JP2899461B2 - Color signal interpolation method, color signal interpolation device, and color correction method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color copying machine, a color facsimile machine, etc.
Y (yellow), M (magenta), C from B (blue) signal
The present invention relates to a color signal interpolation method for obtaining a (cyan) signal, a color signal interpolation device, and a color correction method.

[0002]

2. Description of the Related Art Conventionally, a linear masking method has been known as a color correction method. In this method, when Y, M, and C are ink control amounts and R, G, and B are density signals, Expression (1)
Represented by

[0003]

Y = a10 + a11 × R + a12 × G + a13 × B M = a20 + a21 × R + a22 × G + a23 × B C = a30 + a31 × R + a32 × G + a33 × B. . . Equation (1) Here, a10 to a33 are correction coefficients, and their values can be obtained from the measured values of various color patches by the least square method. No. 1990 pp300).

[0004] This method is practical because the amount of hardware is small, but if sufficient color correction cannot be obtained, the value of R is used.
By using a non-linear masking method such as Equation (2) in consideration of a quadratic term such as a power, G × B, etc. (an even higher-order function may be used), highly accurate color reproduction can be obtained.

[0005]

Y = a10 + a11 × R + a12 × G + a13 × B + a14 × R * 2 + a15 × G * 2 + a16 × B * 2 + a17 × R × G + a18 × G × B + a19 × B × RM M = a20 + a21 × R + a22 × G + a23 × B + a24 × R * 2 + a25 * G * 2 + a26 * B * 2 + a27 * R * G + a28 * G * B + a29 * B * RC = a30 + a31 * R + a32 * G + a33 * B + a34 * R * 2 + a35 * G * 2 + a36 * B * 2 + a37 * R * G + a38 × G × B + a39 × B × R. . . Equation (2) Here, a10 to a39 are correction coefficients, the values of which are obtained from the measured values of various color patches by the least square method. However, x (= R, G, B) * 2 represents the square of x.

Further, interpolation is used as a color correction method for obtaining Y (yellow), M (magenta), and C (cyan) ink control amounts from R (red), G (green), and B (blue) density signals. Several methods have been proposed.

In the first method, the R, G, B space is divided into a plurality of unit cubes, and color correction values obtained by calculation in advance are set at eight grid points of each unit cube, and the grid points are set at the grid points. This is a method of calculating a color correction value of an input color separation signal located in the middle by linearly interpolating eight color correction values. This method requires eight product-sum calculations at the time of linear interpolation, which increases the processing time, complicates the hardware configuration, and changes the interpolated value at the boundary between adjacent unit cubes. There was a problem that the minutes became discontinuous.

As second and third methods for solving this,
There is an interpolation method based on tetrahedral division. The second method is R,
The G and B spaces are divided into a plurality of unit cubes, the divided unit cube is further divided into a plurality of tetrahedrons, and a color correction value at each vertex of each tetrahedron is stored in a memory.
A unit cube is selected by the upper bit data of the B signal, a tetrahedron is selected from the selected unit cube by the lower bit data, and a color correction value corresponding to the input RGB signal is converted to the selected tetrahedron. This is a method of obtaining by linear interpolation using color correction values (Japanese Patent Publication No. 58-16)
No. 180).

In a third method, the R, G, and B spaces are divided into a plurality of unit cubes, and the divided unit cube is further divided into five tetrahedrons, and a correction coefficient is stored in a memory for each tetrahedron. The unit cube is stored, and a unit cube is selected according to the high-order bit data of the input RGB signal. From the selected unit cube,
In this method, a tetrahedron is selected based on lower-order bit data, and color correction is performed by performing a product-sum operation of a correction coefficient of the selected tetrahedron and an input RGB signal (Japanese Patent Laid-Open No. 2-20 / 1990)
No. 6973, Journal of the Institute of Image Electronics Engineers of Japan Vol. 18, No. 5, 1
989 pp 319-328).

[0010]

According to the above-mentioned second method, the calculation formula is simplified as compared with the eight-point interpolation, so that the calculation is performed at high speed, but the number of multipliers is four. Therefore, there is a problem that the amount of hardware increases, and the third method has a simpler hardware configuration because the number of multipliers and adders for performing the product-sum operation is each three. However, since the entire bit width of the RGB signal is input to the multiplier as one input to the multiplier, there is a problem that the scale of the multiplier increases.

On the other hand, the setting of the Y, M, and C values for each grid point in the above-described interpolation method is performed, for example, by obtaining the coefficient of a nonlinear function as shown in the above equation (2) by the least square method. Was determined. That is, the conventional grid point value setting method is based on the premise that the number of space divisions is sufficiently large, and thereby performs interpolation within a unit section with high accuracy. However, in the conventional method, when the number of divisions is small, the accuracy of interpolation deteriorates, and a method of setting grid point values on the assumption of division has not been adopted.

Further, the data used by the least squares method in the conventional setting method includes ink amount data of a large number of patterns (color patches) printed with a predetermined ink amount in the entire printable space. Is used, and the density data obtained by color separation is used.

In this case, the distribution of the density data of a large number of read print patterns does not exist over all the regions of the R, G, and B spaces, but the density data is near the diagonal line of R = G = B. The distribution is large, and decreases as the distance from the diagonal increases (hereinafter, this is referred to as an end). FIG. 19A shows this state, where R = G = B.
The density data 192 is distributed in the vicinity of the diagonal line 191 in FIG.

When the entire R, G, B space is approximated by a non-linear function using such data, the grid point value becomes very large or diverges in a space portion where no data exists or in a small space. Value. In general, the color gamut output by a printer is narrower than that of a photo document. For this reason, the distribution of density data when a photo original is read extends to the end of the R, G, B space as shown by a broken line 193 in FIG. When the Y, M, and C values are set, there is a problem that an unnatural or unnatural color is reproduced at the hatched portions in FIG. 20, that is, at the ends of the R, G, and B spaces.

Further, by using the above-mentioned conventional setting method, the reproduction error (density and color difference) can be minimized on the average in the entire color space. In an achromatic region or a highlight region where the color difference is noticeable, there is a problem that the density is not always smaller than a predetermined color difference, and a sufficient image quality cannot be obtained.

The present invention has been made to solve the problems in the conventional interpolation method and color correction method described above.

An object of the present invention is to provide a color signal interpolation method and a color signal interpolation device which can simplify the hardware configuration and enable highly accurate color correction.

Another object of the present invention is to provide a grid point value setting method suitable for a division type interpolation method.

Still another object of the present invention is to provide a color correction method in which the color reproducibility at the edge of the color space is improved.

Still another object of the present invention is to provide a color correction method which improves color reproduction of a local area where the color difference between an original image and a reproduced image is conspicuous.

[0021]

In order to achieve the above object, according to the present invention, an input first color signal is provided.
Signal interpolation method for finding output value of second color signal corresponding to
A three-dimensional color space consisting of the first color signal.
Divided into multiple cubes, and each cube is further divided into multiple triangles.
Divided into pillars and using the input first color signal
Select one of the triangular prisms, and in the selected triangular prism,
In the triangular prism indicated by the input first color signal
The output value of the second color signal corresponding to the position
Using the specified value set for the selected triangular prism, draw a line
It is characterized by being calculated by shape interpolation .

According to the second aspect of the present invention, the first input
Signal for obtaining the output value of the second color signal corresponding to the color signal of
Signal interpolating device, the three-dimensional signal comprising the first color signal
The color space is divided into multiple cubes, and each cube is further duplicated.
Divided into a number of triangular prisms and a predetermined value for each of the divided triangular prisms
Using the input first color signal
Means for selecting one of the triangular prisms by
The position in the triangular prism indicated by the first color signal is
The output value of the corresponding second color signal is
Linear interpolation using a predetermined value read from a prism
And means for calculating by means of

According to the third aspect of the present invention, the first input
Signal for obtaining the output value of the second color signal corresponding to the color signal of
In the signal interpolation method, a three-dimensional signal comprising the first color signal
The color space is divided into multiple cubes, and each cube is further duplicated.
Number of triangular prisms and use the input first color signal
And select one of the triangular prisms,
The three color signals represented by the input first color signal.
Output of the second color signal corresponding to a position in the prism
Value is set for each vertex of the selected triangular prism
Color signal calculated by linear interpolation using output values
An interpolation method, wherein a predetermined data of the first and second color signals is
Using the data pairs, the output value on the vertex is used as a variable,
The coefficient of a linear function that is linear in a triangular prism
Calculated Me multiplicative, calculated by the linear function with the coefficients
The output value to be set is set to each of the vertices .

[0024] In the invention of claim 4, linear or non-linear least squares method using a plurality of pattern data printed at a predetermined ink amount, and density data obtained by color separation of the plurality of patterns In a color correction method for obtaining a correction coefficient of a function and setting Y, M, and C values calculated using the linear or non-linear function in which the correction coefficient has been set to respective grid points in R, G, and B spaces, The Y, M, and C values calculated using a linear function are set to grid points in an area where the density data is equal to or less than a predetermined value. Y calculated using
M and C values are set.

According to the fifth aspect of the present invention, a plurality of pattern data printed with a predetermined amount of ink and density data obtained by color separation of the plurality of patterns are used to form a non-linear function by a least square method. A correction coefficient is obtained, and the Y, M, and C values calculated by using the non-linear function in which the correction coefficient is set are expressed by:
In the color correction method set at each grid point in the R, G, and B spaces, a correction coefficient of a nonlinear function is obtained by a least squares method using data of an achromatic region or a highlight region of the density data. It is characterized in that Y, M, and C values calculated by the non-linear function for which coefficients have been set are set at grid points of the region in the R, G, and B spaces.

[0026]

In the embodiment of the present invention, a triangular prism is selected from the magnitude relation of the input color signals X, Y, and Z, and the slope coefficient ai and the intercept coefficient bi of the selected triangular prism are read out from the memory.
The interpolation calculation unit of the formula (1) calculates the selected slope coefficient ai and the intercept coefficient bi.
Is used to perform interpolation on the sides of the triangular prism. The second interpolation operation unit uses the output of the first interpolation operation unit to perform interpolation on the triangular surface of the selected triangular prism, and to input color signals X, Y,
It outputs Y, M, C signals corresponding to Z. Therefore, according to the present embodiment, since the memory capacity can be reduced as compared with the conventional interpolation method, the entire hardware can be reduced, the LSI can be easily formed, and the color correction can be performed with high accuracy. it can.

In another embodiment of the present invention, the color correction circuit comprises:
Output color patch to input sensor, A / D converter, log
The data is read through the converter and the density data is measured for each patch. The correction coefficient a10 of the linear linear expression is obtained by the least square method using all the data pairs of R, G, B and Y, M, C.
a33 and correction coefficients a10 to a39 of a nonlinear quadratic equation are obtained. The color correction circuit determines whether or not there is more than a predetermined number of data in the vicinity of the grid point of interest. If there is no more than the predetermined number of data in the vicinity of the grid point of interest, the correction coefficient a10 The Y, M, and C values calculated by the linear equation are set using a33, and if there is a predetermined number or more of data in the vicinity of the grid point of interest, the correction coefficient a
The Y, M, and C values calculated by the nonlinear quadratic equation are set using 10 to a39. As described above, since the linear masking is applied to the end of the color space, a color without a sense of incongruity is reproduced. Further, a reproduced image having a small color difference between the original image and the reproduced image is obtained by the non-linear masking.

In yet another embodiment of the present invention, Y, M, C
A color patch is output according to the signal, and the color correction circuit outputs the output color patch to an input sensor, an A / D converter,
The density data (R, G, B) is measured for each patch by reading through an og converter. A correction coefficient of a nonlinear function is obtained by the least squares method using all data pairs of R, G, B and Y, M, C. The correction coefficient (a) of the nonlinear function is obtained by the least squares method using only the data within a predetermined value in the data pair. A correction coefficient (a) is given to a lattice point satisfying the condition of R = G = B and a lattice point satisfying the condition of (R <threshold Tr1 & G <threshold Tg1 & B <threshold Tb1).
Are used to set the calculated Y, M, and C values. Therefore,
The color reproducibility of the achromatic region and the highlight region can be improved.

[0029]

Embodiment 1 An embodiment of the present invention will be specifically described below with reference to the drawings. First, the division method of the present invention will be described. In the present invention, the X, Y, and Z spaces are divided into a plurality of unit triangular prisms as shown in FIG. Then, when calculating the value of the output P at the input X, Y, and Z coordinates, a unit triangular prism including the input X, Y, and Z coordinates is selected, and the six triangular vertices of the selected unit triangular prism are selected. An output value at the output P is obtained by linear interpolation based on the output value (this value is a known value obtained by a predetermined method). Here, when applied to color correction, X, Y, Z are input R (red), G (green), B
(Blue) signal, the output P is Y (yellow), M (magenta), C for controlling ink in the case of a four-color printer.
(Cyan) and Bk (black) signals.

FIG. 3 is a diagram for explaining interpolation on the sides of the triangular prism. The output value at each vertex coordinate of the triangular prism is represented by Pi.
(I = 1 to 6, the values are known), the output values at the intersection coordinates of the triangular surface 31 and the triangular prism 32 including the coordinates of the output P to be obtained are Pa, Pb, and Pc, respectively.
The value of a is obtained by linear interpolation using the line segment ratio between P1 and P2. Similarly, the value of Pb is P3 and P4, and the value of Pc is P
5. It is obtained from P6 by linear interpolation. Specifically, for example, for the value of Pa, the length between P1 and Pa is m
1. If the length between Pa and P2 is m2,

[0031]

Pa = (m1 × P2 + m2 × P1) / (m1 + m2) The same applies to Pb and Pc.

Next, on the triangular surface 31 shown in FIG.
Output values Pa, Pb, Pc of each vertex obtained as described above
Is linearly interpolated to obtain the value of the output P. That is, linear interpolation is performed by obtaining the contribution ratio of the output value at the coordinates of each vertex to P by the area ratio of the triangle formed by the coordinates of P and each vertex. Let the coordinates of each vertex of the triangle be A,
When B and C are set, the contribution of Pa to P is represented by the area Sa of the triangle PBC at the diagonal position of Pa, and the contribution of Pb to P is represented by the area S of the triangle PAC at the diagonal position of Pb.
b, The contribution of Pc to P is determined by the ratio of the area Sc of the triangle PAB at the diagonal position of Pc.

[0033]

## EQU00004 ## P = (Sa.times.Pa + Sb.times.Pb + Sc.times.Pc) / (Sa + Sb + Sc) The method of interpolating according to the area ratio is represented by coordinates A,
This is equivalent to obtaining P by a plane equation satisfying the three points B and C. The above-mentioned division method is referred to as type 1 (FIG. 5A is the XY plane , and each vertex forms a square lattice ).
Then , as another division method of the present invention, type 2
Is shown in FIG. 5B ( each vertex forms a triangular lattice ).

Hereinafter, a specific embodiment of the present invention will be described in detail with reference to FIG. FIG. 1 shows the input color signals X, Y, Z
1 is a block diagram of a color correction apparatus according to the present invention, which outputs a color correction signal of, for example, Y (yellow). The color correction devices for M (magenta) and C (cyan) have exactly the same configuration. In FIG. 1, a triangular prism selection unit and a memory unit 1 select a triangular prism from the magnitude relation of the input color signals X, Y, and Z, and read out a slope coefficient ai and an intercept coefficient bi of each side of the selected triangular prism. The interpolation calculation unit 2 is a calculation unit that performs interpolation on a triangular prism selected using the slope coefficient ai and the intercept coefficient bi, and includes a multiplier 201 and an adder 202. The interpolation operation unit 3 is an operation unit that performs interpolation on the triangular surface of the selected triangular prism using the output of the interpolation operation unit 2, and includes subtractors 301 and 302, a multiplexer 303, and multipliers 304 and 305. 306.

Selection of triangular prism: Each coordinate of the input space X, Y, Z is divided into 2 n powers, and the space X, Y, Z is divided into 2 3n cubes. Now, assuming that the bit width of X, Y, Z is f bits, X, Y, Z
Z

[0036]

X = x + Δx, Y = y + Δy, Z = z + Δz, where f bits are upper n bits and lower (f−n)
And the upper n bits of X, Y, and Z are x,
In y and z, lower (f−n) bits are represented by Δx, Δy, Δz
, One of the cubes previously divided into 2 3n powers is selected by the upper n bits of X, Y, and Z. The lower (f−n) bits are relative positions Δx, Δy, Δz in the selected cube.
Will be shown.

Selection of type 1 triangular prism; When a cube is selected as described above, two of the three sides of the triangular prism in the Z-axis direction are identified, and the lower bits Δx, Δy
The remaining one side is identified by the magnitude relation of. That is, FIG. 6A shows the selected cube, and FIG.
(B) is a plan view of the upper bits x, y, and z of X, Y, and Z, and the two sides A and C of the triangular prism are converted to the upper bits x, y, and z.
The remaining one side B or D (B when Δx ≧ Δy) is selected according to the magnitude relationship between y and z and the lower bits Δx and Δy.

FIG. 7 shows a configuration for generating a type 1 triangular prism selection signal in the triangular prism selecting unit and the memory unit 1. That is, the memory 71 is referred to by the upper bits x, y, and z to read out the two sides A and C of the triangular prism, and the memory 7 is read by the upper bits x, y, and z and the lower bits Δx, Δy.
2 is read, the remaining one side B of the triangular prism is read out when Δx ≧ Δy, and the one side D is read out when Δx <Δy. Based on these side information, a selection signal for selecting the triangular prism ABC or the triangular prism ACD is obtained. Generated.

Selection of triangular prism of type 2 Similarly, in the case of type 2, when a cube is selected, one of the three sides of the triangular prism in the Z-axis direction (the side passing through the center of the cube) is identified. The remaining two sides are identified by the magnitude relationship between the lower bits Δx and Δy. That is, FIG.
Indicates a selected cube, and FIG. 8 (b) is a plan view of the cube. One side E of the triangular prism is formed by upper bits x, y, z of X, Y, Z and upper bits x, y, z. Lower bits Δx, Δ
The remaining two sides are selected according to the magnitude relation of y.

FIG. 9 shows a configuration for generating a type 2 triangular prism selection signal in the triangular prism selecting unit and the memory unit 1. That is, the memory 91 is accessed by the upper bits x, y, and z, and one side E of the triangular prism is read, the memory 92 is accessed by the upper bits x, y, and z and the lower bits Δx, Δy, and Δx ≧ Δy When side B is read, Δx <
At the time of Δy, the side D is read. Similarly, sides A and C are read from the memory 93 according to the magnitude relationship (Δx + Δy ≧ 1 or Δx + Δy <1) between the upper bits x, y, z and the lower bits Δx, Δy. A selection signal for selecting the four triangular prisms ABE, BEC, CED, and AED is generated based on the side information.

When Type 1 and Type 2 are compared, when the number of grid points is almost the same, Type 2 has a smaller memory capacity, and Type 2 is advantageous when high-speed processing such as color copying is considered.

In the following description, type 1 is used, but type 2 is basically the same as type 1, except that the coefficients α, β, and γ when performing the interpolation operation on the triangular surface are different from type 1. is there.

Interpolation on Side of Triangular Prism; There are two methods for storing in the memory 1 a known value for each triangular prism when performing an interpolation calculation. In the first method, output values of six vertices of a triangular prism (which are known values calculated by a predetermined method) are directly stored in a memory 1.
This is a method of storing the information. According to this method, when the output values at two vertices of one side of the selected triangular prism are P1 and P2, the output value Pa corresponding to the lower bit Δz of the input Z is obtained.
Is obtained by interpolating on the side as shown in FIG.
Next formula

[0044]

Pa = P2 + Δz × (P2-P1) The same applies to the other two sides.

The second method is to store (P2-P1) in the memory 1.
This is a method of storing (this is represented by ai and is referred to as a slope coefficient) and P2 (this is represented by bi and is referred to as an intercept coefficient). According to this method, there is an advantage that it is not necessary to calculate P2-P1 by a subtraction circuit. In the present invention, the second
I will adopt the method of.

In summary, in FIG. 1, one triangular prism is selected by the upper and lower bits of the inputs X, Y, and Z, and the memory 1 stores the coefficient (a1) of the first side of the triangular prism according to the triangular prism selection signal. , B1), the coefficient (a2, b2) of the second side, and the coefficient (a3, b3) of the third side are read out, and the interpolation calculation unit 2 uses each coefficient to calculate P
X = ai × Δz + bi (i = 1, 2, 3, X = A, B,
C) to calculate the interpolation output values PA, P on the sides of the triangular prism.
B and PC are output.

Interpolation on a triangular surface on a triangular prism: These output values PA, PB, and PC are linearly interpolated by a line segment ratio to finally obtain an output P to be obtained. FIG. 11 is a diagram illustrating interpolation on a triangular surface of a type 1 triangular prism. When the output P is within the triangular plane 1 (Δx ≧ Δy),
When linear interpolation is performed using the interpolation method on the side of the triangular prism described above,
The output P is

[0048]

P = (PB−PA) × Δx + (PC−PB) × Δy + PA

When the output P is within the triangular plane 2, (Δx <
Δy), similarly,

[0050]

P = (PC−PB) × Δx + (PB−PA) × Δy + PA, and the coefficients of Δx and Δy of the triangular surface 1 are interchanged.

Therefore, the interpolation operation unit 3 in FIG.
First, the subtracter 301 subtracts (PB-PA) from PA, PB, and PC output from the interpolation calculator 2,
To calculate (PC-PB) and output it to the multiplexer 303. If the lower bit of the input signal is Δx ≧ Δy, the multiplexer 303 sends the output (PB-PA) of the subtractor 301 to the multiplier 304 based on the control signal indicating Δx ≧ Δy, and performs multiplication. The output (PC-PB) of the subtractor 302 is sent to the unit 305. The result of the multiplication and the PA are added by the adder 306, and the output P is calculated.

On the other hand, when Δx <Δy, the multiplexer 303 sends the output (PC-PB) of the subtractor 302 to the multiplier 304 and sends the output (PB-PB) of the subtractor 301 to the multiplier 305. PA), and the result of the multiplication is added to the PA by the adder 306 to calculate the output P. This output is, for example, a Y (yellow) signal, and the M (magenta) and C (cyan) signals are also generated by the same circuit as in FIG.

The above-mentioned interpolation formula is expressed as P = α × Δx + β × Δy
When expressed as + γ, the coefficients α, β, and γ at the time of performing the interpolation calculation on the triangular surface of the type 2 triangular prism are as follows. FIG. 12 is a diagram for explaining interpolation on triangular surfaces 0 to 3 in a type 2 triangular prism, and vertices A, B, and C of each triangular surface are illustrated.
If the output values at are PA, PB, and PC, the triangle plane 0
Α is PB-PC, β is 2PA-PB-PC, γ is PC
Becomes Α of triangle 1 is 2PA-PB-PC, β is PB
-PC, γ is PC, α of triangle 2 is PB + PC-2P
A and β are PC-PB, γ is 2PA-PC, α of triangle 3
Is PC-PB, β is PB + PC-2PA, γ is 2PA−
PC. Then, an interpolation operation is performed using these coefficients.

Embodiment 2 FIG. 13 is a block diagram of a second embodiment of the present invention. In FIG. 13, reference numeral 131 denotes a ROM that stores output values on grid points. For example, when X, Y, and Z are divided into three ( that is, 4 × 4 × 4 = 64 grid points ),
131 stores grid point information of 64 bytes per color. Reference numeral 132 denotes an interpolation processing unit as shown in FIG.
RA at which grid point information of ROM 131 is loaded at the time of execution
M133, and a Y processing unit 13 for generating Y, M, and C signals by referring to the RAM 133 from the input RGB signals.
4, an M processing unit 135 and a C processing unit 136.
137 is a CPU for controlling the whole.

In this embodiment, the overlap (for example, 512 bytes) of the grid point information in the ROM 131 is loaded into the RAM 133 at the time of execution in consideration of simultaneous access to adjacent grid points. According to the second embodiment, the minimum necessary grid point information is stored in the ROM.
M can be configured with a small capacity.

<Embodiment 3> FIG. 14 is a block diagram of a third embodiment of the present invention. The difference from the second embodiment is that an image processing unit 138 is provided. . This image processing unit 13
Reference numeral 8 denotes, for example, color conversion or fine adjustment of the color of the grid point information in the ROM 131, which makes it possible to perform color conversion flexibly.

<Embodiment 4> In the above embodiment, the value (lattice point value) of the output value Pi (i = 1 to 6) at each vertex coordinate of the triangular prism has been described as being known. Relates to a method of setting grid point values suitable for a division type interpolation method. As described above, the conventional grid point value setting method is based on the premise that the number of space divisions is sufficiently large, and performs interpolation within a unit section with high accuracy. However, in such a conventional method, when the number of divisions is small, the accuracy of interpolation deteriorates, and a method of setting grid point values on the premise of division has not been adopted.

The grid point value setting method of the present embodiment is a method capable of performing high-precision interpolation even when the number of divisions is small, on the premise of division type interpolation. The setting method in triangular prism division type interpolation will be described in detail below. explain. The present invention is not limited to this, and can be applied to the above-described eight-point interpolation method or four-point interpolation method.

In a known digital color copy,
For example, data obtained by dividing Y, M, and C signals of 256 gradations into, for example, 16 steps are gamma-converted, subjected to halftone processing, and subjected to hard copy (color patch). Then, the output color patches are read by a scanner, and the density data (R, G, B) is measured for each patch. Thereby, (Rp, Gp, Bp) and (Yp, Mp, Cp) for each color patch
The correspondence of p) is obtained (p is the number of color patches). Using all the R, G, B and Y, M, C data pairs, the coefficients of the linear function in the divided section are determined by the least squares method.
A grid point value calculated by a linear function using the coefficient is set to a grid point of the divided section.

That is, the X axis is divided into L, the Y axis is divided into M,
By dividing the Z axis into N, the X, Y, and Z spaces are divided into unit rectangular parallelepipeds, and the unit rectangular parallelepiped is further divided into two units and divided into unit triangular prisms. Inputs x, y, and z are 0 to 255, respectively.
, The side lengths of the unit triangular prism in the X-axis, Y-axis, and Z-axis directions are dx = 256 / L, dy = 256 / M,
dz = 256 / N.

The values of each grid point are represented by Pi, j, k (i = 0 to
Assuming that L, j = 0 to M, k = 0 to N) (P = Y, M, C), the x, y, z coordinates of Pi, j, k are

[0062]

(X, y, z) = (i × dx, j × dy, k × dz)

Now,

[0064]

## EQU10 ## If Δxijk = (xi−dx) / dx, Δyijk = (y−i × dy) / dy, Δzijk = (zi−dz) / dz, then (i × dx, j × dy) , K × dz) as a starting point, the area of the triangular prism 1 is represented as D (1) ijk = 1, and the other area is represented as D (1) ijk = 0.

That is, the region where D (1) ijk = 1 is

[0066]

(11 × dx ≦ x <(i + 1) × dx j × dy ≦ y <(j + 1) × dy k × dz ≦ z <(k + 1) × dz Δyijk <Δxijk) The triangle prism ABC in FIG. (Or a triangular surface 1 in FIG. 11).

Also, (i × dx, j × dy, k × dz)
Is represented as D (2) ijk = 1 in the area of the triangular prism 2 and D (2) ijk = 0 in other areas.

That is, the region where D (2) ijk = 1 is

[0069]

(I × dx ≦ x <(i + 1) × dx j × dy ≦ y <(j + 1) × dy k × dz ≦ z <(k + 1) × dz Δyijk ≧ Δxijk) and the triangular prism ACD in FIG. (Or a triangular surface 2 in FIG. 11).

D (1) triangular prism with ijk = 1 (Δyijk
Assuming that each grid point value of <Δxijk) is as shown in FIG. 15, output values PA, PB, and PC at coordinates of the intersection of the triangular surface and the triangular prism including the coordinates of the output P are calculated on the sides of the triangular prism described above. It is obtained in the same way as interpolation,

[0071]

PA = Pi, j, k + Δzijk × (Pi, j, k + 1, −Pi, j, k) PB = Pi + 1, j, k, + Δzijk × (Pi + 1, j, k + 1, −Pi + 1, j, k ) PC = Pi + 1, j + 1, k, + Δzijk × (Pi + 1, j + 1, k + 1, −Pi + 1, j + 1, k)

The output P is obtained by the above-described interpolation on the triangular surface.

[0073]

P (Y, M, C) = (PB−PA) × Δxijk + (PC−PB) × Δyijk + PA = Δxijk × Δzijk × (Pi + 1, j, k + 1, −Pi + 1, j, k, −Pi, j, k + 1, + Pi, j, k) + Δxijk (Pi + 1, j, k, −Pi, j, k) + Δyijk × Δzijk × (Pi + 1, j + 1, k + 1, −Pi + 1, j + 1, k, −Pi + 1, j, k + 1, + Pi + 1, j, k) + Δyijk (Pi + 1, j + 1, k, −Pi + 1, j, k) + Pi, j, k + Δzijk (Pi, j, k + 1, −Pi, j, k) Then

[0074]

P (Y, M, C) = Pi, j, k × (Δxijk × Δzijk−Δxijk−Δzijk + 1) + Pi + 1, j, k × (Δxijk−Δxijk × Δzijk−Δyijk + Δyyjk × Δzijk) + Pi + 1 j + 1, k × (Δyijk−Δyijk × Δzijk) + Pi, j, k + 1 × (Δzijk−Δxijk × Δzijk) + Pi + 1, j, k + 1 × (Δxijk × Δzijk−Δyijk × Δzijk) + Pi + 1, j + 1k × Δzijk).

A similar calculation is made for the triangular prism of D (2) ijk = 1, and P is represented by D (1) ijk and D (2) ijk.

[0076]

P = １６ [D (1) ijk × ｛Pi, j, k × (Δxijk × Δzijk−Δxijk−Δzijk + 1) + Pi + 1, j, k × (Δxijk−Δxijk × Δzijk−Δyijk + Δyijk × Δzijk) + Pi + 1, j + 1, k × (Δyijk−Δyijk × Δzijk) + Pi, j, k + 1 × (Δzijk−Δxijk × Δzijk) + Pi + 1, j, k + 1 × (Δxijk × Δzijk−Δyjk × Δzijk) +1, Pi + 1 × (Δyjk × Δzijk)｝ D (2) ijk × ｛Pi, j, k × (−Δyijk + Δyijk × Δzijk−Δzijk + 1) + Pi + 1, j + 1, k × (Δxijk−Δxijk × Δzijk) + Pi, j + 1, × Δxijk × Δzijk−Δxijk + Δyijk−Δyijk × Δzijk) + P i, j, k + 1 × (Δzijk−Δyijk × Δzijk) + Pi + 1, j + 1, k + 1 × (Δxijk × Δzijk) + Pi, j + 1, k + 1 × (Δyijk × Δzijk-Δxijk × Δzijk)}] = {fijk, z) × Pijk, so that it becomes linear with respect to Pijk, and a linear least squares method can be applied. That is, using the sample data described above, the coefficient of the linear function is obtained by the least squares method, and the lattice point value is calculated from the linear function thus set.

Incidentally, (Rp,
In the corresponding data of (Gp, Bp) and (Yp, Mp, Cp), the (Rp, Gp, Bp) data is not distributed over the entire R, G, B space, and there is a unit space without data. . Therefore, a solution may not be obtained. As a countermeasure, it is necessary to replenish the data in the unit space. As a method, for the data in the whole area of the R, G, B space or the data in the vicinity of the unit space without data,
There are a method of setting a non-linear expression by the least square method and supplementing data based on the data, and a method of setting a linear format and supplementing data based on the data.

When data is extrapolated by the former nonlinear equation,
Very large or divergent values may occur due to non-linearity.
Such a situation can be prevented. Therefore, in this embodiment,
If the data in the unit space is small, the extrapolated data is added nonlinearly, and if there is no data in the unit space, the data is extrapolated linearly. May be supplemented with linear extrapolated data).

<Embodiment 5> This embodiment relates to a color correction method in which the color reproducibility at the edge of the color space is improved. FIG. 16 shows a color copy to which the color correction method of this embodiment is applied. FIG. 2 is a block diagram of the machine. In FIG. 16, an input sensor 161 has a photoelectric conversion element such as a CCD camera,
It outputs three color separation signals (red), G (green), and B (blue).
The A / D converter 162 converts the R, G, and B signals into, for example, 8
Convert to a bit digital signal. log converter 163
Converts the digital signals of R, G, and B into density and outputs them.

The color correction circuit 164 is a circuit to which the color correction method of the present embodiment is applied, and generates Y, M, and C ink amount control signals from the density data (R, G, B) of the log converter 163. I do. That is, for example, the R, G, and B spaces are divided into a plurality of unit cubes in the same manner as in the above-described first method, and calculated in advance (calculated by the least square method) at eight grid points of each unit cube. Memory for color correction values (Y, M, C values)
The Y, M, and C signals are calculated by linearly interpolating the color correction values of the input color separation signal located in the middle of the grid points and the eight color correction values. The present embodiment relates to a method for setting Y, M, and C values for grid points.

The UCR circuit 165 performs under color removal processing (UC
R processing), and Y, M,
With respect to the C signal, a signal obtained by removing a part or all of the C signal and a signal K of black ink corresponding to the removed amount are generated. That is,

[0082]

K = α × min (Y, M, C), Y1 = Y−
K, M1 = M−K, C1 = CK, where α is a predetermined constant (for example, 0.5). As a result, black in a high-density region that cannot be reproduced even by overprinting the Y, M, and C inks is reproduced, and an effect is obtained in which dark portions in the picture are tightened and the amount of ink is reduced.

The dither circuit 166 includes Y1, M1, C1,
A circuit for binarizing the K1 signal by the systematic dither method. The output signals Y2, M2, C2, and K2 are sent to the color printer 167 bit by bit, and a color image is reproduced by turning on / off the ink dots of each color. Is done.

In this embodiment, the first interpolation method described above, that is, the R, G, B space is divided into a plurality of unit cubes, and eight grid points of each unit cube are calculated in advance. The present invention is applied to a method of setting color correction values (Y, M, C values) and calculating the color correction values of the input color separation signal located in the middle of the grid points by linearly interpolating the eight color correction values. The case will be described.

The method of setting the Y, M, and C values for each grid point according to this embodiment will be described below. In the color copying machine shown in FIG. 16, a color patch is formed by the Y, M, and C signals for which UCR processing is not performed. Output. Here, as the Y, M and C signals, a total of 4913 data obtained by dividing Y, M and C of 256 gradations into 16 stages are used. That is, Y: 0, 15, 31,. . . 239, 255 M: 0, 15, 31,. . . 239, 255 C: 0, 15, 31,. . . FIG. 17 shows an output example of color patches of 239, 255 (a total of 17 to the third power).

Next, the color correction circuit 164 converts the output color patch into the input sensor 161 and the A / D converter 16.
2. Read through the log converter 163 and measure the density data (R, G, B) for each patch. The correction coefficients a10 to a33 of the above equation (1) and the correction coefficients a10 to a39 of the above equation (2) are obtained by the least square method using all the R, G, B and Y, M, C data pairs. Respectively.

The color correction circuit 164 determines whether there is a predetermined number or more of color patch data in the vicinity of the grid point of interest in the R, G, and B spaces (or whether there is data).
FIG. 18 is a G, R plan view near a lattice point. Now, assuming that the coordinate value of the grid point 181 of interest is (r, g, b),
The number of data included in the neighboring area (the hatched portion 182 in FIG. 18) is obtained. That is, rl <R <r + 1, gl
Data that satisfies <G <g + 1, bl−B <B <b + 1 is counted, and it is determined whether or not there is a predetermined number or more.

If there is no more than a predetermined number of data in the vicinity of the grid point of interest, the correction coefficients a10 to a33
Are used to set the Y, M, and C values calculated by the above equation (1). If there is a predetermined number or more of data in the vicinity of the grid point of interest, the correction coefficients a10 to a39 are assigned to the grid point of interest.
Are used to set the Y, M, and C values calculated by the above equation (2). In this way, by setting the Y, M, and C values for each grid point, color reproducibility can be improved even at the ends of the R, G, and B spaces.

Although the fifth embodiment is applied to the eight-point interpolation method, the present invention is not limited to this. The four-point interpolation method, which is the method described above, or It can also be applied to the proposed six-point interpolation method using triangular prism division.

Embodiment 6 This embodiment relates to a color correction method in which color reproduction is improved in a local region, for example, an achromatic region or a highlight region in which a difference in color between an original image and a reproduced image is conspicuous. The first method of setting the Y, M, and C values for each grid point according to the present embodiment is similar to that of the fifth embodiment described above. A color patch is output by a signal (FIG. 17), and as Y, M, and C signals, a total of 4913 data obtained by dividing Y, M, and C of 256 gradations into 16 stages are used. That is, Y: 0, 15, 31,. . . 239, 255 M: 0, 15, 31,. . . 239, 255 C: 0, 15, 31,. . . Next, the color correction circuit 164 reads the output color patches via the input sensor 161, the A / D converter 162, and the log converter 163, and for each patch, Measure the density data (R, G, B). All R, G, B and Y,
Equation (2) using the least squares method using the M and C data pairs
Are obtained.

Then, 4913 R, G, B and Y,
The correction coefficient of the above equation (2) is obtained by the least squares method using only the data in which Δ (R, G, B) is within a predetermined value out of the M and C data pairs. This correction coefficient is a
a391. here,

[0092]

Δ (R, G, B) = max (RG, GB, BR), which is data near an achromatic color.

In this embodiment, the Y, M, and C values calculated by the above equation (2) using the correction coefficients a101 to a391 are set at the lattice points satisfying the condition of R = G = B. Are calculated using the correction coefficients a10 to a39.
The Y, M, and C values calculated in (2) are set. Also, R
= G = B, the correction coefficients a101-a
391 may be used to set the Y, M, C values calculated. As described above, by setting the Y, M, and C values, it is possible to improve color reproducibility in the vicinity of an achromatic color.

FIG. 21 is a plan view of the G and R grid points.
k1 is a lattice point where R = G, and k2 and k3 are R = G
This is a nearby grid point.

Second setting method of Y, M, C values for each grid point according to the present embodiment; In the second setting method, 4913 R, G, B and Y, M, C data pairs Of these, only the data in which R, G, and B are within a predetermined value (R <threshold Tr & G <threshold Tg & B <threshold Tb), that is, the data in the highlight area, are used in the least squares method, and Find the correction coefficient. The correction coefficients are denoted by a102 to a392.

(R <threshold Tr1 & G <threshold Tg1 & B <
A lattice point satisfying the condition of the threshold value Tb1) has a correction coefficient a1
Y calculated from the above equation (2) using 02 to a392,
M and C values are set, and correction factors a
Y, M, calculated by the equation (2) using 10 to a39.
The C value is set. As described above, the color reproducibility of the highlight area can be improved by setting the Y, M, and C values.

FIG. 22 is a plan view of the G and R lines, from h1 to h4.
Is a region (R) where the G and R density values are each equal to or less than a predetermined threshold.
<63 & G <63) are the grid points in the highlight area.

Although the sixth embodiment described above is a case in which the present invention is applied to the eight-point interpolation method, the present embodiment is not limited to this, and the four-point interpolation method, which is the method described above, or the present applicant. Can also be applied to the six-point interpolation method based on triangular prism division proposed by R.

[0099]

As described above, according to the first and second aspects of the present invention, since the memory capacity is reduced as compared with the conventional interpolation method, the entire hardware is reduced and the LSI is easily implemented. Color correction can be performed with high accuracy.

According to the third aspect of the present invention, since the grid point values are set on the premise of division type interpolation, color correction can be performed with high accuracy even when the number of divisions is small.

According to the fourth aspect of the present invention, since the linear masking is applied to the end of the color space, a color without a sense of incongruity is reproduced, and a reproduced image having a small color difference between the original image and the reproduced image is obtained by the nonlinear masking. Can be In particular, it is possible to improve the color reproducibility of an original such as a silver halide photograph having a wider color gamut than a printer.

According to the fifth aspect of the present invention, the color reproducibility of the achromatic region and the highlight region can be improved without increasing the amount of hardware.

[Brief description of the drawings]

FIG. 1 is a block diagram of a color correction apparatus according to an embodiment of the present invention.

FIG. 2 is a diagram in which an X, Y, and Z space is divided into a plurality of unit triangular prisms.

FIG. 3 is a diagram illustrating interpolation on a side of a triangular prism.

FIG. 4 is a diagram illustrating interpolation on a triangular surface.

FIG. 5 (a) is a diagram illustrating a type 1 division method according to the present invention. (B) is a diagram showing a type 2 division method according to the present invention.

FIG. 6A is a diagram illustrating a selected type 1 cube. (B) is a plan view thereof.

FIG. 7 is a configuration diagram for generating a type 1 triangular prism selection signal.

FIG. 8A is a diagram showing a selected type 2 cube. (B) is a plan view thereof.

FIG. 9 is a configuration diagram for generating a type 2 triangular prism selection signal.

FIG. 10 is a diagram illustrating interpolation on the sides of a triangle.

FIG. 11 is a diagram illustrating interpolation on a triangular surface of a type 1 triangular prism.

FIG. 12 is a diagram illustrating interpolation on triangular surfaces 0 to 3 in a type 2 triangular prism.

FIG. 13 is a block configuration diagram according to a second example of the present invention.

FIG. 14 is a block configuration diagram according to a third embodiment of the present invention.

FIG. 15 is a diagram for explaining a method of setting grid point values.

FIG. 16 is a block diagram of a color copying machine to which the color correction method of the present invention is applied.

FIG. 17 is a diagram illustrating an output example of a color patch.

FIG. 18 is a G, R plan view near a lattice point.

FIG. 19A is a diagram showing a density data distribution near a diagonal line of R = G = B. (B) is a G, R plan view of the density data distribution.

FIG. 20 is a diagram illustrating an end of a R, G, B space.

FIG. 21 is a diagram showing an achromatic region of a grid point on the G and R planes.

FIG. 22 is a diagram showing a highlight area of a grid point on the G and R planes.

[Explanation of symbols]

 1 Triangular prism selection unit and memory unit 2, 3 Interpolation operation unit 201, 304, 305 Multiplier 202, 306 Adder 301, 302 Subtractor 303 Multiplexer 161 Input sensor 162 A / D converter 163 log converter 164 Color correction circuit 165 UCR circuit 166 Dither circuit 167 Color printer

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁶ Identification symbol FI H04N 1/46 Z (31) Priority claim number Japanese Patent Application No. 3-96221 (32) Priority Date Hei 3 (1991) April 2 (33) Priority claiming country Japan (JP) (56) References JP-A-60-14372 (JP, A) JP-A-2-226867 (JP, A) JP-A-2-16875 (JP, A) JP-A-63-208370 (JP, A) JP-A-57-198463 (JP, A) JP-A-4-119765 (JP, A) JP-B-58-16180 (JP, B2) (58) Int.Cl. ⁶ , DB name) H04N 1/40-1/409 H04N 1/46-1/64 G06F 17/00-17/18 G06F 1/02 G06F 7/548

Claims (5)

(57) [Claims] 1. A second color signal corresponding to an input first color signal.
A color signal interpolation method for obtaining an output value of a color signal of the first color signal, wherein the three-dimensional color space composed of the first color signal is divided into a plurality of cubes, and each of the cubes is further divided into a plurality of triangular prisms; One of the triangular prisms is selected using the first color signal to be obtained, and the second triangle corresponding to a position in the triangular prism indicated by the input first color signal in the selected triangular prism. A color signal interpolation method, wherein an output value of a color signal is calculated by performing linear interpolation using a predetermined value set for the selected triangular prism. 2. A second color signal corresponding to an input first color signal.
A color signal interpolating apparatus for calculating an output value of a color signal of the first color signal, wherein the three-dimensional color space including the first color signal is divided into a plurality of cubes, and each of the cubes is further divided into a plurality of triangular prisms; Means for storing a predetermined value for each of the triangular prisms, means for selecting one of the triangular prisms by using the input first color signal, and the means indicated by the input first color signal. Means for calculating an output value of the second color signal corresponding to a position in the triangular prism by linearly interpolating using a predetermined value read from the selected triangular prism. Interpolator. 3. A second color signal corresponding to an input first color signal.
In the color signal interpolation method for obtaining the output value of the color signal, the three-dimensional color space composed of the first color signal is divided into a plurality of cubes, and each of the cubes is further divided into a plurality of triangular prisms, and the input is performed. One of the triangular prisms is selected using a first color signal, and in the selected triangular prism, the second color signal corresponding to a position in the triangular prism indicated by the input first color signal A color signal interpolation method for calculating the output value of the first and second color signals by linearly interpolating the output value of each of the selected triangular prisms using the output value set at each vertex of the selected triangular prism. Using a data pair, the output value on the vertex is a variable, the coefficient of a linear function that is linear in the triangular prism is obtained by the least square method, and the output value calculated by the linear function using the coefficient is Setting each vertex Color signal interpolation method according to claim. 4. A correction coefficient of a linear or non-linear function is obtained by a least square method using a plurality of pattern data printed with a predetermined amount of ink and density data obtained by color separation of the plurality of patterns. A color correction method for setting Y, M, and C values calculated using the linear or non-linear function with the correction coefficient set at each grid point in the R, G, and B spaces. Y, M, and C values calculated using a linear function are set for grid points in the following areas, and grid points for areas where the density data is equal to or greater than a predetermined value are set as grid points.
A color correction method comprising setting Y, M, and C values calculated using a non-linear function. 5. A correction coefficient of a nonlinear function is obtained by a least squares method using a plurality of pattern data printed with a predetermined amount of ink and density data obtained by color separation of the plurality of patterns. In a color correction method for setting Y, M, and C values calculated using the nonlinear function in which a correction coefficient is set for each grid point in the R, G, and B spaces, an achromatic region or The correction coefficient of the nonlinear function is obtained by the least squares method using the data of the highlight area, and the Y, M, and C values calculated by the nonlinear function in which the correction coefficient is set are calculated as follows:
A color correction method, wherein the color correction method is set at a grid point of the area in the R, G, and B spaces.