JPH02226869A - Color decomposition picture correction method and device - Google Patents

Abstract

PURPOSE:To improve the consistency at a boundary face in a split space even when the relation of input and output is nonlinear by using a matrix coefficient decided at every split space and applying matrix arithmetic operation, thereby outputting a color decomposition picture data. CONSTITUTION:A matrix coefficient decided at every triangular pyramid is stored in lookup tables(LUT) 15-18. When the input signals R, G, B are supplied, the LUT 15-18 supply output signals corresponding to the triangular pyramid with the input signals B, G, R included therein to multipliers 20-22 and an adder 224 respectively and Y, M, C signals are outputted sequentially from an output terminal 26 leading from an adder 25 every time one of the input signals B, G, R are supplied. Since the matrix coefficient is used to apply matrix arithmetic operation and the color decomposition picture data subjected to color correction is obtained, even when the relation of input and output is nonlinear, the consistency with the boundary faces in the split spaces is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION Field of Industrial Use] The present invention relates to a color separation 1i color correction method and apparatus suitable for use in color correction of video printers, digital color copies, and the like.

[Background of the Invention] When making a hard copy of a television image using a video printer, digital color copying device, etc., each color system is different, so in order to match the reproduced colors, color separation III image correction is necessary. is used.

As is well known, a color masking device, which is one type of color separation image correction device, is a device for reproducing correct colors by canceling sub-absorption of color materials such as toner and ink.

For example, from the color masking IIIO shown in FIG. 5, new color correction data (W1 image data after color correction, in this example, yellow Y , magenta M, cyan C
image data) is output.

In this case, the color correction data is input ii! It can be obtained by calculating the iil data using equation 1.

For example, color correction data is obtained by dividing the color space formed by input image data into multiple basic grids (cubes) and performing matrix operations on the input image data using matrix coefficients that are not determined for each basic grid. It has been proposed (see Japanese Patent Laid-Open No. 61-60068).

In addition, the color correction data is stored in a lookup table (LU
It is conceivable to store the color correction data in R, .
When G and B image data are each 8-bit data, the combination of each image data is 2s 28
- 2B = 224, and in order to obtain three color I-correct data, a huge memory capacity of 2" x 3 = 50.3 Mbytes is required.

Therefore, in order to reduce the memory capacity, the applicant has
Divide the color space formed by input stroke data into a plurality of basic grids (cubes), 1. , UT stores color correction data for the combination of input image data located at the vertices, and when there is no color correction data corresponding to input III image data, this input image data (interpolation point) is included. It was proposed to obtain correct color data by weighted averaging of color correction data of the vertices of the basic lattice (patent application No. 6).
3-314636 sound N).

Furthermore, in order to shorten the interpolation processing time, we decided to obtain color correction data by weighted averaging of the color correction data of the vertices of triangle 1), which is a divided space with the minimum number of vertices obtained by dividing the basic grid ( Patent Application No. 1983-238507 N).

The issue reference is trying to solve! fl] By the way, color correction data for the combination of input image data located at the vertices of the basic grid is stored in the LUT in advance,
When color correction data is obtained by weighted averaging with reference to this LUT, the input image data is divided into the upper N bits (representing the reference point of the basic grid that includes the interpolation points determined by the input image data) and the lower When storing the color correction data of the vertices of the basic lattice in the LUT separated into M bits (representing the position of the interpolation point in the basic lattice), each color has (2'+1)
For example, when input image data is 8 bits and is separated into upper 4 bits and lower 4 bits, 0, 16. 32. ..., 224,240,2
Color correction data for 56 vertices is stored in the LtJT, requiring <24 + 1) = 17 addresses for each color.

However, since memory generally has a capacity of bytes that is a power of 2, it is inefficient to implement it in hardware as it is, and the memory cannot be used effectively.To avoid this, it is necessary to store large or small digital values. ) is rounded to the nearest appropriate value.0 For example, in the upper tide, values 241 to 255 are treated as 240 and processed using 24 addresses. However, this method has the disadvantage that errors occur due to rounding processing.

On the other hand, the color space formed by the input image data is 3j1
According to methods that obtain color correction data by dividing input image data into a number of basic grids and performing matrix operations using matrix coefficients determined for each basic grid, there are methods that use LUTs as described above. However, if the relationship between input and output is nonlinear, the coefficients of the matrix operation change at the boundary of the divided basic grid, so if there is a smooth color change such as a gradation, it will pass through the boundary. There was a drawback that the change in time was discontinuous.

Therefore, the present invention aims to eliminate the above-mentioned drawbacks.

[Means for Solving Eight Problems] The color separation image correction method according to the present invention minimizes the color space formed by n color separation image data (n is an integer of 2 or more) that can be input for color correction. The color-corrected color-separated image data can be obtained by dividing the input color-separated image data into divided spaces with a number of vertices of This is a characteristic feature.

Further, the color separation image II correction device according to the present invention transforms a color space formed by n color separation image data (n is an integer of 2 or more) that can be inputted for color correction into a divided space with a minimum number of vertices. matrix coefficient storage means for storing the matrix coefficients determined for each divided space; and matrix coefficients for reading out the matrix coefficients of the divided space including the input n color separation image data from the matrix coefficient storage means. It is characterized by comprising a readout means, and a matrix calculation means for performing a matrix calculation on input 11 color separation image data using matrix coefficients read from the storage means and outputting color-corrected color separation image data. It is something to do.

[Operation] In the above configuration, the color space formed by the n-color separation image data is divided into divided spaces having the minimum number of vertices. For example, when n=3, it is divided into pentagonal pyramids.

When color-separated image data is obtained by matrix calculation using the seven matrix coefficients determined for each divided space, consistency at the boundaries of the divided spaces is improved even if the relationship between input and output is nonlinear.

[Example] Hereinafter, the color separation image correction method and Waa according to the present invention will be described.
An example will be explained.

In this example, blue B. In this example, signals of green G9 and red R are input, and one word of yellow Y, magenta M, and cyan C is output. In this case, the three-dimensional color space formed by B, G, and F't is divided into a divided space with the minimum number of vertices, that is, a pentagonal pyramid, and the input signals B, G,
R is subjected to matrix operation using the matrix coefficients determined for each pentagonal pyramid, and one output word Y, M,
C is obtained.

First, the principle of this example will be explained. B.G.

Assume that the two spaces R, Y, M, and C are each divided into pentagonal pyramids, and the vertices of the pentagonal pyramids correspond to each other as follows (Figure 1).

B, G, R'! ! ! m Y, M, C empty rlflPI
(81, Gl, R1) PI' (Yl,
Ml. CI) R2 (82, G2. R2) P2'
(Y2゜R2,C2)R3(83,G3.R
3) P3' (Y3. R3, C3)
R4 (84, G4. R4) P4' (Y4.
R4, C4) R5 (85, GS, R5) P
5' (Y5. R5, C5) At this time, B,
Assume that two pentagonal pyramids in G and R spaces that touch each other are formed using the vertices of Pl, R2, R3, and R4 and Pl, R2, R3, and R5, respectively.

In this case, the vertices Pi R2P of B, G, R space
Y, M, corresponding to the pentagonal pyramid formed by 3P4,
(including the points in the pentagonal pyramid formed by the vertices Pl'P2'P3'P4' of the C space) can be calculated using the following matrix formula.

This equation (1) is derived as follows.

That is, point Pi (B
i.

Gin R+ ) and point P in the space of Y, M, and C.
The relationship with i' (Yi0M1i Cj) can be expressed by the following matrix operation.

At this time, there are 12 unknowns aiJ, so the four-dimensional simultaneous 1
This aIJ can be determined using the following equation.

Therefore, the triangular pyramid in B, G, and R space has the vertex PIP
When it is formed by 2 F3 F4,...
・ (3) Then, ・ ・ ・ (4) By substituting equation (4) into equation (2) and rewriting it in the form of matrix operation, equation (1) is derived.

By the way, Pi (Yi, Mi, Ci) is B.

On the surface (triangle) where two triangular pyramids touch in G and R space, it is necessary to obtain the same value no matter which matrix coefficient of the triangular pyramid is calculated, and formula (1) satisfies this condition. This will be explained next.

That is, in the example shown in the mi diagram, the triangle PIF2P3 is the surface where the two triangular pyramids touch. At this time, the triangle Pi
The point on F2 F3 is the triangle mIPIP2 F3 F4
It will be shown below that the same result is obtained no matter which matrix coefficients of the triangular pyramid PI F2 F3 F5 are used.

Point P on triangle PI F2 F3 is: B=α81+βB2+γB3 G=αG! +βC2+γG3 R=αR1+βR2+γR3. Substituting this into equation (1) and expanding it yields, and the term (Y4.M4°C4) disappears. This is Y4. M4. C4 to Y5. It is the same even if MS and C5 are replaced, and it is proven that the celestial point of the triangular pyramid calculated by equation (1) will have the same value no matter which matrix coefficient of the triangular pyramid is used for calculation. .

In this example, first, input signals B, G, and R are separated into upper bits and lower bits, and input signal B is obtained.

The color space formed by G and R is divided into basic grids (cubes). In this case, the basic lattice is specified by the upper bits.

Next, the lower bits of input signals B, G, and H are compared, and it is determined which of the six triangular pyramids created by further dividing the basic lattice is included.

That is, as shown in FIG. 2, a total of six pentagonal pyramids are formed by one-dot chain lines for the basic lattice composed of vertices a to h. The coordinates (X,
Y, Z') are respectively shown in FIG. The conditions for specifying in which of the pentagonal pyramids of 61il the lower bits of input signals B, G, and H are included are BL and OL, respectively.

If it is RL, it will be as shown in FIG. In other words, a pentagonal pyramid is specified depending on whether this condition is satisfied.

Next, according to equation (1), input signals B, G,
Output signals Y, M, and C corresponding to H are calculated.

,', V= cllB+ c12G+ c13R+
cX4M= c21B+ c22G+ c23R+
c24C= c31B+ c32G+ c33R+ c
34. . . (7) Here, the matrix coefficients all to c34 are determined in advance for each pentagonal pyramid.

From equation (7), Y, M, and C are calculated by three multiplications and three additions for each color.

FIG. 4 shows input signals B and G as described above.

This is an example of a color masking device 10 that obtains output signals Y, M, and C corresponding to R.

The 8-bit input signals B, G, and R supplied to terminals 118.11G and IIR are separated into upper 4 bits and lower 4 bits, respectively. In this case, the upper 4 bits represent the basic lattice (cube) containing the input signals B, G, and R, and the lower 4 bits represent the position within the basic lattice.

The lower 4 bits of signals BL, G1 . ．． RL is selectively supplied to comparators 12-14. Comparator 12
Then, whether BL≧GL or not, comparator 13 determines CL
The comparator 14 determines whether RL≧BL, and when it is satisfied, a high level "1" signal is output, and when it is not satisfied, a low level "θ" signal is output. The input signals B, G, and R are determined by the 3-bit signals from these comparators 12 to 14.
For example, when Nal is included in the pentagonal pyramid of Nal shown in FIG. 3, the outputs of comparators 12 and 13 will be high. Level “1” The output of comparator 14 is low 1
The novel becomes "0", and the input signals B and G.

It can be seen that R is included in a pentagonal pyramid with positive].

The 3-bit signals from these comparators 12-14 are supplied to LUTs 15-18. Also, these Lt
For JT 15 to 18, the upper 4 bit signals BH, GH,
R)l is supplied. Here, the top 4 bits No. 18B
11. G)1. RH represents the basic lattice as mentioned above,
Furthermore, the 3-bit signals from the comparators 12 to 14b) represent the pentagonal pyramid in the basic lattice as described above, and therefore, these upper 4-bit signals B11 . G11.
The 3-bit signals from RH and comparators 12 to 14 specify which pentagonal pyramid of the basic lattice the input signals B, G, and R are included in.

Matrix coefficients are stored in LUTs 15-18.

That is, LUT 15 stores coefficients of cll, c21°c31, and LUT1B stores coefficients of c12. c”22*
The coefficient of c32 is stored, and LUT 17 contains the coefficient of c13°c23. c 33 coefficients are stored, LUTlB
CI4+ c24. c34 is stored. In this case, the matrix coefficients are determined for each pentagonal pyramid, and each of the coefficients ell to c34 has 6 types x the number of basic lattices.

Further, these LUTs 15 to 18 are connected to Y from terminal 19.
, M, and C color cutoff signals are supplied.

When one input signal B, G, R is supplied, that input signal B is output from LUTs 15 to 18.

Corresponding to the pentagonal pyramid containing G and R, the coefficients ell to c14 related to Y are first referenced, and then the coefficient c2 related to M is
1 to c24 are referenced, and finally the coefficients c31 to e related to C are
34 is referred to.

Also, the output signals of LUTs 15 to 17 are multiplied by 'a'
20 to 22 to receive input signals B and G, respectively.
, multiplied by R. and multiplication 1120 and 2
The output signals of 1 are supplied to the adder 23 and added, and this addition! The output signal of 23 is supplied to an adder 26. Also, the output signal of addition@22 and the output signal of LUT 18 are supplied to addition@24 and added together, and this addition 1!24
The output signal of is provided to a summation 825. Then, an output terminal 26 is derived from the adder 26.

Here, when input signals B, G, and R are supplied,
First, coefficients C-th to c14 are output from LUTs 15-18. Therefore, the multipliers 820, 21, and 22 output c 11B + c 12G, c 13R, respectively, and the coefficient c 21 ~C
24 is output. Therefore, Ei! 20. 21
From ゜22, c 21B, c 22G, respectively.
c 23R is output, and the output terminal 26 has (7
) is obtained. Finally LLIT 15
Coefficients c31 to c34 are output from .about.18. Therefore, from multipliers 20, 21, and 22, c 31B
, c32G, and c33R are output, and C No. 18 shown in equation (7) is obtained at the output terminal 26. That is, each time one input signal B, G, R is supplied, Y, M, C signals are sequentially output from the output terminal 26.

In this way, according to this example, matrix calculation is performed using the matrix coefficients determined for each pentagonal pyramid to calculate Y,
Since the f-words of M and C are obtained, even if the relationship between input and output is nonlinear, there is good consistency at the boundary surface, and if there is a smooth color change such as a certain gradation, the boundary surface can be There is no discontinuous change when passing through.

In addition, since color correction data for silk matching of input image data located at the vertices of the basic grid is stored in advance in the LUT, and color correction data is not obtained by weighted averaging by referring to this LUT, it is not possible to obtain color correction data by weighted averaging. There are no drawbacks such as errors.

In the example in Figure 4, the calculation of equation (7) is performed in parallel, but a multiplier is used separately to calculate each term of equation (7) sequentially. It's good to stutter.

In addition, in the example of FIG. 4, input numbers B and G are input.

Although R is 8 bits and processed separately into the upper 4 bits and lower 4 bits, it is of course not limited to this.

Also, in the example in Figure 4, there are 3 inputs, but 4
The present invention can be similarly applied to other cases such as input. In the case of 4 inputs, a 4-dimensional pentagonal pyramid (Mi
The matrix coefficients are determined for each triangular pyramid.

[Effects of the Invention] As explained above, according to the present invention, color-separated image data that has been color-corrected by performing a matrix operation using matrix coefficients determined for each divided space where the number of vertices is the minimum can be obtained. Therefore, even if the relationship between input and output is nonlinear, if there is good consistency at the interface between the divided spaces and there is a smooth color change such as a certain gradation, then when passing through the interface, The changes will not be discontinuous, and the color correction data for the combination of input image data located at the vertices of the basic grid is stored in advance in the LUT, and color correction is performed by referring to this LUT by weighted average. Since no data is obtained, there are no disadvantages such as errors caused by rounding processing.

[Brief explanation of the drawings] Figures 1 to 41! 1 is a diagram for explaining one embodiment of the present invention, and FIG. 5 is a diagram for explaining the invention in detail. 11B, IIG, IIR --- Input terminal 12 to 14 # -- Comparator 15 to 18 20 to 22 ・ 23 to 25 # Lookup table (LUT) -- Multiplier -- Adder / Output terminal

Claims (2)

[Claims] (1) n color separation image data that can be input for color correction (
(n is an integer greater than or equal to 2)) is divided into divided spaces with a minimum number of vertices, and input color separated image data is subjected to matrix calculation using matrix coefficients determined for each divided space. A method for correcting a color-separated image, the method comprising: obtaining color-corrected color-separated image data. (2) n color separation image data that can be input for color correction (
(n is an integer of 2 or more); matrix coefficient storage means for dividing a color space formed by a color space formed by a color space (n is an integer of 2 or more) into divided spaces having a minimum number of vertices, and storing matrix coefficients determined for each divided space; matrix coefficient reading means for reading out matrix coefficients of the divided space containing color separation image data from the matrix coefficient storage means; and matrix operation for the input n color separation image data using the matrix coefficients read from the storage means. 1. A color-separated image correction apparatus, comprising: a matrix calculation means for outputting color-separated image data that has been color-corrected.