JP3287016B2 - Image processing device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus which outputs full-color image data to an output device having a low gradation expression capability or converts the data into a smaller number of gradations for the purpose of reducing the data capacity. .

[0002]

2. Description of the Related Art A typical form of full-color image data is one in which a storage capacity of 8 bits is assigned to each color component of red R, green G, and blue B of each pixel. Each color component of this image has 256 levels of brightness,
Express up to about 16.78 million colors (256 to the power of 3).
Such full-color image data can be easily obtained by optically reading a document image using a scanner or the like. On the other hand, CR of personal computer
At T, it is rare that such a full-color image can be displayed as it is. As a typical example, a 4-bit image storage memory is provided for each pixel, and 16 colors (2
(4th power) can be displayed, but with regard to the 16 colors to be used, a system has become widespread in which any 16 colors can be selected from a total of 4096 colors of 16 gradations of R, G, B colors.

When it is desired to display a full-color image consisting of a maximum of about 16.78 million colors on a CRT capable of outputting only 16 colors at the same time, the number of gradations of full-color image data is reduced and converted so that the data falls within 16 colors. Is required. As a method of converting the number of gradations, R and B components are allotted one bit at a time and 2 bits each.
There is a method in which gradation is displayed, and only G is assigned to 2 bits and displayed in four gradations. In this case, the reason why the four gradations are assigned only to G is that the number of reproduction gradations of G having the largest influence is increased by utilizing the property that the resolution of the human eye is higher in the order of G, R, and B. It is.

[0004]

However, although the resolution of the human eye is considerably lower for the B component than for the other two components, the difference between the R and G components is not so large.
The above-described gradation number conversion method has a problem in that even if there are four gradations of the G component, the number of gradations of the R component is only two. In this regard, a system in which each of R and G has three gradations and B has two gradations can improve the image quality. In this case, however, 18 colors (3 × 3) are used.
× 2 = 18), which is not possible with a system that can display only 16 colors.

The present invention solves such a problem, and an object of the present invention is to obtain a high-quality color image with a small number of colors used.

[0006]

[MEANS FOR SOLVING THE PROBLEMS] Less image data
In an image processing device that converts the number of tones,
Of the other color component b in accordance with the gradation number conversion result of the color component a.
Characterized by having a conversion instruction means for determining the number of conversion tones.
This is an inkjet printing device that performs
Image processing method to convert data to smaller number of gradations
Extracting the color component a of the target pixel;
Taking out a certain color component b;
Converting the number a into a number of gradations;
A process for determining the number of gradations of the color component b according to the gradation value of the component a
And characterized in that:

[0007]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the image processing apparatus according to the present invention will be described below in detail.

FIG. 1 is a diagram showing a first embodiment of the present invention. It consists of three primary color components of R, G and B, and each color 25
Image data having 6 tones is provided with a 4-bit image storage memory for each pixel, and 16 colors can be displayed at the same time. For the 16 colors, each of R, G, and B colors is 0 to 1
CRT of personal computer that can select any 16 colors from a total of 4096 colors with 16 gradations up to 5
It is an example of the case of displaying above.

In this embodiment, a G component is used as a color component a and an R component is used as a color component b.
The number of gradations is converted into gradations, and the number of converted gradations of the R component is determined to be 2 or 3 according to the result, and the number of gradations is converted.

First, the G data 101 is input to the G gradation number conversion means 102, and its output Gx103 is output to G1, G2,
G3 (here, G1 <G2 <G3) is ternarized to one of three values. Here, it is assumed that G1 = 0, G2 = 8, and G3 = 15.

Next, the number of conversion gradations of the R data 106 is determined. The conversion instruction means 104 determines that the value of Gx103 is G1 or G
In the case of 3, the conversion gradation number of R becomes 3 and the value of Gx becomes G
When the number becomes 2, a conversion instruction signal 105 for instructing to make the number of conversion gradations of R two is output.

The R gradation number conversion means 107 converts the conversion instruction signal 1
05, and if the R conversion gradation number is 3, the R data 1
06 is ternarized to any one of R1, R2, and R3 (here, R1 <R2 <R3), and when the number of R conversion gradations is 2, R
Binarize to either 1 or R3, and convert the result to Rx10
Output as 8. Here, it is assumed that R1 = 0, R2 = 8, and R3 = 15.

The B data 109 is binarized by the B gradation number conversion means 110 into one of two values B1 and B2 (here, B1 <B2) and output as Bx111. Here, B1 = 0 and B2 = 15. In the present embodiment, the gradation number conversion of the B data is R,
Performed independently of G data.

When R and G are ternarized and B is binarized by an ordinary method, 3 × 3 × 2 = 18 and a maximum of 18 colors are used. In the example, when the conversion result of G is G2, R is not converted to R2, so that the conversion result is (R2, G2, B
No two kinds of combinations such as 1) or (R2, G2, B2) appear, and 18-2 = 16, and only 16 colors are used at the maximum.
Display can be performed on a display system that can simultaneously display only colors.

In this embodiment, when G is converted to G2, the number of converted gradations of R is set to 2. This is because only when G is converted to G1, the number of converted gradations of R is set to 2. Or a configuration in which the number of converted gradations of R is set to 2 only when G is converted to G3.

In this embodiment, after the G data is converted into a ternary value,
The R data is binarized according to the result,
The relationship between R and R may be reversed and R may be ternary first, and then G may be binarized. If the original image data is an image including more red halftone areas than green, the image quality after gradation number conversion is improved.

FIG. 2 is a diagram showing a second embodiment of the present invention. It consists of three primary color components of R, G and B, and each color 25
Full-color image data having a maximum of about 16.7 million colors at 6 gradations can be converted into any 256 colors from about 16.7 million colors.
This is an example in the case of performing gradation number conversion for outputting to an output device capable of simultaneously outputting colors. In this embodiment, not only one color component is used as the color component a, but also the R component and the G component are considered, and the conversion gradation number of the B component is determined by a combination of both gradation number conversion results. I do. In this embodiment,
In particular, a case where it is desired to improve the tone reproducibility of the skin color region will be described.

Now, in the output device for outputting the number-of-tones conversion result of the present embodiment, R and G of color data which can be called flesh color are
As a result of examining the components, R was mainly from 140 to 16
It is assumed that 0 and G are in the gradation range of 40 to 80.

The G data 101 is converted into G gradation number conversion means 102
, And its output Gx103 is G1, G2, G3, G
The gradation is converted into six values to any one of 4, G5 and G6. Here, it is assumed that G1 = 0 G2 = 40 G3 = 80 G4 = 130 G5 = 190 G6 = 255.

The R data 106 is converted into R gradation number conversion means 107.
, And its output Rx108 is R1, R2,.
, R8 is converted into eight gradations. Here, R1 = 0 R2 = 40 R3 = 80 R4 = 120 R5 = 160 R6 = 190 R7 = 220 R8 = 255.

Next, the conversion instruction means 104 for determining the number of conversion gradations of the B data 109 includes Rx 108 and Gx 103
In response, the number of gradations of B is determined.

First, when Rx 108 is R4 or R5 and Gx 103 is G2 or G3, it is determined that this color may be a color of a flesh-tone area, and the number of gradations of B is increased to a higher sixth level. Determine the key. The decision result is the conversion instruction signal 10
5 is transmitted to the B gradation conversion means 110, and the B gradation number conversion means 110 converts the B data 109 into B1, B2, B3, B
The value is converted into any one of 4, B5, and B6 into six gradations and output as Bx111. Here, B1 = 0 B2 = 50 B3 = 100 B4 = 160 B5 = 190 B6 = 255.

The conversion result of the number of gradations of R and G, Rx10
If the combination of 8 and Gx103 is other than the above, it is determined that the possibility of the skin color area is low, the number of gradations of B is determined to be a smaller four gradations, and any of B1, B3, B4, and B6 is determined. The value is converted into four gradations.

The number of colors used by the above conversion method is 2 × 2 × 6 + (6 × 8−2 × 2) × 4 = 200, and the maximum is 200 colors, and 56 colors are used for other applications. In the flesh color area, the B component has been successfully provided with a smooth gradation characteristic while having a margin.

In the above embodiment, the conversion result of the color component a is divided into two cases and the conversion gradation number of the color component b is switched to two. The gradation number conversion result of a may be divided into three or more types, and the number of conversion gradations of the color component b may be switched to three or more accordingly. For example, in the second embodiment, as the third case, particularly when Rx108 is R4 and Gx103 is G2,
It is also possible to increase the number of conversion gradations of B to 8 gradations.

FIG. 3 illustrates a third embodiment of the present invention. In this embodiment, two colors to be removed from 18 colors are optimally added to the first embodiment according to the image data by adding a color distribution checking means for checking the color distribution of the original image data in advance. It is to try to.

First, the color distribution examination means 302 examines G data 3 by extracting original image data by appropriate means.
01 and the estimated distribution of the G component. 1. the number Glo of pixels in a low-luminance level color region where G is 0 ≦ G <64; 2. the number of pixels Gmid in a color region of an intermediate luminance level where G is 64 ≦ G <192; The number of pixels Ghi in the high-luminance level color region where G is 192 ≦ G <256 is counted. Now, G data is G1, G2, G3
(Here, G1 <G2 <G3), the number of gradations can be converted into any one of three values, G1 = 0, G2 = 8, G3 = 15
And However, 15 is the maximum luminance level.
The color distribution examination means 302 compares Glo, Gmid, and Ghi and examines which color region is estimated to have the smallest number of pixels. First, an example in the case where Gmid is minimum will be described. At this time, it is possible to estimate that the probability that G will be converted to G2 is the lowest. Therefore, the color distribution investigation unit 302 determines that the conversion result of the number of gradations of G is
When G2 is reached, the number of convertible gradations of R is changed to two, and when the result of conversion of G is G1 or G3, the number of convertible gradations of R is changed to three. Is set, and the result is transmitted to the conversion instruction unit 104. So far,
This is the stage of color distribution investigation, and the subsequent tone number conversion work is exactly the same as in the first embodiment.

In the above, among Glo, Gmid and Ghi, Gm
Although the case where id is minimum has been described, when Glo is minimum, the color distribution investigation unit 302 sets the number of converted gradations of R to two gradations of R1 and R3 only when G105x becomes G1. And the conversion result is (R2, G1, B1) or (R
2, G1, B2) should not appear. Similarly, if Ghi is minimal, Gx
Only when 105 becomes G3, the number of R conversion gradations is determined to be two gradations, and the conversion result is (R2, G3, B1) or (R2,
G3, B2) should not appear.

The ratio of pixels ternarized to the respective values of G1, G2, and G3 is close to the ratio of Glo, Gmid, and Ghi. Thus, as described above, the color component a G
, Only in the gradation region that can be estimated to be the least frequent,
By reducing the number of converted gradations of R, which is the color component b, to two, it is possible to reduce the number of colors used while minimizing deterioration of image quality.

In this embodiment, G is first converted into three gradations as a color component a, and R is converted into a color component b by two or three colors.
Although the gradation is previously determined, the color distribution investigating unit investigates a plurality of color components and determines colors to be actually assigned as the color components a and b according to the result. Is also possible. For example, the color component investigating means includes: 1. the number of pixels Rmid in a color region of an intermediate luminance level where R is 64 ≦ R <192; The number of pixels Gmid in the color region of the intermediate luminance level where G is 64 ≦ G <192 is checked, and according to the result, If Rmid ≦ Gmid, the halftone expression of the G component is given more importance than R, and after G is ternarized to one of G1, G2, or G3, R is determined according to the result of converting the number of gradations of G. , 2
Or ternary conversion 2. If Rmid> Gmid, the emphasis is placed on the halftone representation of the R component over G, and after R is ternarized to any of R1, R2, and R3, G is determined according to the result of conversion of the number of gradations of R. , 2
Or ternarization.

FIG. 4 is a diagram showing a fourth embodiment of the present invention, which is obtained by adding color distribution checking means for checking the color distribution of original image data in advance to the second embodiment.
1 is converted into any of the values of G1, G2, G3, G4, G5, G6 into six gradations, and the R data 106 is represented by R1, R2,.
The same applies to the case where eight gradations are set to any value of R8.

As described in the second embodiment, in the output device of this embodiment, the color data which can be called flesh color has R of 140.
To 160 gradation areas, G is 40 to 80 gradation areas,
B is assumed to be a gradation region of 30 to 70. Therefore, the color distribution investigating means 302 examines all the pixels of the original image data and extracts the G data for inspection 30 created by pixel extraction and the like.
1. Investigate the investigation R data 401 and the investigation B data 402, and estimate the ratio of the skin color data.

If the ratio of the flesh color data is equal to or greater than the fixed value Ph, the gradation number conversion of this image is performed in the flesh color emphasis mode, and otherwise, the normal mode is set. Here, it is assumed that Ph = 0.05 (5%). First, the case where the skin color priority mode is entered will be described. At this time,
The conversion instructing means 104 works in the same manner as in the second embodiment, in which R is converted to R4 or R5 in gradation number, and G is G
When the number of gradations is converted to 2 or G3, B is changed to B1, B2, B
In the case where R, G, and R are converted to a different number of gradations by any other combination, B is set to B, B4, B5, or B6.
The correspondence is determined so that one of the values of 1, B3, B4, and B6 is converted into four gradations.

In the normal mode which is not the skin color emphasizing mode,
The emphasis is not so much placed on the skin color, and the operation is different from that of the second embodiment. The conversion instructing means 104 converts R into any of R6, R5, and R4, and outputs G as G2,
When the gradation number is converted to either G3 or G4, B becomes B
The correspondence is determined so that any one of 1, B2, B3, B5, and B6 is converted into five gradations. If the combination of R and G is other than that, B is determined to be converted into any of B1, B3, B5, and B6 in four gradations. The number of colors used in this correspondence is 3 × 3 × 5 + (6 × 8−3 × 3) × 4 = 201.

After the correspondence is determined as described above,
The gradation number conversion process for each color component is the same as in the second embodiment. According to the conversion method of this embodiment, the number of colors used is a maximum of 200 colors in the skin color emphasis mode and a maximum of 201 colors in the normal mode, and the gradation is optimized according to the color distribution of the original image data with substantially the same number of colors used. Number conversion can be performed.

By the way, in all of the first, second, third and fourth embodiments, the details of the number-of-tones conversion method such as binarization, ternary conversion, and hexadecimal conversion have not been described. A simple gray scale number conversion method may be used.

For example, in order to simply convert the gradation data D into N gradations (N is an integer of 2 or more), N types of output data D
1, D2,..., DN (where Dk <Dk + 1 and k is an integer of 1 ≦ k ≦ N), and N−1 threshold strings S1, S
2,..., SN-1 (however, Sk <Sk + 1) is set, and if SN-1 <D≤SN, the data D is compared with the threshold value sequence as in DN as shown in FIG.
What is necessary is just to make gradation.

Further, if dither noise X is added to the data, and if SN-1 <D + X≤SN, N gradation may be performed as in DN. This method is generally called a dither method, and is effective in improving the gradation expression ability. If Y = -X and SN-1 + Y <D.ltoreq.SN + Y, this is equivalent even if it is considered that the dither noise Y is added not to the data but to the threshold value like DN. Dither noise X
There are various methods of generating the image. For example, when X is random noise, a random dither method is used. It becomes a typical dither method. In addition, taking into account the influence of an error generated when the surrounding pixels are converted into N values as X,
When a value that cancels the average error is used, it is called an average error minimum method or an error diffusion method. The average error minimum method corrects the data value of the next pixel with a weighted average value of the error of a plurality of pixels. In the error diffusion method, an error generated in a certain pixel is diffused to a plurality of subsequent pixels. The average error minimization method and the error diffusion method are completely equivalent except for handling at the image edge. As the error diffusion method, there are methods such as JP-A-03-18177, JP-A-03-34767, JP-A-03-73670, and JP-A-03-80767.

In the present invention, the pixel in which the color component a has been converted to a specific value may be disadvantageous in that the number of convertible gradations of the color component b is reduced. By using the error minimization method or the error diffusion method, this disadvantage can be greatly reduced. As an example, G in the first embodiment
Tone number converter 102 and R number converter 107
Considers the case where the number of gradations is converted using the error diffusion method.

Now, both the G data and the R data of the original image are 0.
Let us consider a case where there is an area near 64 out of ~ 255. The converted data has a scale range of 0 to 15, and 64/255 is close to 4/15. Therefore, it is ideal that both R and G are converted to data near 4.

First, since the G component can be ternarized to any one of 0, 8, and 15, the gradation level is usually converted to 0 for about half the pixels and to 8 for the remaining pixels. Next, the R component is
0, 8, 1 where the G component is converted to 0
Although it can be ternarized to any of 5, G can only be binarized to 0 or 15 where G is converted to a gray level of 8. For this reason, in a simple probability calculation, R is converted into a gradation level of about 1/8 pixel to 15, about 1/4 pixel to 8, and about 5/8 pixel to 0. , The tone level is converted to 15 which is considerably different from the ideal value of 4, and that portion may look like noise.

However, when the error diffusion method is used, the error is accumulated and becomes large, and G is set to 0 in the vicinity of the pixel before the situation where R is converted into the maximum number of gradations of 15 is reached.
If there is a pixel that can convert R to 8,
In most cases, R is converted to 8 and the error is eliminated. This is because the error diffusion method works to minimize the average error in consideration of the neighboring pixels. Therefore, the probability that R is converted to 8 in a pixel whose G is converted to 0 is larger than a simple calculated probability value, and is converted to 15 having a large difference from the ideal value of 4. Becomes smaller conversely. As a result, smoother gradation number conversion is performed.

In the above embodiment, examples of the color components include:
R, G, B additive color mixture of three primary color components was used.
Color component systems of three primary colors of subtractive color mixture of cyan, magenta, and yellow; four color component systems of black added to cyan, magenta, and yellow; CIE's XYZ color system and L * a * b *
Which color component of any color system, such as a color system, is a color component a,
The color component b may be subjected to the gradation number conversion of the present invention. Also, the output device is not limited to the CRT, and a hard copy device such as a printer may be used. Also, in FIG. 1 to FIG. 4, the R gradation number conversion means 107, the G gradation number conversion means 102, the B gradation number conversion means 110, and the like are illustrated as being separate, but physical May be configured such that the same gradation number conversion means is used by switching.

[0044]

According to the configuration of the present invention, since there is provided conversion instructing means for determining the conversion gradation number of another color component b for each pixel according to the gradation number conversion result of a certain color component a, A color conversion result with high quality image quality is obtained while using a small number of colors. For example, in a system capable of simultaneously displaying 16 colors, an image quality close to that obtained when 18 colors of R and G are used for 3 gradations and B for 2 gradations is obtained.

Further, when the conversion result of the gradation number of the color component a is a specific color area, the conversion gradation number of the color component b can be increased. Etc., the reproducibility of a specific important color region can be improved.

Further, the number of colors is reduced by optimally setting the correspondence between the gradation number conversion result of the color component a and the convertible gradation number of the color component b based on the investigation result by the color component investigation means. The deterioration of the image quality due to this is minimized.

Further, by using an error diffusion method or an average error minimization method as the number-of-tones conversion means, the image quality deterioration due to the reduction in the number of colors used according to the present invention can be extremely reduced. This is because the mechanism for minimizing the average error by considering the neighboring pixels of the error diffusion method or the average error minimization method and the mechanism for switching the number of conversion gradations by the conversion instruction means of the present invention are effectively cooperated. To work.

[Brief description of the drawings]

FIG. 1 is a diagram showing a first embodiment of the present invention.

FIG. 2 is a diagram showing a second embodiment of the present invention.

FIG. 3 is a diagram showing a third embodiment of the present invention.

FIG. 4 is a diagram showing a fourth embodiment of the present invention.

[Explanation of symbols]

 101 G data 102 G gradation number conversion means 104 Conversion instruction means 106 R data 107 R gradation number conversion means 109 B data 110 B gradation number conversion means 302 Color distribution investigation means

Claims (4)

(57) [Claims] In an image processing apparatus for converting image data into a smaller number of gradations, a conversion gradation number of another color component b is determined according to a gradation number conversion result of a certain color component a of a target pixel. An image processing apparatus comprising a conversion instruction unit. 2. An image processing apparatus according to claim 1, further comprising color distribution checking means for previously checking a color distribution of the original image data, and based on the result, a conversion result of the number of gradations of the color component a and a color component. An image processing apparatus for setting correspondence between b and the number of convertible gradations. 3. The image processing apparatus according to claim 1, wherein an error diffusion method or an average error minimization method is used as a method of converting the number of gradations. 4. The image data is converted into a smaller number of gradations.
In the image processing method, the color component a of the target pixel is extracted.
A step of extracting a color component b of the pixel of interest, a step of converting the extracted color component a into a number of gradations, and a step of converting a color component according to the gradation value of the color component a having the number of gradations converted.
determining the number of gradations of b .