JPH079671B2 - Color image converter - Google Patents

Description

Description: TECHNICAL FIELD The present invention relates to a color map used in a color image display device having a frame buffer, and in particular, an important color is automatically extracted from the color image to form a color map. The present invention relates to a color image conversion device that registers and approximates a color image using this color map.

[Conventional technology]

Color image display devices having a frame buffer are used in various fields, and the demand at home is also increasing. It is said that human vision can distinguish at least 50,000 or more colors. To express such a color, 1
A frame buffer with more than 15 bits per pixel is required.

Displaying a color image using a color map has an advantage that various color expressions can be performed with a smaller frame buffer, and a composite image can be efficiently displayed by a computer. That is, the color map is a correspondence table of index numbers and colors, and the number of indexes is 256.
, Then all Rs that may appear in the image
256 from any combination of (red), G (green), and B (blue) color components
Colors can be extracted and registered in the color map.
256 colors can be displayed simultaneously. However, in order to input an actual color image and display this input image using the color map, it is necessary to extract the colors to be registered in the color map from the input image. However, there is a problem that the deterioration of the image quality must be minimized. Regarding this problem, from the viewpoint of color quantization, it has been attempted to automatically extract a color to be registered in the color map from the input image and use the color to represent the input image.

FIG. 9 is, for example, the entire flow chart of the system shown in “Color image conversion device” by the same inventor previously applied, and FIGS. 10 and 11 are the color map and color shown in step 91 of FIG. It is a more detailed flowchart of the space division map creation unit. The former is called a frequency priority method and the latter is called a distance priority method.

In FIG. 9, step 31 is to display a color image on a TV camera,
The R (red) component, the G (green) component, and the B (blue) component input from a magnetic disk or the like are written in the image memory as M-bit data. Step 32 scans all or some pixels of the input image to create a histogram, that is, a color occurrence frequency distribution. In step 33 extracts color below 2 ^3M color from among 2 ^3M possible colors appearing in the input image as candidate colors. For example, if each axis of RGB 3D space is Nr, Ng,
And Nb division into Nr × Ng × Nb rectangular parallelepipeds, and a method of extracting a color having the highest occurrence frequency in each rectangular parallelepiped as a candidate color is conceivable, but the method is not limited to this. Step 34 is the initial integration, and the set of candidate colors is Is a three-dimensional vector represented by R, G, B components ), The RGB three-dimensional space C ³ (C ³ is a set composed of 2 ^3M colors) is a candidate color. Is divided into N regions Si associated with. However, Si
Is a set of colors that satisfies the following equation (1).

^{C 3 = Ui ▲ N-1} = 0 ▼ Si (1) Si ∩ Sj = φ (i ≠ j) In addition, the 2 ^3M color of the region number (Si defined by the formula in Step 34 (1) i Corresponding number) is described in a color space division map as shown in FIG. 14, and one representative color is extracted from each area. Step 35 creates an adjacency table for the areas of RGB space. The method of checking the adjacency is to formally divide the color space division map into S (r, g, b) (color (r, g, b) is the area S.
(It means that it belongs to (r, g, b)) can be considered as follows. First, all colors (r, g,
Regarding b), S (r, g, b) = S (r + 1, g, b) (2) S (r, g, b) = S (r, g + 1, b) (3) S (r, g, b) = S (r, g, b + 1) (4) is examined. If the formula (2) is not satisfied, the region S (r, g, b) and the region S (r + 1, g, b) are adjacent to each other, and if the formula (3) is not satisfied, the region S (r, g g, b)
And the area S (r, g + 1, b) are adjacent to each other, and when the expression (4) is not satisfied, the area S (r, g, b) and the area S (r, g, b + 1) are adjacent to each other. If the above process is performed for all the colors in the RGB space, the adjacency relationship of regions can be examined. FIG. 12A shows an example of the configuration of the adjacency relationship table. In the figure, the region i is adjacent to the region of the column number where the logic 1 stands when the row i is viewed in the horizontal direction, and the region i is adjacent to the regions 2 and j. ing. It is always 0 on the diagonal. Step 91 is a color map and color space division map creation unit, which selects two areas from the area, integrates both areas, and extracts a representative color of one color from the integrated area. To reduce the number of colors. And finally the representative color that remained remains
As shown in Fig. 13, the color map is registered, and the correspondence relationship between all colors in the RGB space and the index number of the color map is registered in the color space division map. The processing method of this part will be described separately in detail with reference to the flowcharts of FIGS. 10 and 11. In step 38, the R, G, B component data of each pixel of the input image on the image memory is read out, converted into the index number of the color map using the color space division map, and the index number is written in the image memory.

FIG. 10 is a detailed flowchart of the frequency priority method which is the first processing method of step 91 shown in FIG. In the figure, step 101 determines whether or not the number of representative colors has reached a predetermined number as a result of region integration.If the number of representative colors does not reach the predetermined number, steps 102 to 104 are executed, and the predetermined number is determined. When the number becomes, the steps 105 to 106 are executed and the process is terminated. In step 102, the minimum frequency region in which the color occurrence frequency is the minimum is extracted from each region, and step 103
Then, the area adjacent to the minimum frequency area is obtained by referring to the adjacent relationship table, and the area is integrated with the area having the representative color with the minimum distance. The new area created at this time is the union of both areas, the frequency is given by the sum of the frequencies of both areas, and the representative color is given by the weighted average of the representative colors of both areas. In step 104, the adjacency relationship table is updated. FIG. 12 shows a specific example of updating the adjacency table. The update result when the area i is integrated into the adjacent area j in FIG. 9A is shown in FIG. That is, row i
And the row j are ORed and the result is written in row j, and the column i and the column j are ORed and the result is written in column j, and
Write a logical 0 in row i, column i and diagonal components. This is the end of updating the adjacency table. The above steps 101 to 104 are repeated. When the number of colors reaches a predetermined number, the representative color of each area is registered in the color map in step 105, and the color space division map is created in step 106.

FIG. 11 is a detailed flowchart of the distance priority method which is the second processing method of step 91 shown in FIG. In the figure, step 111 refers to the adjacency relation table and calculates all the distances between the representative colors of the adjoining regions.
Step 112 determines whether or not the number of representative colors has reached a predetermined number as a result of integrating the areas.If the number of representative colors does not reach the predetermined number, steps 113 to 115 are executed to reach the predetermined number. If so, steps 105 to 106 are executed and the process ends. In step 113, the area pair with the smallest distance is extracted from the distances between the adjacent areas calculated in advance,
At 114, the region pairs are integrated. The parent area created by the integration is the union of both areas, the frequency is given by the sum of the frequencies of both areas, and the representative color is given by the weighted average of the representative colors of both areas. In step 104, the adjacency relation table is updated. Step 115 calculates the distance between the representative colors of the area adjacent to the new area and the parent area. Steps 112-1 above
If the number of colors reaches a predetermined number by repeating the processing of 15, step 105
While registering the representative color of each area in the color map with,
In step 106, a color space division map is created.

A conventional color image conversion device is configured to execute the above algorithm, and when a color image is input, a representative color is automatically registered in a color map and the input image is converted and displayed. It has become.

[Problems to be solved by the invention]

In the conventional color image conversion device as described above, when the frequency priority method, which is the first processing method, is used as the color map creation method, the low frequency area is integrated into the high frequency area. Although the image quality of the image does not deteriorate significantly, very close colors may be registered in the color map in a high frequency area, and as a result, for example, it may be painted in one color on the image plane. There is a problem that two or more colors that are close to each other appear like noise in a desired area. Further, regarding the color reproducibility, there is a problem that an infrequent color disappears or the color changes.

On the other hand, when the distance priority method, which is the second processing method, is used as the color map creation method, colors that are close to each other are not registered in the color map. Colors do not appear like noise (has a noise removal effect), and even if the frequency is low, colors that are distant from other colors are registered in the color map, so the reproducibility of such colors is good. Has become. However, in the case where colors that are close to each other exist in series in the color space, the integration of regions is performed in a chain reaction, so that colors that are far from each other are represented by the same representative color. was there.
Further, if there is even one color that is very far away due to the effect of noise during image input, there is a problem that that color is registered in the color map.

The present invention has been made to solve such a problem, and when an input color image is represented by a limited number of representative colors, an appropriate color is registered as a representative color in a color map, A color image that prevents colors that are very close to each other from being registered in the color map, and that if they are not noise components, colors that are far from other colors even if the frequency is somewhat small are registered in the color map. The purpose is to obtain a conversion device.

[Means for solving problems]

(1) Color image input device 1 and color image memory 2
A histogram creating means 3 and a histogram counter group 4, a candidate color extracting means 5, an initial integrating means 6, and an area having the smallest occurrence frequency of colors among the areas, and extracting the minimum frequency area as a color space. The number of regions is reduced to a predetermined number by combining the adjacent regions above with other regions and expressing the integrated region in one color, and the colors representing the resulting regions are indexed. First color map creating means 9 for creating a correspondence between all colors that can appear in the input image and index numbers of the color map in the color space division map 12 while registering in the numbered color map 11. , Combining the pair of areas with the smallest distance between the representative colors among the adjacent areas in the color space, and repeating expressing the integrated area with one color The number of areas is further reduced to a predetermined number of colors, and the colors representing the respective areas generated as a result are registered in the color map 11, and all colors that can appear in the input image are associated with index numbers of the color map. Second color map creating means 10 for creating a relationship in the color space division map 12, and color image conversion for converting each pixel of the input image into an index number of the color map according to the color space division map and storing it in an image memory. Means 13
The adjacency relationship table creating means 7, the adjacency relationship table 8, the display device 14, the adjacency relationship table 8, the color map 11 and the color space division map 12 into the first or second color map creating means 9, In a color image conversion device equipped with a series of changeover switches 15 for changing over to any one of 10, a first color image is initially inputted, and after the processing by the initial integration means 6 is completed, a first or second color map is created. The processing by the means 9 is continued until a predetermined number of colors is reached according to a predetermined threshold value, and further switched to reduce the number of colors by the second or first color map creating means 10 to reduce the color map and the color. A color image conversion device characterized by creating a space division map.

(2) Color image input device 1 and color image memory 2
A histogram creating means 3 and a histogram counter group 4, a candidate color extracting means 5, an initial integrating means 6, and an area having the smallest occurrence frequency of colors among the areas, and extracting the minimum frequency area as a color space. The number of regions is reduced to a predetermined number by combining the adjacent regions above with other regions and expressing the integrated region in one color, and the colors representing the resulting regions are indexed. First color map creating means 9 for creating a correspondence between all colors that can appear in the input image and index numbers of the color map in the color space division map 12 while registering in the numbered color map 11. , Combining the pair of areas with the smallest distance between the representative colors among the adjacent areas in the color space, and repeating expressing the integrated area with one color The number of areas is further reduced to a predetermined number of colors, and the colors representing the respective areas generated as a result are registered in the color map 11, and all colors that can appear in the input image are associated with index numbers of the color map. Second color map creating means 10 for creating a relationship in the color space division map 12, and a color for converting each pixel of the input image into an index number of the color map according to the color space division map 12 and storing it in the image memory 2. The image conversion means 13, the adjacency relationship table creating means 7, the adjacency relationship table 8, the display device 14, the adjacency relationship table 8, the color map 11 and the color space division map 12 are used as the first or second color map. In a color image conversion device equipped with a series of changeover switches 15 for changing over to any one of the creating means 9 and 10, one color image is initially input, After the processing by the initial unifying unit 6 is completed, the number of colors is reduced by interactively switching between the first color map creating unit 9 and the second color map creating unit 10 so that the color map and the color space division map are reduced. A color image conversion device characterized by creating and.

[Action]

In the present invention, the first color map creating means has an action of integrating the low frequency area into the higher frequency area, and the second color map creating means creates the areas having the adjacent representative colors. Has a function of integration. Therefore,
By providing the first and second color map creating means, the features of both means can be brought out.

Further, in another invention of the present invention, the selection means is
Since the second color map creating means can be switched and used, appropriate processing can be performed according to the type of the input image.

〔Example〕

 Hereinafter, the present invention will be described based on examples.

FIG. 1 is an overall configuration diagram of a color image conversion apparatus according to an embodiment of the present invention. The image input device 1 writes a color image from a TV camera, a magnetic disk or the like into the image memory 2 as M-bit data for each of R, G and B components. The histogram creating means 3 corresponds to step 32 of the conventional example, and scans all or some of the pixels of the input image stored in the image memory 2 to display a histogram, that is, a color occurrence frequency distribution on the histogram counter group 4. create. Candidate color extraction means
5, the initial integration unit 6, the adjacency relation table creation unit 7, and the color image conversion unit 13 perform the same processes as Step 33, Step 34, Step 35, and Step 38 of the conventional example, respectively. The adjacency relationship table 8 is the adjacency relationship table creating means 7
The table in the format shown in Fig. 12 created by is stored. The color map 11 and the color space division map 12 correspond to FIGS. 13 and 14, respectively. That is, the color map 11 is a correspondence table of index numbers and representative colors represented by R, G, B components, and the color space division map 12 is a correspondence table of all colors in RGB space and index numbers of the color map 11. Is. The first color map creating means 9 performs the same processing as the conventional frequency priority method shown in the flowchart of FIG.
Performs the same processing as the conventional distance priority method shown in the flowchart of FIG. The display device 14 displays the input image on the image memory 2 or converts the processing result represented by the index number on the image memory 2 into RGB values via the color map 11 and displays the RGB values. 15,15,15 are the adjacency relation table 8, color map 11 and color space division map
12 is a changeover switch for changing over to either the first or second color map forming means 9 or 10, and in the initial state, all the switches 15 are the first color map forming means 9
Or all switches 15 are connected to the second color map creating means 10 side. However, the switch 15 is not mechanical but software or electrical.

FIG. 2 is a system configuration diagram of the embodiment of FIG. 1, and the same reference numerals as those in FIG. 1 indicate the same or corresponding parts. The CPU 21 creates a histogram according to the control program stored in the program / working memory 22, extraction of candidate colors, initial integration, creation of an adjacency table, first color map creation means, second color map creation means, and color image creation. Converts and controls various devices.

FIG. 3 is a flow chart of this embodiment.
In step 37, the number of representative colors is reduced to a predetermined number by the first color map creating means 9, and then in step 37, the number of representative colors is further reduced by the second color map creating means 10 to obtain the color map 11 and the color. The space division map 12 is created. In this case, all the switches 15 in the overall configuration diagram of FIG. 1 are all connected to the first color map creating means 9 side in the initial state, and at the same time when the processing of step 36 is completed, all the second color map creating means is created. It will be connected to the means 10 side. Incidentally, in FIG. 3, steps 31 to 35, 38 correspond to the same reference numerals in the conventional example (FIG. 9). Further, steps 36 and 37 perform the same processing as the frequency priority method (FIG. 10) and the distance priority method (FIG. 11) of the conventional example, respectively.

FIG. 4 is also a flow chart of this embodiment. After the number of representative colors is reduced to a predetermined number by the second color map creating means 10, the number of representative colors is further reduced by the first color map creating means 9. By reducing the number, the color map 11 and the color space division map 12 are created. In this case, all the switches 15 in the overall configuration diagram of FIG. 1 are all connected to the second color map creating means 10 side in the initial state, and at the same time when the processing of step 37 is completed, all the first color map creating It will be connected to the means 9 side.

According to this embodiment, since the first and second color map creating means are connected in cascade, the first color map creating means 9 inputs a completely different color due to noise at the time of image input. Therefore, the color is extracted as a candidate color even though the frequency is low, and when a region is formed, the region is integrated with another region, and the second color map creating means 10
Then, the part that was originally painted with one color on the image plane is input as two or more colors that are relatively similar in color due to unevenness in the document or errors in A / D conversion, and these colors are used as the representative colors. Such regions can be integrated when different regions are formed. Then, in the method shown in FIG. 3 of the present embodiment, first, the first color map creating means 9 clusters the regions of low frequency to form the regions of large volume. This means that when a large number of similar colors having a low frequency appear, such as a natural scene image, these colors form one region. Therefore, it is possible to prevent the chain-reactive regions from being integrated when the adjacent colors are reduced by the second color map creating means 10. In other words, this method is suitable for processing natural scene images. On the other hand, in the method of FIG. 4, first, the regions having the representative colors that are close to each other are integrated by the second color map creating means 10, so that the color distribution exists as several clusters, and each cluster is formed. It is suitable for processing in the case of separation in the RGB space. Such an image is equivalent to an illustration or a cartoon drawn in several highly saturated colors.

FIG. 5 is an overall configuration diagram of another embodiment of the present invention. In the figure, 1 to 15 have exactly the same configuration as the same reference numerals in FIG. The switch switching unit 16 interactively inputs or automatically generates selection information indicating whether to use the first or second color map creating unit and the number of representative colors, and specifies by the selected unit. Instruct to perform area integration until the number of representative colors is reached. Here, the changeover switch 15 and the switch changeover means 16 constitute the selection means of the present application.

FIG. 6 is a system configuration diagram of the embodiment of FIG.
A keyboard 61 is added as an interactive input device to the system configuration diagram in the figure.

FIG. 7 is a flow chart of this embodiment, step 31.
.About.35 and 36 to 38 have the same configurations as the same reference numerals in FIGS. 3 and 4 which are flowcharts of the first embodiment of the present invention. Therefore, the same processing as in the first embodiment of the present invention is performed until the adjacency relation table 8 is created in step 35. Then, in this embodiment, after that, in step 71 and step 72, the selection information indicating which of the first or second color map creating means to use and the number of representative colors are interactively input, and the step which is the selected means 36 or step 37 is used to integrate the regions until the number of representative colors is reached. Then, in step 38, the RGB component data of each pixel of the input image on the image memory 2 is converted into the index number of the color map 11 using the color space division map 12, and the index number is written in the image memory 2. The converted image at this time is converted into RGB component values via the color map 11 and displayed on the display device 14. In step 73, the number of representative colors is reduced to a predetermined number,
It is determined whether or not all the processing is completed. If not completed, the processing from step 71 is repeated.

As described above, according to the present embodiment, since the color map creating means can be selected by the interactive means while watching the processing result, it is appropriate depending on the nature of the input image and the state of the image in the middle of processing. It is possible to perform various processes.

In the present embodiment, the interactive means is used as the switch switching means 16, but this can be done automatically. FIG. 8 is a flow chart of an embodiment in which the switches are automatically changed, and steps 31 to 35, 36 to 38 and 73
Corresponds to the same reference numeral in the flowchart of FIG. In step 81, initial values are given to the threshold value Th for the minimum frequency and the threshold value Td for the minimum distance. Step
Reference numeral 82 denotes switching of the switch 15. In the first loop, an instruction to perform processing corresponding to the initial state of the switch 15 is given. In the second and subsequent loops, the switch 15 is switched to the opposite side, and Instruct to perform processing corresponding to the state. Steps 83 and 84 are the updating of the threshold value, and the increment Δh (>
0) and Δd (> 0) are added. Step 36 is the first color map creating means, and if the minimum frequency exceeds Th or the number of representative colors finally becomes equal to the number of colors to be registered in the color map, the process of step 36 is ended. Step 37 is a second color map creating means, and if the minimum distance between the representative colors of the adjacent areas exceeds Td or the number of representative colors becomes equal to the number of colors to be finally registered in the color map, The process ends.

As described above, according to this embodiment, it is possible to automatically select the first and second color map creating means.

〔The invention's effect〕

As described above, according to the present invention, as the color map creating means, the first color map creating means based on the frequency priority method and the second color map creating means based on the distance priority method are provided. This is effective in preventing a noise-like color having a low frequency from being registered in the color map, and preventing a similar color of a plurality of colors from being registered in the color map.

Further, according to another invention of the present invention, by providing the above-mentioned one with a selecting means for selecting either the first or second color map creating means, the same effect as the above can be obtained, and the input image can be obtained. There is an effect that appropriate processing can be performed according to the type of.

[Brief description of drawings]

1 is an overall configuration diagram showing an embodiment of the present invention, FIG. 2 is a system configuration diagram of the embodiment of FIG. 1, FIGS. 3 and 4 are flow charts of the embodiment of FIG. 1, and FIG. FIG. 6 is an overall configuration diagram showing another embodiment of the present invention, FIG. 6 is a system configuration diagram of the embodiment of FIG. 5, FIGS. 7 and 8 are flowcharts of the embodiment of FIG. 5, and FIG. Is a flow chart of the conventional example, FIGS. 10 and 11 are flow charts of the color map creating means of the conventional example, FIGS. 12 (a) and 12 (b) are examples of the constitution of the adjacency relation table, and FIG. Configuration diagram, No.
Figure 14 is a block diagram of the color space division map. In the figure, 1 is an image input device, 2 is an image memory, 3 is a histogram creation means, 4 is a histogram counter group, 5 is a candidate color extraction means, 6 is an initial integration means, 7 is an adjacency relation table creation means, and 8 is Adjacency table, 9 first color map creating means, 10 second color map creating means, 11 color map, 12 color space division map, 13 color image converting means, 14 display device, 15 A changeover switch, 16 is a switch changeover means which constitutes a selecting means together with the changeover switch 15, 21 is a CPU, 22 is a program working memory, and 61 is a keyboard. In the drawings, the same reference numerals indicate the same or corresponding parts.

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical display location G09G 5/06 8121-5G (72) Inventor Satoru Tomita 8-1-1 Tsukaguchihonmachi, Amagasaki City, Hyogo Prefecture Issue Sanryo Denki Co., Ltd. Applied Equipment Research Laboratory (56) References 1985 IEICE General Conference, 1173 "A Study on Colormap Creation" Published March 5, 1985

Claims (2)

[Claims] 1. A color image input device 1, a color image memory 2, a histogram creating means 3 and a histogram counter group 4, a candidate color extracting means 5, an initial integrating means 6,
By extracting an area having the smallest occurrence frequency of color from each area, integrating the area having the smallest frequency with other areas adjacent in the color space, and repeating the expression of the integrated area in one color The number of areas is reduced to a predetermined number of colors, and the colors representing the respective areas generated as a result are registered in the color map 11 with index numbers, and all colors and color maps that can appear in the input image are displayed. The first color map creating means 9 for creating the correspondence with the index number in the color space division map 12 and the area pair having the smallest distance between the representative colors among the areas adjacent in the color space are integrated. At the same time, the number of regions is reduced to a predetermined number of colors by repeatedly expressing the integrated region with one color, and the colors representing the respective regions generated as a result are registered in the color map 11. Together with a second color map creating means 10 for creating a correspondence relationship between all colors that can appear in the input image and index numbers of the color map in the color space division map 12, and each pixel of the input image is in the color space. A color image conversion means 13 for converting into an index number of a color map according to a division map and storing it in an image memory, an adjacency relation table creating means 7, an adjacency relation table 8, a display device 14, the adjacency relation table 8, color In a color image conversion apparatus equipped with a series of changeover switches 15 for changing the map 11 and the color space division map 12 to either the first or second color map creating means 9 or 10, one color image is initially input. After the processing by the initial integration means 6 is completed, the processing by the first or second color map creation means 9 can be summarized. The process continues until the number of colors reaches a predetermined value according to the defined threshold value, and the number of colors is reduced by the second or first color map creating means 10 to create a color map and a color space division map. A color image conversion device characterized by: 2. A color image input device 1, a color image memory 2, a histogram creating means 3 and a histogram counter group 4, a candidate color extracting means 5, an initial integrating means 6,
A region is extracted by extracting a region having the smallest occurrence frequency of color from each region, integrating the region having the smallest frequency with other adjacent regions in the color space, and expressing the integrated region in one color. The number of colors is reduced to a predetermined number, and the colors representing the respective regions generated as a result are registered in the color map 11 to which index numbers are added, and all colors that can appear in the input image and the index of the color map. The first color map creating means 9 for creating the correspondence with numbers in the color space division map 12 and the area pair having the smallest distance between the representative colors among the areas adjacent to each other in the color space are integrated. , The number of areas is reduced to a predetermined number of colors by repeatedly expressing the integrated area with one color, and the colors representing the respective areas generated as a result are registered in the color map 11. Both, second color map creating means 10 for creating a correspondence relationship between all colors that can appear in the input image and index numbers of the color map in the color space division map 12, and each pixel of the input image in the color space. A color image conversion means 13 for converting into an index number of a color map according to the division map 12 and storing it in the image memory 2, an adjacency relation table creating means 7, an adjacency relation table 8, a display device 14, and the adjacency relation table 8 ,
In a color image conversion device provided with a series of changeover switches 15 for changing over the color map 11 and the color space division map 12 to either of the first or second color map creating means 9 and 10, one color image is initialized. After the input and the processing by the initial integration means 6 are completed, the number of colors is reduced by interactively switching between the first color map creation means 9 and the second color map creation means 10 to reduce the color map and color. A color image conversion device characterized by creating a space division map.