JPH10276337A - Color image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To keep high image conversion accuracy without considerably increasing required memories by newly setting lattice points only to a range, where much higher accuracy is required or characteristics are changed by degradation with the passage of time, so as to improve accuracy within the range. SOLUTION: A color corrector 12 stores the relation between the image data of source image and the image data of copy image at the respective lattice points of three-dimensional lattice composed of plural solids dividing a color converting signal space as a look-up table and determines a color correct signal CMY by performing three-dimensional interpolation. A selector 14 judges whether or not the number of lattices is to be increased corresponding to the change of γ characteristics at an input device 11 and when that number is changed because of any change with the passage of time, the number of lattices is increased by setting the new lattice point at the center of changed range and outputted to the color corrector 12. Thus, color conversion accuracy can be improved in the range where the γ characteristics are rapidly changed.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a color image processing apparatus used for a color image forming apparatus such as a color copying machine or a color printer having a color correcting function.

[0002]

2. Description of the Related Art Conventionally, there is a color image forming apparatus such as a color copying machine or a color printer which reads a color original image by an internal or external reading means such as a color scanner and outputs a duplicate image of the color original image by an output means.

However, in the color image forming apparatus, even if the input image data is directly supplied to the output means, a duplicate image having a tone different from that of the original image is output in most cases. this is,
While the image input unit in color image formation creates image data by color separation processing by additive color mixture,
It is considered that this is because the image output unit performs the color forming process by subtractive color mixture. Therefore, in order to improve the reproducibility of the color tone in the duplicated image, it is necessary to correct the image data of the original image input from the image input unit and output it to the image output unit.

Therefore, as a color correction technique for obtaining a duplicate picture faithfully reproducing the color tone of the original picture, the Image Electronics Society of Japan, Vol. 29, No. 3, “Color Correction Technique for Faithful Color Reproduction”, A color correction process based on mathematical expression using a polynomial regression analysis and a color correction process based on non-formulaization using a three-dimensional interpolation or a neural network using a lookup table are disclosed. Japanese Patent Application Laid-Open No. 6-55779 discloses a configuration for correcting aging of printer characteristics. Further, Japanese Patent Application Laid-Open No. H4-221191 discloses a method in which a different bit number is allocated to each input signal in a table for storing a color conversion value, and a method of interpolating by setting a maximum number of green signals is disclosed in Japanese Patent Application Laid-Open No. H11-216191. I have.

[0005]

However, since there are variations even within the same coordinate axis, errors in the conversion accuracy increase depending on the grid used for the interpolation calculation. Also,
The conventional color correction technique cannot cope with a case where only a specific section of the coordinate axis has deteriorated as it is used.

An object of the present invention is to set a new grid point only in a range where higher accuracy is required or in a range in which characteristics have changed due to aging or the like.
It is an object of the present invention to provide a color image processing apparatus capable of improving the accuracy of the range and maintaining a high image conversion accuracy without significantly increasing the required memory.

[0007]

According to the present invention, there is provided a look-up table for storing a correspondence between a color separation signal and a color correction signal at each grid point of a three-dimensional grid for dividing a color space. A color image processing apparatus that performs a color correction process on the image data of the original image input from the image input unit with reference to the look-up table and outputs the processed data to an image output unit; It is characterized in that a lattice point changing process for changing the space coordinates is performed.

According to the first aspect of the present invention, the coordinates of the grid points in the look-up table that affect the accuracy of the color conversion process are changed, and the deterioration of the conversion accuracy due to the temporal change in the γ characteristic of the image input unit is corrected. Or a case where high-precision conversion processing is required for a specific color.

The invention described in claim 2 is characterized in that the grid point changing process is a process of increasing or decreasing the number of grid points in the look-up table.

According to the second aspect of the present invention, the number of grid points in the look-up table which affects the accuracy of the color conversion processing is increased or decreased, and the number of grid points is increased in a range where the conversion accuracy needs to be improved. By reducing the number of grid points in a range that does not need to be improved, it is possible to correct the conversion accuracy and improve the accuracy for a specific color without increasing the storage capacity of the lookup table.

The invention described in claim 3 is characterized in that the grid point changing process is a process of changing an interval between grid points of the lookup table.

According to the third aspect of the present invention, the interval between the grid points in the look-up table is changed to affect the accuracy of the color conversion processing, and the color conversion processing is continuously performed according to the γ characteristic of the image input unit. Accuracy can be changed.

According to a fourth aspect of the present invention, the grid point changing process includes a process of determining whether or not to change the color space coordinates of the grid point based on the detection result of the γ characteristic of the image input unit. It is characterized by.

According to the fourth aspect of the present invention, a decrease in color reproducibility due to a temporal change in the γ characteristic in the image input unit is corrected by changing the lattice points of the look-up table.

According to a fifth aspect of the present invention, the grid point changing process determines whether or not to change the color space coordinates of the grid points based on the history of the original image input from the image input unit. It is characterized by including.

According to the fifth aspect of the present invention, the grid points of the look-up table are changed in accordance with the frequency of appearance of the color in the original image, and the reproducibility of the color having a high appearance frequency can be improved.

According to a sixth aspect of the present invention, there is provided a neural network which uses, as learning data, image data of an original image included in one or a plurality of specific grids preset in the look-up table. The image data is subjected to a color correction process using the look-up table, and the image data of the original image in the specific grid is subjected to a color correction using a neural network.

In the invention described in claim 6, color correction by a neural network is performed instead of color correction based on a look-up table for colors in a specific grid set in advance, and colors in a specific range are further increased. Accurate color conversion processing can be performed.

The invention described in claim 7 is characterized in that the neural network includes image data of grid points of the specific grid, points on a boundary side, and points on a boundary surface in learning data. .

In the invention described in claim 7, since the image data of the boundary portion of the specific grid set in advance is included in the learning data of the neural network, the image data color-corrected based on the look-up table and The continuity of the color tone with the image data color-corrected by the neural network can be maintained.

[0021]

FIG. 1 is a block diagram showing a configuration of a color image processing apparatus according to an embodiment of the present invention.
The color image processing apparatus 1 outputs an image input device 11 to which image information of an original image is input, a color correction device 12 for correcting color data of the input image information, and image information including the color corrected color data. It comprises an image output device 13 for obtaining a duplicate image, and a selection device 14 for selecting whether or not to rewrite the contents of a look-up table stored in the color correction device 12.

The image input device 11 includes a color scanner as image reading means, scans a color original image, and separates R (red), G (green), and B (blue) color separation signals (hereinafter, referred to as R).
It is called a GB signal. Get) The color scanner converts the scanned original image into at least 256 colors for each of RGB.
Color separation into gradations is possible.

The image output device 13 includes a color printer as an output means, and outputs density signals of C (cyan), M (magenta), Y (yellow), and [K (black)] (hereinafter, CMY (K)). An image is obtained based on the signal. This color printer is CMY (K) 3 (4)
An image is output at a density of 256 gradations or more using color inks.

As shown in FIG. 3 (C), the color correction device 12 converts the image data of the original image and the duplicated image at each grid point of a three-dimensional grid formed by a plurality of cubes dividing the color conversion signal space. The relationship with the image data is stored as a look-up table, and the color correction signal CMY (CMYK) is determined by performing three-dimensional interpolation. This three-dimensional interpolation calculation will be described using a certain unit rectangular parallelepiped in the table shown in FIG. In FIG. 2, each vertex is represented by P1.
To P7, the relative coordinates of the point to be subjected to the interpolation calculation when the origin is P0 are (r, g, b), and the length of each side is L1, L
2, L3, the output value of each point is C (Pi), M (Pi), Y
(Pi). At this time, the output CMY signal is represented by the following equation.

[0025]

(Equation 1)

According to the above equations (1) to (3), the interpolation calculation can be performed for all the grids.

The selection device 14 determines whether or not it is necessary to increase or decrease the number of grids, and if determined to be necessary, outputs the value of the grid to be increased or decreased to the color correction device 12. I do.

FIG. 4 is a flowchart showing a basic processing procedure of the color image processing apparatus. First, an original image is scanned by the image input device 11, and original image data of 256 gradations for each of RGB colors is input (s1). Next, the selection device 14 determines whether or not it is time to check whether the number of grids needs to be changed (s2). Here, for example, it is checked whether or not the number of grids needs to be changed based on whether or not a predetermined number of duplicate images have been formed or whether or not the power has just been turned on.

If it is determined in s2 that it is time to check whether the number of grids needs to be changed,
It is checked whether the number of grids needs to be increased or changed (s3). If it is determined that the number of grids needs to be increased or changed, the grid to be changed and the table value of the grid point (color conversion) are determined. Calculate (coordinate values in the signal space) (s
4). Thereafter, or if it is determined in s2 that the timing is not to be checked, and if it is determined in s3 that the change in the number of grids is not necessary, the color correction device 12 performs color correction (s5). Then, a duplicate image is output by the image output device 13 (s6).

The following describes how to determine the necessity of changing the number of grids in the processes of s3 and s4, and how to change the number of grids.

In the color image processing apparatus according to the embodiment of the present invention, it is determined whether or not the number of grids should be increased based on a change in the γ characteristic in the input device 11. For example, if the initial component γ characteristic of the R component in the document reading unit of the input device 11 is in the state shown in FIG. 3A, the input characteristic of the R signal is linear. Look-up tables with multiple cubic grids are suitable for interpolation calculations.

In s3, first, a sample image in which each component of RGB changes linearly and continuously is read as an original image. At this time, the γ characteristic should ideally show a linear shape as shown in FIG. 3 (A), but if the characteristic changes due to aging, the linearity of the input signal is lost.
Assuming that the characteristic has a steep slope in a range surrounded by a broken line as shown in FIG. 3B, sufficient accuracy cannot be obtained in this range depending on the table shown in FIG. Therefore, in such a case, FIG.
As shown in (D), a new grid point is set at the center of the range where higher accuracy is required, and the number of grids is increased.

In s4, first, a color sample to be used as data for obtaining an output value of a grid point is output to the output device 13 (step S4).
Is output. At this time, when calculating the table value of the grid point to be newly increased, it is possible to output a color sample only around the value to be increased. For example, when one new grid point is set near R = 239, the range of 0 to 255 for G and B is divided by 9 into 243 (3 × 3) for three points of R = 223, 239, and 255.
The RGB signals for 9 × 9) colors are converted into CMY signals based on the lookup table before the change, and color samples are output. When one grid point is newly set and all table values are changed at the same time, 810 (10
(* 9 * 9) The RGB signals for the colors are converted into CMY signals based on the look-up table before conversion, and color samples are output.

At this time, the value supplied to the output device 13 is (D
_ci , D _mi , D _yi ), and the values obtained by reading the output color samples are (D _ri , D _gi , D _bi ), and table values are obtained by a linear masking method.

[0035]

(Equation 2)

A _ij is determined by the least square method so that the following holds. For example, a new grid point (239, 255, 2
55) is (D _ci , D _mi , D _yi ): D _r = 2
39, D _m = 255, D _y = 255, and can be obtained by substituting into the above equation (4).

By the above processing, γ of the image input device 11
By increasing the number of grids that divide the color conversion signal space in a range where the characteristic changes steeply (see FIG. 3D), the color conversion accuracy in that range can be improved.

It is also possible to determine whether or not the number of grids should be changed according to the frequency of use of colors in the color image processing apparatus 1. In this case, in s2, the necessity of changing the number of grids is checked every fixed number N,
Every time the original image is input, several tens of preset colors in the image or several tens representative colors in the image are extracted. The values of the RGB components of the extracted dozens of color data are counted, and after N sheets are input, the count number of each component is checked.
If there is a section with a large count, the section is determined to be frequently used, and the number of grids is increased. For example, if the histogram of the count of the R component after input of N sheets is in the state shown in FIG. 5A, it is determined that the frequency of use near the dotted line in FIG. 5A is high, and as shown in FIG. Increase the number of grid points. The table value of the new grid point can be determined by the method described above.

Further, it is possible to determine whether or not to increase the number of grids according to the conversion accuracy of each grid. In this case, if it is determined that a check is necessary in s2,
Output the color of the points inside each grid, for example, the point of the center of gravity.
At this time, let the value of the point of each center of gravity be ( _Dri , _Dgi , _Dbi ),
The value obtained by interpolating the center of gravity value is defined as (D _ci , D _mi , D _yi ). The output color sample is read by the image reading means, and the read value is defined as (D _ri ', D _gi ', D _bi '). In this case, if the color correction system is ideal, (D _ri , D
_gi , D _bi ) = (D _ri ′, D _gi ′, D _bi ′), but this is not true due to calculation errors, deterioration over time, and the like. Therefore, first, in order to make a determination in the R-axis direction, the error of each lattice is determined by ΔE = (D _ri ′ −D _ri ) ² + (D _gi ′ −D _gi ) ² + (D _bi ′ −D _bi ) ^2.・ ・ Calculate by (5).

_Assuming that the error of the (x, y, z) -th lattice from the origin (0, 0, 0) is ΔE _xyz , the n-th error from the origin of the R axis shown in FIG. expressed.

[0041]

(Equation 3)

If the value of E _Rn is within a predetermined allowable range, it is not necessary to increase the number of grids. Conversely, if the value is outside the allowable range, grid points are added inside. For example, as shown in FIG. 6, when the sum of errors (E _R5 ) of the fifth grid from the origin of the R axis exceeds the allowable range, grid points are added as shown in FIG. 5B. Similarly, the number of lattice points can be increased for the G axis and the B axis. The table value of the new grid point can be determined by the method described above.

If the γ characteristic changes from the state shown in FIG. 3B to the state shown in FIG.
In the range of 0, the value hardly changes, and in this range, very high conversion accuracy is not required. Therefore, FIG.
As shown in (B), the required memory capacity can be reduced by reducing the number of grids in a range where high conversion accuracy is not required. The process of decreasing the number of grids can be executed together with the process of increasing the number of grids in another range.

In addition, grid points for improving conversion accuracy are set in advance in a grid in a range including a specific color, and whether to use the grid points is selected according to conditions. FIG.
If all the grids are cubic as shown in (B),
In the equations (1) to (3), L1 = L2 = L3, and the calculation can be simplified. Therefore, for example, when the skin color is a specific color, the skin color region is (190, 1) in FIG.
20, 110), the number of pixels included in this area is counted. If the counted number of pixels of the skin color is within a preset threshold value, FIG.
The calculation is performed using the equally-spaced cubic lattice shown in FIG. 8B, and when the threshold is exceeded, interpolation calculation is performed by adding grid points set in advance as shown in FIG. In this way, for example, the conversion accuracy can be improved only when there are many specific colors that the human eyes are sensitive to, such as skin colors.

The lookup table used for the color conversion uses a cubic lattice at regular intervals. However, considering that the RGB signals input from the image input device 11 are often non-linear, first, FIG. 9A shows an evaluation sample for measuring the γ characteristic of the image input device 11 (for example, in order to measure the γ characteristic of the R signal, FIG.
The sample of (A) is made in red. ) Is read by the image input device 11, and as shown in FIG. 9 (B) or (C), a look-up table in which the interval of R is determined so that the output signal R 'or G' of the original reading means becomes constant. Can be used. By determining the interval between the signals in this manner, the interval is set narrower in a range where the change of the input value is sharp, and the interval is set wider in a range where the change of the input value is slow. As a result, color correction can be performed with the same accuracy on any grid, and it is not necessary to γ-correct the output signal of the image input device 11 in advance.

Further, the grid interval of the look-up table can be determined according to the magnitude of the conversion error. That is, the value of the lookup table is often calculated by a linear masking method or a neural network,
Even with such a calculation method, a certain degree of conversion error occurs. Therefore, for example, when the table value is determined by the linear masking method of the above equation (4), consider an area 61 shown in FIG. 10, and let D _x 'be a representative ideal value of several to several tens of points in this area 61. , when the calculated value D _x by equation (4),

[0047]

(Equation 4)

Is an error in the area 61. By performing this calculation for the entire range of the R signal, for example, a conversion error with respect to the R signal shown in FIG. When this conversion error is integrated by R, FIG.
The accumulation error for the R signal shown in FIG. The grid width of the R component in the look-up table is determined so that the interval of the accumulated error is constant. By determining the grid width in the lookup table in this way,
In a range where the conversion error is large, the grid width can be set small, and in a range where the conversion error is small, the grid width can be set large, and the same conversion accuracy can be obtained in all the grids.

Further, the grid width of a range including a specific color in the lookup table can be set small. For example, a region including a skin color requiring high conversion accuracy (for example, (190, 120, 110)).
Is represented in the lookup table as shown in FIG. 12A, and the grid width is determined according to the distance from this area. That is, as shown in FIG. 12B, in the lookup table, the R axis is narrow near R = 190, and wide near R = 0 and R255. Thereby, the color conversion accuracy of the specific color can be increased.

The processing for setting the grid width in the look-up table as described above can also be used for correcting a change over time in characteristics in the color image forming apparatus 1. That is, the evaluation sample image shown in FIG. 9A is read together with the original image, and the grid width of the lookup table is calculated so that the interval between the output signals of the image reading unit is constant. At this time, the calculated grid width is compared with the previous grid width, and when the difference between the two exceeds the threshold, the grid width is changed to the newly calculated grid width. Thus, for example, when the γ characteristic of the image reading unit changes from the state shown in FIG. 13A to the state shown in FIG. 13B, the grid width of the lookup table is changed as shown in FIG. The state is changed from the state to the state shown in FIG.

Each time the original image is read by the image input device 11, the number of pixels used in the original image is counted, and a histogram for each of the RGB colors is created. The grid width can also be corrected. That is, if the grid width obtained by integrating the histogram created each time the predetermined number of reading processes is completed has changed greatly from the previous grid width, a new grid width may be set. it can. By this processing, for example, if the histogram of the R component is in the state shown in FIG. 14A, the histogram can be integrated as shown in FIG. 14B to correct the grid width of the lookup table. .

Further, a representative color sample uniformly distributed over the entire color space, for example, 9 pixels for each color of RGB.
The grid width of the look-up table can be corrected by reading a total of 729 (9 × 9 × 9) color samples as reference images. That is, the read value of the color sample as the reference image is represented by (D _ri ', D _gi ', D
_{As bi} ′), the color is corrected based on the current look-up table, and the duplicate image is output via the output device. This duplicate image is read again, and the read value at this time is (D _ri , D _gi , D
_bi ). Error between the two readings

[0053]

(Equation 5)

As a result, for example, the region 6 in FIG.
Calculated as the error at 1. This calculation is, for example, R
By running over the full range of the signal, FIG.
A conversion error with respect to the R signal shown in FIG. 1A is obtained, and this conversion error is integrated with R to obtain an accumulation error with respect to the R signal shown in FIG. The grid width of the R component in the look-up table is determined so that the interval of the accumulated error is constant. Compare the grid width in the look-up table determined in this way with the previous grid width,
If the difference between the two exceeds the threshold, a new grid width is set.

In addition, by reading the evaluation sample shown in FIG. 9A as the reference image, the table value of each grid point of the lookup table whose grid width has been changed may be corrected. That is, the read value of the color sample as the reference image is (D _ri ′, D _gi ′, D _bi ′), and the read value is color-corrected based on the lookup table at that time, and is output via the output device. Output a duplicate image.
This duplicate image is read again, and the read value is changed to (D _ri ,
D _gi , D _bi )

[0056]

(Equation 6)

In

[0058]

(Equation 7)

A _{ij is calculated} by the least squares method so that
Determine the value of. The table value of each grid point after changing the grid width is determined using the equation (9) determined by the equation (10).

FIG. 15 is a diagram showing a configuration of a color image processing apparatus according to another embodiment of the present invention. The color image processing device 21 includes a designation device 24 instead of the selection device 14 of the color image processing device 1 shown in FIG. 1 and includes a color correction device 22 different from the color correction device 12. The specification device 24 is a device for externally specifying color data to be subjected to color correction by a neural network.

FIG. 16 is a diagram showing a configuration of a color correction device included in the color image processing device. Color correction device 22
Is 256 input from the scanner of the image input device 11.
A discrimination unit 31 for selecting which of a neural network and a table interpolation method to perform color correction on the RGB signals of gradation, a neural network operation unit 32 for performing color correction using the neural network, and a neural network The storage unit 33 includes a storage unit 33 that stores a learning result, and a table interpolation calculation unit 34 that performs color correction by table interpolation based on a lookup table.

The color correction device 22 includes a table interpolation calculator 3
In FIG. 4, each grid point of a grid formed by dividing the inside of a cube having each of the three input values of RGB as axes as shown in FIG. The color correction signal CMY (K) signal is determined by performing three-dimensional interpolation. Here, a grid in the look-up table 41 including a color area that needs high-precision color correction in order to be sensitively recognized by human eyes due to low saturation such as skin color, light blue or green is previously specified as a specific grid. The data such as the grid coordinates of the specific grid is stored in the color correction device 22. At the same time, a color correction neural network using the color data in the specific grid as learning data is formed for each grid or each grid group described later, and the data is stored in the storage unit 36.

From the RGB signal components which have been decomposed into 256 gradations inputted from the input device 11,
The determination unit 31 in the color correction device 22 determines whether the signal is color data in a specific grid. RG input
When it is determined that the B signal is color data in a specific grid, the neural network operation unit 32 performs color correction using a neural network, and when it is determined that there is color data outside the specific grid, the table interpolation calculation unit 34 performs the color correction. By performing color correction by an interpolation method such as 8-point interpolation, CM
Determine the Y (K) signal.

By the above processing, the RGB signal specified in advance as a color requiring high-precision color correction is
By performing the color correction processing using the neural network, it is possible to perform color correction with a small interpolation error.

The color data in the grid including the color data of the points at the boundary such as the grid point 43 of the specific grid 42, the point 44 on the boundary side, and the point 45 on the boundary surface is converted into the neural network operation unit. By including the color correction data at the boundary of the specific grid 42 in the learning data of the neural network for color correction of the specific grid at 32, the color correction processing by the neural network reflecting the color correction data of the point at the boundary can be performed. , Table interpolation calculator 3
4, the continuity between the color-corrected image data and the color-corrected image data by the neural network operation unit 32 can be maintained.

In the case where a plurality of specific grids are located close to each other in the look-up table 41, the plurality of specific grids are used as one grid group 46 on condition that the color correction accuracy does not decrease. Accordingly, it is not necessary to form a neural network by duplicating color data of a boundary portion common to a plurality of adjacent specific grids, thereby reducing the amount of memory used in the storage unit 36 and shortening the time required for color correction processing. can do.

Further, according to the processing shown in the flowchart of FIG. 18, the input RGB signal is judged to be the color data in the specific lattice by the discriminating section 31 in the color correcting device 12, and Only when the color data is within the specified grid specified from the above, the color correction by the neural network is executed by the neural network operation unit 32 (s12 → s13).
→ s14), the time required for the color correction process can be further reduced by performing the color correction by the neural network only on the color image that really needs high-precision color correction. In addition, by setting a specific grid even in an achromatic region in the look-up table, it is possible to reproduce gray scale characteristics of a black-and-white image with high accuracy.

[0068]

According to the first aspect of the present invention, by changing the coordinates of the grid points in the look-up table which affects the accuracy of the color conversion processing, the conversion due to the temporal change of the γ characteristic of the image input unit is achieved. To compensate for the loss of accuracy,
It is possible to cope with a case where high-precision conversion processing is required for a specific color.

According to the second aspect of the present invention, by increasing or decreasing the number of grid points in the look-up table which affects the accuracy of the color conversion processing, the number of grid points can be increased in a range where the conversion accuracy needs to be improved. In addition, it is possible to reduce the number of grid points in a range where the conversion accuracy does not need to be improved, and to correct the conversion accuracy and improve the accuracy for a specific color without increasing the storage capacity of the lookup table.

According to the third aspect of the present invention, the γ of the image input unit is changed by changing the interval between the grid points in the look-up table which affects the accuracy of the color conversion processing.
The accuracy of the color conversion processing can be continuously changed according to the characteristics.

According to the fourth aspect of the present invention, it is possible to correct a decrease in color reproducibility due to a temporal change in the γ characteristic in the image input unit by changing the lattice points of the look-up table.

According to the fifth aspect of the present invention, the reproducibility of a color having a high frequency of appearance can be improved by changing the lattice points of the lookup table in accordance with the frequency of color appearance in the original image.

According to the sixth aspect of the present invention, a color in a specific range of colors within a specific range is corrected by performing a color correction using a neural network instead of a color correction based on a look-up table with respect to colors in a specific grid set in advance. Further, a highly accurate color conversion process can be performed.

According to the seventh aspect of the present invention, the image data at the boundary of the specific grid set in advance is included in the learning data of the neural network, so that the color-corrected image data based on the look-up table is obtained. The continuity of the color tone between the image data and the image data color-corrected by the neural network can be maintained.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating a configuration of a color image processing apparatus according to an example of an embodiment of the present invention.

FIG. 2 is a diagram illustrating table interpolation processing in the color image processing apparatus.

FIG. 3 is a diagram illustrating a correction process of a lookup table in the color image forming apparatus.

FIG. 4 is a flowchart showing a processing procedure in a lookup table correction process in the color image forming apparatus.

FIG. 5 is a diagram illustrating a lookup table correction process for increasing the number of grids according to the frequency of use of colors in the color image forming apparatus.

FIG. 6 is a diagram illustrating a lookup table correction process for increasing the number of grids in accordance with the conversion accuracy of each grid in the color image forming apparatus.

FIG. 7 is a diagram illustrating a lookup table correction process for increasing or decreasing the number of grids according to a change in the γ characteristic of the image input device in the color image processing device.

FIG. 8 is a diagram illustrating a correction process of a lookup table for increasing the number of grids in a range including a specific color in the color image processing apparatus.

FIG. 9 is a diagram illustrating a process of changing a grid interval of a lookup table using an evaluation sample in the color image processing apparatus.

FIGS. 10 and 11 are diagrams illustrating a process of changing a grid interval of a lookup table based on an error in a color conversion process in the color image processing apparatus. FIGS.

FIG. 12 is a diagram illustrating a process of changing a grid interval of a lookup table centering on a range including a specific color in the color image processing apparatus.

FIG. 13 is a diagram illustrating a process of changing a grid interval of a lookup table based on a change in γ characteristics of an image input device in the color image processing device.

FIG. 14 is a diagram illustrating a process of changing a grid interval of a lookup table based on the frequency of appearance of colors in the color image processing apparatus.

FIG. 15 is a diagram illustrating a configuration of a color image processing apparatus according to another embodiment of the present invention.

FIG. 16 is a diagram illustrating a configuration of a color correction device included in the color image processing device.

FIG. 17 is a diagram showing a grid and a grid group in a lookup table to be subjected to color correction processing by a neural network in the color image processing apparatus.

FIG. 18 is a flowchart illustrating a color correction processing procedure in the color image processing apparatus.

[Explanation of symbols]

 11-input device 12-color correction device 13-output device 14-selection device

──────────────────────────────────────────────────の Continued on the front page (51) Int.Cl. ⁶ Identification symbol FI H04N 1/407 G06F 15/66 310 1/46 H04N 1/40 101E 1/46 Z (72) Inventor Tatsuya Tanaka Osaka-shi, Osaka (22) Inventor Takahiro Odo 22-22 Nagaikecho, Abeno-ku, Osaka-shi, Osaka

Claims (7)

[Claims] A lookup table for storing a correspondence between a color separation signal and a color correction signal at each grid point of a three-dimensional grid that divides a color space; and referring to the lookup table from an image input unit. In a color image processing apparatus that executes color correction processing of input image data of an original image and outputs the result to an image output unit, the color image processing apparatus may perform a grid point changing process of changing a color space coordinate of a grid point of the lookup table. Characteristic color image processing device. 2. The color image processing apparatus according to claim 1, wherein the grid point changing process is a process of increasing or decreasing the number of grid points in the look-up table. 3. The process according to claim 1, wherein the grid point changing process is a process for changing an interval between grid points in the lookup table.
3. The color image processing apparatus according to 1. 4. The grid point changing process according to claim 1, further comprising the step of determining whether or not to change the color space coordinates of the grid point based on the detection result of the γ characteristic of the image input unit. 3. The color image processing apparatus according to 1. 5. The grid point changing process according to claim 1, further comprising a step of determining whether or not to change the color space coordinates of the grid points based on the history of the original image input from the image input unit.
The color image processing device according to any one of the above. 6. A neural network which uses image data of an original image included in one or a plurality of specific grids preset in the lookup table as learning data. 6. The color image processing apparatus according to claim 1, wherein a color correction process is performed using a table, and color correction is performed on the image data of the original image in the specific grid using a neural network. 7. The color image processing apparatus according to claim 6, wherein the neural network includes, as learning data, image data of a lattice point of the specific lattice, a point on a boundary side, and a point on a boundary surface.