19 calculates three-dimensional Euclidean distances between 256 representative colors. For example, two points (r, g,
The three-dimensional Euclidean distance between b) and (r', g'+ b') is D= (r-r')2+(gg')2+(bb'
) becomes 2.

Then, codes of representative colors whose distances between the representative colors are within a certain threshold are selected by the registered representative color selection means 2° and registered in the representative color relationship table 21. As shown in FIG. 3, this representative color relationship table 21 has another representative color code registered for each representative color code, and the number of registered colors differs depending on the representative color. In other words, in areas with gradation of an image, the dispersion becomes large on the histogram, so it is easy to divide the image into small areas, and many similar representative colors are extracted, and a related table of selected representative colors is created. It becomes easy to register many different representative colors.

Through the above control processing, representative colors are determined and a representative color relationship table is created.

Once the representative colors have been determined in this way, the assignment of the representative colors is then executed.

That is, the color image scanner 12 performs a second reading and outputs digital image data.

After each of the 8 bit data of R, G, .B is converted into 6 bits each by the upper 6 bit detection circuit 13,
The dimensional/one-dimensional color space conversion table 11 provides color values in a one-dimensional space.

The color values in this one-dimensional space are supplied to a temporary representative color conversion table 22, and a temporary representative color is determined. Here, the provisional representative color is the same as the representative color assigned by the conventional method of assigning the color of each pixel of the original image or the representative color of the area to which it belongs as is to the corresponding pixel of the limited color image.

Subsequently, for allocation, a three-dimensional Euclidean distance is calculated by the distance calculation means 23 between the 8-bit original pixel of each color, the temporary representative color, and the representative colors registered in the temporary representative color association table 21. Further, for the shortest distance representative color assignment, the representative color with the shortest distance is assigned by means 24 to create a limited color image.

In this embodiment with such a configuration, first, the resolution is reduced and the image is read by the color image scanner 12, and after being converted into color values in the one-dimensional space by the three-dimensional/one-dimensional color space conversion table 11, The histogram creation means 14 creates a histogram distribution on a one-dimensional space. Then, the area dividing means 15 divides the area into 256 areas.

Subsequently, the representative color selection means 16 calculates the average value within each region, selects the representative color of the region, and inputs the representative color code, R, G, and R of the representative color to the color lookup table 18.
The value of B is registered. Further, the representative color distance calculating means 19 calculates three-dimensional Euclidean distances between the 256 representative colors.

Then, codes of representative colors whose distances between the representative colors are within a certain threshold are selected by the registered representative color selection means 20,
It is registered in the representative color relationship table 21.

Once the representative colors are determined in this way, the second reading is then performed by the color image scanner 12 and assignment of the representative colors is executed.

That is, the digital image data from the color image scanner 12 is converted into six bits each by the upper six bit detection circuit 13, and then converted into color values in a one-dimensional space by the three-dimensional/one-dimensional color space conversion table 11. Then, the temporary representative color is determined by the temporary representative color conversion table 22.

Subsequently, for allocation, a three-dimensional Euclidean distance is calculated by the distance calculation means 23 between the 8-bit original pixel of each color, the temporary representative color, and the representative colors registered in the temporary representative color association table 21. Further, for the shortest distance representative color assignment, the representative color with the shortest distance is assigned by means 24 to create a limited color image.

By using the representative color related table 21 in this manner, the amount of calculation when assigning representative colors can be reduced, and representative color assignment can be processed at high speed. Furthermore, an appropriate representative color can always be assigned to the original pixel, and in particular, an ideal representative color can be assigned to a gradation image.

[Effects of the invention] As detailed above, according to the present invention, it is possible to reduce the amount of calculations when assigning representative colors and perform high-speed assignment of representative colors. It is possible to provide a limited color expression device for color images, which can allocate representative colors, and in particular, can allocate ideal representative colors to gradation images.

[Brief explanation of drawings]

1 to 3 show an embodiment of the present invention, in which FIG. 1 is a block diagram, FIG. 2 is a diagram showing the configuration of a color lookup table, and FIG. 3 is a diagram showing the configuration of a representative color related table. FIGS. 4 and 5 are block diagrams showing conventional examples. 14... Histogram creation means, 15... Region dividing means, 16... Representative color selection means, 21... Representative color related table, 23... Distance calculation means for allocation, 24... Shortest distance Distance representative color assignment is a means. Applicant's agent Patent attorney Takehiko Suzue Figure 2, Figure 3

Claims (1)

[Claims] A color distribution creation means for creating a color distribution in the color space of a digitized natural color image, and an area for dividing the color space into a limited number of regions based on the color distribution created by the color distribution creation means. A dividing means, a representative color selecting means for selecting a representative color of each area divided by the area dividing means, and a representative color related table for registering similar colors between the representative colors selected by the representative color selecting means. An allocation distance that calculates the three-dimensional Euclidean distance between the color of each original pixel, the representative color of the area to which the original pixel color belongs, and the representative color registered in the association table of the representative colors of the area to which it belongs. 1. A limited color expression device for a color image, comprising a calculation means and a representative color allocation means for allocating a representative color having the shortest distance to an original pixel based on a calculation result by the allocation distance calculation means.