JP2007189487A - Image processor - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image processor for performing the flexible color correction of an image at high speed. <P>SOLUTION: The image processor for for automatically correcting colors of an image represented in a digital color space having N colors includes: a memory; a general processor which derives (n) grid correction color data representing correction colors corresponding to (n) (n<N) grid colors to generate a three-dimensional color conversion table holding the grid correction color data by the grid colors in the memory; and a dedicated processor which acquires original color data stored in the memory, acquires unit data composed of a plurality of grid correction color data corresponding to a plurality of grid colors close to the color of a pixel of an interest from the three-dimensional color conversion table based upon the original color data, and derives correction color data corresponding to the original color data by linear interpolation operation using the unit data to generate a corrected image in which colors of respective pixels are represented with the correction color data. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image processing apparatus, and more particularly to color correction of an image using a three-dimensional color conversion table.

  In color correction of photographic images, various color conversions such as dynamic range improvement by tone curve conversion and level conversion, brightness and contrast adjustment, color balance adjustment for color cast removal, and color conversion for specific hues, etc. Done. Both of these color conversions and color conversions combining them are data (hereinafter referred to as correction color data) in which data representing the color before conversion (hereinafter referred to as original color data) represents another color according to some conversion rule. ). Conventionally, color correction using a three-dimensional color conversion table that holds corrected color data representing a corrected color corresponding to a color before correction is known. In color correction using a three-dimensional color conversion table, correction color data representing the color resulting from correcting the color of the target pixel can be acquired simply by reading the correction color data from the memory for each target pixel. Color correction is possible. In color correction using a three-dimensional color conversion table, it is not necessary to obtain a correction result for each color component, so that color correction can be performed at a higher speed than color correction using a one-dimensional color conversion table. However, when correction color data corresponding to all colors is held in the three-dimensional color conversion table, the data amount of the three-dimensional color conversion table becomes enormous. In Patent Document 1, correction color data is held in a three-dimensional color conversion table only for colors corresponding to grid points in the color space (hereinafter, colors corresponding to grid points in the color space are referred to as grid colors) (hereinafter referred to as “color colors”). The correction color data held in the three-dimensional color conversion table is referred to as grid correction color data.), And a technique for obtaining correction color data corresponding to an arbitrary color other than the grid color by linear interpolation is disclosed.

  By the way, in order to automatically correct the color of a photographic image, for example, it is necessary to adjust the color balance according to the color temperature of the illumination light or to adjust the luminance according to the brightness of the photographic image for a failed photograph with color cast. is there. In order to use the 3D color conversion table for the purpose of automatically correcting the color according to the image quality of the photographic image, it is necessary to prepare in advance a large number of 3D color conversion tables that can be selected according to the image quality of the photographic image. There is. However, it is not realistic to hold a large number of three-dimensional color conversion tables with a large amount of data in the image processing apparatus under conditions where the storage capacity is limited. In addition, the design change is not easy because the three-dimensional color conversion table needs to be updated in accordance with the change of the correction algorithm and the change of the correction amount.

JP-A-3-244292

  The present invention has been created in view of the above problems, and an object thereof is to provide an image processing apparatus capable of performing flexible color correction of an image at high speed.

(1) An image processing apparatus for achieving the above object is an image processing apparatus for correcting the color of an image represented in a digital color space having N colors, and includes a memory, n (n <N) N grid correction color data representing correction colors corresponding to the grid colors are derived, and a three-dimensional color conversion table that holds the grid correction color data for each grid color is generated in the memory. A plurality of grid correction color data corresponding to a plurality of grid colors that are close to the color of the target pixel, and obtains original color data representing the color of the target pixel of the original image stored in the memory; Unit data configured is acquired from the three-dimensional color conversion table based on the original color data, and linear interpolation is performed using the unit data acquired from the three-dimensional color conversion table. Correction color data corresponding to the original color data is derived, and the correction color data for the pixel of interest is stored in the memory. By repeating a series of processes, the color of each pixel is represented by the correction color data. A dedicated processor for generating an image.
According to the present invention, since the general-purpose processor derives grid correction color data representing a correction color corresponding to the grid color, an arbitrary three-dimensional color conversion table can be generated. That is, the image processing apparatus according to the present invention can freely generate an arbitrary three-dimensional color conversion table according to the image quality of the correction target image and the target image quality. Color correction of an image using a three-dimensional color conversion table is performed by acquiring original color data representing the color of the target pixel, specifying an address of grid correction color data representing a grid correction color close to the color of the target pixel, A series of processes of acquisition, linear interpolation calculation, and correction color data storage are repeated. These series of processing functions can be realized by a relatively small circuit having a small number of stages. Further, when these series of processing functions are realized by hardware, when the data line width is ignored, the memory access includes acquisition of original color data, acquisition of grid correction color data, and storage of correction color data. Times. That is, if these series of processing functions are realized by hardware, the number of memory accesses necessary for the interpolation calculation is drastically reduced. Therefore, in the present invention, these series of processing functions are realized by a dedicated processor to speed up the color correction of the image. Therefore, the image processing apparatus according to the present invention can perform flexible color correction of an image according to the image quality of the original image to be corrected and the target image quality at high speed.

(2) The general-purpose processor may analyze the original image and derive each grid correction color data corresponding to each grid color based on the analysis result of the original image.
Analyzing the original image with a general-purpose processor and deriving grid correction color data corresponding to the grid color according to the analysis result, thereby generating a three-dimensional color conversion table according to the image quality of the original image to be corrected and the target image quality can do.

(3) The general-purpose processor may analyze the original image in a plurality of color spaces.
By analyzing the original image in a plurality of color spaces, it is possible to precisely analyze the image quality of the original image. Specifically, for example, by analyzing the original image in the luminance space, RGB color space, and HSB color space, the average luminance, color balance, average saturation, and the like of the original image can be individually analyzed. Therefore, by analyzing the original image in a plurality of color spaces and deriving each grid correction color corresponding to each grid color based on the analysis result, a precise three-dimensional color conversion table that faithfully matches the image quality of the original image Can be generated.

(4) The dedicated processor derives the correction color data by hexahedral interpolation calculation as the linear interpolation calculation in the hexahedral interpolation calculation mode, and by tetrahedral interpolation calculation as the linear interpolation calculation in the tetrahedral interpolation calculation mode. The correction color data may be derived. The general-purpose processor may control switching between the hexahedral interpolation calculation mode and the tetrahedral interpolation calculation mode.
In the hexahedral interpolation calculation, it is possible to generate a corrected image with a smooth gradation compared to the tetrahedral interpolation calculation. In the tetrahedral interpolation calculation, color twist due to color conversion can be suppressed as compared with the hexahedral interpolation calculation. Thus, the conversion characteristics are different between the hexahedral interpolation calculation and the tetrahedral interpolation calculation. According to the present invention, the dedicated processor can derive the correction color data by both the hexahedral interpolation calculation and the tetrahedral interpolation calculation, and the general-purpose processor operates in the hexahedral interpolation calculation mode or the tetrahedral interpolation calculation mode. Can be controlled. According to this image processing apparatus, the color of the image can be corrected by properly using the hexahedral interpolation calculation and the tetrahedral interpolation calculation in accordance with the target image quality of the corrected image.

(5) The general-purpose processor stores the unit data composed of the grid correction color data corresponding to the eight grid colors adjacent to each other at different consecutive addresses of the memory to store the three-dimensional color. A conversion table may be generated. The dedicated processor includes data acquisition means for simultaneously acquiring a plurality of grid correction color data corresponding to one unit data from a plurality of consecutive addresses in the memory, and in the hexahedral interpolation calculation mode, Interpolation calculation means for deriving the correction color data by hexahedral interpolation calculation.
According to the present invention, the dedicated processor can acquire eight pieces of grid correction color data necessary for deriving correction color data by hexahedral interpolation calculation at high speed from the memory. It can be corrected.

(6) The interpolation calculation means sequentially acquires the eight grid correction color data constituting the unit data from the data acquisition means in the hexahedral interpolation calculation mode and the tetrahedral interpolation calculation mode, A coefficient corresponding to the grid correction color data acquired from the data acquisition unit is generated based on the original color data, and the eight grid correction color data constituting the unit data are weighted with the corresponding coefficient. Then, the correction color data is derived by accumulating each component. The interpolation calculation means generates 0 as the coefficient corresponding to the four grid correction color data determined by the original color data among the acquired unit data in the tetrahedral interpolation calculation mode. The general-purpose processor controls switching between the hexahedral interpolation calculation mode and the tetrahedral interpolation calculation mode.
In the image processing apparatus according to the present invention, the dedicated processor can derive the correction color data by both the hexahedron interpolation calculation and the tetrahedron interpolation calculation, and the general-purpose processor can select the dedicated processor in any of the hexahedron interpolation calculation mode and the tetrahedron interpolation calculation mode. Since the operation can be controlled, the color of the image can be corrected at high speed by either the hexahedral interpolation calculation or the tetrahedral interpolation calculation according to the target image quality of the corrected image.
Eight grid correction color data are required for the hexahedral interpolation calculation. Four pieces of grid correction color data are required for the tetrahedral interpolation calculation. Therefore, if there are 8 grid correction color data corresponding to 8 grid colors adjacent to each other, the color area of the hexahedron having these 8 grid colors as vertices (hereinafter, 8 grid colors as vertices). Correction color data corresponding to any color included in the hexahedral color area can be derived by either hexahedral interpolation calculation or tetrahedral interpolation calculation. The circuit scale of the dedicated processor can be reduced by promoting the common use of the circuit operating in the hexahedral interpolation calculation mode and the circuit operating in the tetrahedral interpolation calculation mode. According to the present invention, since the dedicated processor generates 0 as a coefficient corresponding to the four grid correction color data determined by the original color data, the eight grid corrections are performed in both the hexahedral interpolation calculation mode and the tetrahedral interpolation calculation mode. Color data can be acquired and 8 grid correction color data can be multiplied by 8 coefficients. Therefore, according to the present invention, it is possible to reduce the manufacturing cost of a dedicated processor capable of performing hexahedral interpolation calculation and tetrahedral interpolation calculation.

(7) A cumulative addition circuit for deriving the correction color data by weighting the grid correction color data acquired from the data acquisition unit with the corresponding coefficient and performing cumulative addition for each component. A comparison circuit that compares the magnitude relationship of the lower bit patterns of each component of the original color data representing the color of the target pixel, and the lower bit pattern of each component of the original color data representing the color of the target pixel. And a coefficient generation circuit for generating the coefficient corresponding to the grid correction color data acquired by the cumulative addition circuit. In the hexahedral interpolation calculation mode, the coefficient generation circuit performs all the acquisition by the cumulative addition circuit regardless of the magnitude relation of the lower bit pattern of each color component of the original color data representing the color of the target pixel. The coefficient corresponding to the grid correction color data is generated, and in the tetrahedral interpolation calculation mode, the original color data representing the color of the target pixel among the one unit data acquired by the cumulative addition circuit. 0 is generated as the coefficient corresponding to the four pieces of grid correction color data determined according to the magnitude relation of the lower bit pattern of each color component.
A unit that is a hexahedron region can be divided into six tetrahedron regions that share one diagonal of the unit. Depending on which tetrahedron region in the unit the color of the pixel of interest is included in, four grid correction color data that should have a weight of 0 in the tetrahedron interpolation calculation mode are determined. The tetrahedron region in the unit including the color of the pixel of interest is determined according to the magnitude relationship of the lower bit patterns of each component of the color of the pixel of interest. According to the present invention, the dedicated processor generates 0 as a coefficient corresponding to the four grid correction color data according to the result of comparing the magnitude relations of the lower bit patterns of the components of the color of the target pixel. It is possible to generate 0 as a coefficient of the four grid correction color data whose weight should be 0 in the interpolation calculation mode.

(8) The dedicated processor may operate with wired logic.
A dedicated processor operating with wired logic can generate a corrected image at a higher speed than a dedicated processor operating with a microprogram control system.

(9) The dedicated processor may derive an approximate value of a division result by a combination of a shift operation circuit and a multiplication circuit.
Division is essential for linear interpolation operations. Division by 2 ⁿ can be realized by a simple shift operation circuit, but the number of circuit stages required for division by a number other than 2 ⁿ increases. Division by an arbitrary number can be approximated by a combination of division and multiplication by 2 ⁿ . Therefore, the number of stages of the dedicated processor circuit can be reduced by deriving an approximate value of the division result by combining the shift operation circuit and the multiplication circuit in the dedicated processor.

Embodiments of the present invention will be described in the following order.
[Hardware configuration of image processing apparatus]
[Software configuration of image processing device]
[3D color conversion table]
・ ・ Analysis of original image and setting of correction parameters ・ ・ Derivation of grid correction color data corresponding to grid color ・ ・ Configuration of 3D color conversion table [Configuration and operation of color conversion block]
・ ・ Overall structure ・ ・ Data transfer ・ ・ Interpolation calculation ... tetrahedral interpolation ... hexahedral interpolation ・ ・ Dither processing [Flow of image correction]
[Other Embodiments]

[Hardware configuration of image processing apparatus]
FIG. 2 is a block diagram showing a schematic configuration of the printer 1 to which the present invention is applied. The printer 1 is a so-called stand-alone printer that reads a photographic image of a general format such as JPEG or RAW data from the removable memory 10 and can automatically print a photographic image whose color is corrected. The printer 1 directly inputs images in a general format or RAW data from an external system such as a digital camera 30, a PC (Personal Computer) 32, a camera-equipped mobile phone terminal 34, etc., and prints a photographic image whose color is automatically corrected. Is possible.

  The external interface (IF) 20 includes a USB controller, a USB connector, and the like for communicating with an external system such as the digital camera 30, the PC 32, and the camera-equipped mobile phone terminal 34. Input parameters to the printer 1.

  A removable memory controller (RMC) 12 is connected to the removable memory 10 via a connector (not shown), controls data transfer between the removable memory 10 and the RAM 14, and inputs images and RAW data in a general format to the printer 1. .

  The image processing unit 16 performs processing for generating print data from an image to be printed such as JPEG decoding, color correction using a three-dimensional color conversion table, sharpness correction processing, color separation processing, halftoning, interlace processing, and the like. And an image processing LSI or DSP for high-speed execution in cooperation with

A color conversion block 17 (see FIG. 3) constituting a part of the image processing unit 16 is a dedicated processor for color correction using a three-dimensional color conversion table. The color conversion block 17 performs color correction of the image using the three-dimensional color conversion table generated by the CPU 22 on the original image specified by the CPU 22, thereby generating a corrected image in which the color is automatically corrected. Following description, it is assumed that the color of the original image is also represented by the RGB digital color space colors 256 ^3-color corrected image, the number of colors of the color space of the original image, nor the number of colors of the corrected image color space, It is not limited to 256 ^three colors. Further, the color space color system is not limited to the RGB color system, and any color system may be used.

  The printing unit 18 includes a recording head for forming an image on a sheet by an inkjet method based on print data, a reciprocating mechanism of the recording head, a paper supply / discharge mechanism, and the like. As the printing method, any printing method such as an ink jet method, a laser method, a thermal method, or a dot impact method can be adopted.

The operation unit 26 includes various buttons for accepting user menu operations, print setting requests, and print start requests. When a specific button is pressed in a specific mode, various requests corresponding to the mode are input to the printer 1.
The display unit 28 includes an FPD (Flat Panel Display) such as an LCD for displaying menus and images to be printed, a graphic controller, and the like.

The RAM 14 is a volatile storage medium that temporarily stores a program and data to be processed by the program, such as an original image, a corrected image, and a three-dimensional color conversion table.
The CPU 22 as a general-purpose processor controls each part of the printer 1 by executing a control program stored in the flash memory 24 to control printing. The control program may be transferred from a computer-readable storage medium to the flash memory 24, or may be transferred from a remote server to the flash memory 24 via a network.

[Software configuration of image processing device]
FIG. 4 is a block diagram showing a configuration of a control program executed by the CPU 22.
The print control module 40 performs JPEG decoding, color correction, and color separation processing in accordance with print setting parameters set by an external system such as the digital camera 30 connected to the operation unit 26 or the printer 1, the PC 32, and the mobile phone 34. This is a program component for controlling execution of processing for generating print data such as halftoning and interlace processing.

  The color correction control module 42 is a program component for controlling the color correction module 46 based on the original image to be printed and the print setting parameters passed from the print control module 40. Specifically, the color correction control module 42 passes correction parameters to the correction method determination module 44 and determines which of the S / W color correction module 52 and the H / W color correction module 54 is to execute color conversion. Depending on the determination result, whether to perform color conversion by either the S / W color correction module 52 or the H / W color correction module 54 is set. A mode in which the S / W color correction module 52 is executed by the CPU 22 is a continuous correction mode. A mode in which the H / W color correction module 54 is executed by the CPU 22 is an interpolation correction mode. The color correction control module 42 causes the CPU 22 to function as mode setting means.

  The color correction module 46 is a program component for generating a corrected image by correcting the color of the original image to be printed. A correction parameter generation module 48 constituting a part of the color correction module 46 is a program component for analyzing the original image and setting correction parameters for color correction based on the analysis result. The correction parameter generation module 48 causes the CPU 22 to function as analysis means. The table generation module 50 is a program component for generating a three-dimensional color conversion table based on the correction parameters. The table generation module 50 causes the CPU 22 to function as a table generation unit. The S / W color correction module 52 is a program component for correcting the color by the CPU 22 based on the correction parameter. The S / W color correction module 52 causes the CPU 22 to function as a continuous correction unit. The H / W color correction module 54 is a program component for causing the color conversion block 17 to correct colors using a three-dimensional color conversion table. That is, the H / W color correction module 54 is a program component for the CPU 22 to control the color conversion block 17. The H / W color correction module 54 causes the color conversion block 17 to function as an interpolation correction unit.

  The correction method determination module 44 is a program component for verifying the correction parameter and the three-dimensional color conversion table against the parameter reference and the table reference, and determining which of the CPU 22 and the color conversion block 17 should perform color correction. is there. The correction method determination module 44 causes the CPU 22 to function as a parameter determination unit and a table determination unit.

[Generation of three-dimensional color conversion table]
.. Outline of 3D color conversion table The 3D color conversion table is corrected for each grid color (Gr, Gg, Gb) which is the color of the grid point in the RGB color space shown in FIG. It is a data structure that holds grid correction color data (Pr, Pg, Pb) (see FIG. 5B) representing the resulting color. Since the three-dimensional color conversion table is generated for each photographic image by the CPU 22, the amount of change before and after correction determined by the three-dimensional color conversion table is changed according to the image quality of the photographic image.

.. Analysis of original image and setting of correction parameters In this embodiment, a plurality of correction parameters are set to automatically correct the color of a photographic image. Factors that determine the image quality of a photographic image include exposure, light source color temperature, contrast, image sensor characteristics of the digital camera, and the color of the subject itself. In order to correct the image quality of a photographic image, it is necessary to specify these elements by analyzing the original image and set correction parameters according to these elements. These elements are specified by analyzing the original image in a plurality of color spaces. The analysis of the original image and the setting of correction parameters are performed by the CPU 22 that executes the correction parameter generation module 48 as follows.

  A luminance correction parameter that is a correction parameter for correcting the luminance determined by the exposure, the color of the subject, and the like is set according to, for example, an average value of luminance. Shadow parameters and highlight parameters, which are correction parameters for correcting the contrast determined according to the color of the subject, the intensity of the light source, and the like, are set according to the highlight point and the shadow point, respectively. In order to set the brightness correction parameter, the shadow parameter, and the highlight parameter, a histogram is generated for the brightness that is the pixel value of the QVGA size gray-tone image, and the average value, the maximum value, and the minimum value of the brightness are respectively based on the histogram. Analyzed. A QVGA size graytone image is generated by sampling the original image. A gray-tone image is a one-component image that represents the gradation level of luminance for each pixel. A difference between the average value and a predetermined fixed value is set as a luminance correction parameter. The brightness correction parameter corresponds to R (Red), G (Green), B (Blue), and a width for moving up and down two control points for each tone curve. A difference between the maximum value and a predetermined fixed value is set as the highlight parameter. The highlight parameter corresponds to a width for raising and lowering one point on the highlight side among two control points for each tone curve of R (Red), G (Green), and B (Blue). A difference between the minimum value and a predetermined fixed value is set as a shadow parameter as a contrast correction parameter. The shadow parameter corresponds to a width for moving up and down one point on the shadow side among the two control points for each tone curve of R (Red), G (Green), and B (Blue).

  A memory color correction parameter, which is a correction parameter for correcting a memory color determined by the subject color, light source color temperature, image sensor characteristics, and the like, is determined according to the average value of each memory color. For setting the memory color correction parameters, R (Red), G (Green), and B (Blue) for a pixel group whose hue is similar to the hues of leaf green, sky blue, red, and flesh color among the RGB tone images of QVGA size. ) Each histogram is generated for each hue, and R (Red), G (Green), and B (Blue) of the pixel group whose hue approximates the hues of leaf green, sky blue, red, and skin color based on each histogram. The average value of is analyzed. A QVGA size RGB tone image is generated by sampling the original image. An RGB tone image is a three-component image representing the gradation levels of R (Red), G (Green), and B (Blue) for each pixel. The difference between each average value and a predetermined fixed value is set as a leaf green memory color correction parameter, a sky blue memory color correction parameter, a red memory color correction parameter, and a skin color memory color correction parameter. Each memory color correction parameter corresponds to a width for moving up and down two control points for each RGB tone curve.

  The color balance correction parameter, which is a correction parameter for correcting the color balance determined according to the light source color temperature, is an average value of each of R (Red), G (Green), B (Blue), and luminance specified by the analysis. It depends on it. In order to set the color balance correction parameters, R (Red), G (Green), and B (Blue) histograms are generated for the QVGA size RGB tone image, and a luminance histogram is generated for the QVGA size gray tone image. The average values of luminance, R (Red), G (Green), and B (Blue) are analyzed based on each histogram. The difference between the luminance average value and the average value of R (Red), the difference between the luminance average value and the average value of G (Green), and the difference between the luminance average value and the average value of B (Blue) are weighted. Set as a balance correction parameter. The color balance correction parameter corresponds to a width for moving up and down the control point on the highlight side among the two control points for each RGB tone curve.

  The skin color correction parameters, which are correction parameters for correcting the skin color of the person determined by the skin color of the person who is the subject, the light source color temperature, etc., are R (Red) and G (Green) in the face area of the person specified by the analysis. , B (Blue) is determined according to the average value. The face area of a person is specified by pattern recognition using a template or the like corresponding to the arrangement of eyes, mouth, and nose. In order to set the skin color correction parameter, a human face region is specified for a QVGA-size gray-tone image. Each average value of R (Red), G (Green), and B (Blue) of 256 pixels is obtained for each face area of a person for an RGB tone image of QVGA size. A skin color correction parameter is set according to the difference between each average value and a predetermined fixed value. The skin color correction parameters correspond to the coordinates of the skin color control points of the tone curves of R (Red), G (Green), and B (Blue) and the coordinates of the interpolation control points that suppress gradation skipping. The skin color control point is a control point for designating a preferable skin color pinpoint.

Derivation of grid correction color data corresponding to the grid color FIG. 6 is a diagram showing a tone curve that defines the correspondence between the color of the original image and the color of the correction image for each component. There are two control points in a linear tone curve that defines the correspondence between the color of the original image and the color of the corrected image for each component of R (Red), G (Green), and B (Blue). (See (A), (B), and (C) of FIG. 6). The coordinates of the two control points are determined in advance as, for example, (64/255, 64/255) on the shadow side and (195/255, 195/255) on the highlight side. The coordinate components of the control points are described in the order of the gradation of the original image (input gradation) and the gradation of the corrected image (output gradation). In the present embodiment, the R, G, and B components are described as having 256 gradations (0-255) in both the original image and the corrected image, but the number of gradations is arbitrary. The brightness correction parameter, shadow parameter, highlight parameter, leaf green memory color correction parameter, sky blue memory color correction parameter, red memory color correction parameter, skin color memory color correction parameter, and color balance correction parameter described above are output as control points. The control point moves by accumulatively adding to the gradation component. Further, the coordinates of the skin color control point and the interpolation control point are determined by the skin color correction parameter. The correction parameter is adjusted according to the print setting parameter. For example, if the control value of the correction parameter is set to +10 according to the print setting parameter, and the correction parameter set by image analysis is +20, the final correction parameter is +30. In this way, a tone curve that smoothly connects each control point determined by the correction parameter and (0/255, 0/255) and (255/255, 255/255) is derived by the curve interpolation process. A tone curve (see (A), (B), and (C) of FIG. 6) that defines the correspondence between the input gradation and the output gradation for each component is provided for each input gradation that is one component of the original color data. It is stored in the RAM 14 as three one-dimensional color conversion tables that hold the output gradation that is one component of the correction color data. By three 1-dimensional color conversion table, 256 ³ corrected color data representing a 256 ³ corrected color for 256 ³ colors are defined. Therefore, the total data size of the three one-dimensional color conversion tables is 6144 bits (256 × 8 × 3). The three one-dimensional color conversion tables correspond to function information.

  The discrete input gradation of the one-dimensional color conversion table that defines the tone curve of the R component is the R component of the grid color, and the discrete output gradation corresponding to the discrete input gradation is the R component of the grid correction color data ( Pr). The same applies to the tone curves of the G and B components. Accordingly, grid correction color data for the grid color is derived by the above-described processing for generating tone curves of R, G, and B components. For example, the R component P0r of the grid correction color data corresponding to the grid color indicated by P0 in FIG. 7 is an output gradation extracted from a one-dimensional color conversion table that defines the tone curve of the R component, and the corresponding input This is an output gradation whose gradation is an R component of the grid color.

.. Configuration of three-dimensional color conversion table The linear interpolation operation is executed for each partial area of the polyhedron constituting the color space. A partial region to be subjected to linear interpolation calculation is referred to as a unit. Essentially, four pieces of grid correction color data are necessary for the tetrahedral interpolation calculation, and eight pieces of grid correction color data are necessary for the hexahedral interpolation calculation. Therefore, the unit that is the target of the tetrahedral interpolation calculation is originally a tetrahedron, and the unit that is the target of the hexahedral interpolation calculation is originally a hexahedron. However, in order to simplify the color conversion block 17 corresponding to both the tetrahedral interpolation calculation and the hexahedral interpolation calculation, the color conversion block 17 executes the interpolation calculation in the same unit for both the tetrahedral interpolation calculation and the hexahedral interpolation calculation. Is done. That is, unit data composed of eight grid correction color data is transferred from the RAM 14 to the color conversion block 17 in both tetrahedral interpolation calculation and hexahedral interpolation calculation.

  FIG. 8 is a schematic diagram showing the relationship between the unit of this embodiment and the RGB color space. As the color space is finely divided and the units are made smaller, the accuracy of the linear interpolation calculation is improved, but the number of grid conversion color data increases, so the data amount of the three-dimensional color conversion table increases. The number of units used to divide the color space is arbitrary and may be determined in accordance with the color separation accuracy. However, in this embodiment, the color space is divided into 4096 units. By sharing each vertex with adjacent units, the data amount of the three-dimensional color conversion table can be reduced without substantially reducing the accuracy of the interpolation calculation. However, in order to convert the original color data of one pixel into the corrected color data, it is necessary to read one unit data from the RAM 14 into the color conversion block 17. Therefore, if the grid correction color data is not stored in the continuous address of the RAM 14 for each unit data, the time required for transferring the unit data from the RAM 14 to the color conversion block 17 becomes long. When each vertex of a unit is shared by adjacent units and grid correction color data is stored in the RAM 14 without redundancy, one unit data composed of grid correction color data for all vertices of one unit is stored in a continuous address. I can't do that. In this embodiment, in order to shorten the unit data transfer time and speed up the color correction of the image, the grid correction color data is stored in the continuous address of the RAM 14 for each unit data, and the vertexes are not shared by adjacent units. As a result, in this embodiment, 4096 × 8 grid correction color data corresponding to 4096 × 8 grid colors are held in the three-dimensional color conversion table, and one unit has a rectangular parallelepiped whose length is 15 It becomes the area of.

  Here, the grid colors corresponding to the vertices of any one unit are represented by P0, P1, P2, P3, P4, P5, P6, P7 according to their positions in the color space, as shown in FIG. Define. P0 is the grid color corresponding to the vertex closest to the origin of the unit. P7 is a grid color corresponding to the vertex farthest from the origin of the unit. The unit data is composed of a grid correction color data group corresponding to eight vertices of one unit.

  FIG. 10 is a table showing the data arrangement of the three-dimensional color conversion table. [A: b] is a numerical value represented by a bit string from the a digit (b = 0, 1, 2,...) To the b digit (a = 0, 1, 2,...) Of the address or data. It shall represent. R component of grid correction color data corresponding to the grid color Pn (n = 0, 1, 2,... 7) corresponds to Pnr, G component of grid correction color data corresponding to the grid color Pn corresponds to Png, and grid color Pn. The B component of the grid correction color data to be written is denoted as Pnb. In this embodiment, the RAM 14 is assigned one address per 8 bits. Since the correction image generated by the color conversion block 17 is an RGB tone image in which each gradation of R, G, and B is expressed by 8 bits, each of the R, G, and B components (Pnr, Png, Each of Pnb (n = 0, 1, 2,... 7) is 8 bits. Accordingly, the R, G, and B components of the grid correction color data are stored at one address. Grid correction color data is not stored in one of the four consecutive addresses. That is, one unit data composed of 8 grid correction color data having a total of 192 bits is allocated to a continuous area of 256 bits. It should be noted that how many bits are used for each component of the grid correction color data is a design matter and may be determined according to the number of colors in the digital color space representing the color of the corrected image. Further, the data arrangement of the three-dimensional color conversion table described above is merely an example, and it is only necessary that the grid correction color data is allocated to the continuous area in units of unit data. Further, by compressing the unit data, one unit data can be read into the color conversion block 17 by one memory access.

[Configuration and operation of color conversion block]
.. Overall Configuration FIG. 11 is a block diagram showing the configuration of the color conversion block 17 as a dedicated processor. The color conversion block 17 sequentially selects the target pixel of the original image, reads the original color data representing the color of the target pixel, specifies the address for reading the unit data corresponding to the color of the target pixel, reads the unit data, and linear interpolation This is a dedicated circuit that autonomously repeats a series of processes of calculation and correction color data writing. Since the host interface block 60, the data interface block 62, the unit address generation block 66, the arbiter block 64, and the interpolation operation block 68 constituting the color conversion block 17 are controlled by wired logic, respectively, the speed is higher than that of the microprogram control method. Operate. The color conversion block 17 may be operated by a microprogram control method.

  The host interface block 60 includes the start address of the original image, the image size of the original image, the start address of the three-dimensional color conversion table, the start address of the corrected image, and the mode of the color conversion block 17 (tetrahedral interpolation calculation mode, hexahedral interpolation calculation). Mode), and a register group for controlling start / interruption of color correction. The terminal ain is a terminal for designating the address of the register group. The terminal din is a terminal for inputting data such as the start address of the original image, the start address of the corrected image, and the start address of the three-dimensional color conversion table stored at the specified address. These data are input by the CPU 22 executing the H / W color correction module 54. The terminal dout is a terminal for outputting register group data to make the CPU 22 recognize the state of the color conversion block 17. From the terminal dout, for example, an address signal of the RAM 14 in which correction color data is written for a pixel of interest at a certain time is output. Therefore, the CPU 22 executing the H / W color correction module 54 can recognize the target pixel at an arbitrary time. It is also possible to switch between the tetrahedral interpolation calculation mode and the hexahedral interpolation calculation mode according to the color of the target pixel. For example, the color near the gray axis can be corrected in the tetrahedral interpolation calculation mode, and the color far from the gray axis can be corrected in the hexahedral interpolation calculation mode.

Data input FIG. 12 is a diagram showing data transfer timing between the RAM 14 and the color conversion block 17. The amount of data that can be transferred between the RAM 14 and the color conversion block 17 in one RAM access cycle is 128 bits. The color conversion block 17 pipelines the color conversion of the four target pixels. Therefore, the four original color data Dnr-0, Dnr-1, Dnr-2, and Dnr-3 representing the colors of the four target pixels are transferred from the RAM 14 to the color conversion block 17 in one RAM access cycle. Subsequently, four unit data (P _n 0, P _n 1, P _n 2, P _n 3, P _n 4, P _n 5, P _n 6, P _n 7 (n = 0, 1, 2, 3) ) Is transferred from the RAM 14 to the color conversion block 17 in eight RAM access cycles, and then four correction color data, Dnw-0, Dnw-1, Dnw-2, and Dnw-3 are stored in one RAM access cycle. Then, the data is transferred from the color conversion block 17 to the RAM 14.

The data interface block 62 (see FIG. 11) outputs to the arbiter block 64 one address signal sd_ra_d for reading four original color data representing the colors of the four target pixels of the original image. When the original color data representing the colors of the four target pixels is input from the terminal sd_d_in, the data interface block 62 outputs one interpolation calculation start request rgb_en for the four target pixels to the unit address generation block 66, and the original color data The components R _in [7: 0], G _in [7: 0], and B _in [7: 0] are sequentially output to the unit address generation block 66 for each pixel.

  The unit address generation block 66 converts the original color data of the target pixel into an address of unit data composed of grid correction color data corresponding to eight grid colors close to the color of the target pixel. The arbiter block 64 outputs a request signal req_sr_l output from the unit address generation block 66 for reading unit data and a request signal req_srd output from the data interface block 62 for reading the original color data of the target pixel. Arbitration is performed, and the address signal sd_sr_l or sd_ra_d corresponding to one of the request signals is output from the terminal sd_ra to the RAMC 15. The address signal output from the terminal sd_ra has a bit pattern of [25: 4] of the physical address in order to read data for 16 addresses at a time. In accordance with the address signal output from the terminal sd_ra, data for four addresses is transferred from the RAM 14 to the data interface block 62 or the unit address generation block 66 at a time. When the address signal sd_ra_d for reading the original color data of the target pixel is output from the terminal sd_ra, the original color data of the target pixel is transferred from the RAM 14 to the data interface block 62. When the address signal sd_sr_l for reading the unit data is output from the terminal sd_ra, the four grid correction color data are transferred from the RAM 14 to the unit address generation block 66. The unit address generation block 66 has a cache memory for holding an address signal and unit data for reading unit data corresponding to the original color of the target pixel. When the unit data corresponding to the target pixel is held in the cache memory of the unit address generation block 66, the request signal req_sr_l and the address signal sd_sr_l held in the cache memory are not output to the arbiter block 64. The data interface block 62, the arbiter block 64, and the unit address generation block 66 correspond to data acquisition means.

  FIG. 13 is a diagram showing the correspondence between original color data and unit data addresses. RAM_address [16: 4] represents a part of the bit pattern of the address signal output from the terminal sd_sr_l of the unit address generation block 66. The address signal output from the terminal sd_sr_l is the sum of the start address of the three-dimensional color conversion table set in the register of the host interface block 60 and RAM_address [16: 4]. The CPU 22 always stores the three-dimensional color conversion table in a continuous area of the RAM 14 starting from the address [16: 4] = 0. Therefore, the start address of the three-dimensional color conversion table is always set to [16: 4] = 0 by the CPU 22.

The correspondence between arbitrary original color data and RAM_address [16: 4] is determined as follows for speeding up and circuit simplification. The bit pattern of [16: 5] of the address signal for reading the unit data is a pattern of the upper 4 bits of each of R, G and B of the original color data (R _in [7: 4], G _in [7: 4], B _in [7: 4]) is the same as the bit pattern directly connected in the order of RGB. [4: 4] bits S of the address signal for reading the grid correction color data corresponding to the grid colors P0, P1, P2, and P3 are 0 and correspond to the grid correction color data P4, P5, P6, and P7. Bits [4: 4] of the address signal for reading grid correction color data are 1.

  As described above, the interpolation calculation block 68 executes the interpolation calculation in units of unit data composed of eight grid colors regardless of whether it is the hexahedral interpolation calculation mode or the tetrahedral interpolation calculation mode. The operation of the data interface block 62, the unit address generation block 66, and the arbiter block 64 is one of the following operations (see FIG. 11) regardless of whether it is the hexahedral interpolation calculation mode or the tetrahedral interpolation calculation mode.

  Specifically, an address signal for reading grid correction color data corresponding to four grid colors P0, P1, P2, and P3 close to the color of the target pixel, and four grid colors P4 close to the color of the target pixel. , P5, P6, and P7, an address signal for reading the grid correction color data group is generated by the unit address generation block 66 based on the original color data for each target pixel, and these address signals are generated by the unit address generation block. 66 terminal sd_sr_l. Since the data transfer unit between the RAM 14 and the color conversion block 17 is 128 bits, data for four addresses is transferred at a time by one address designation between the RAM 14 and the color conversion block 17. The address signal output from the terminal sd_sr_l of the unit address generation block 66 is a [25: 4] bit pattern of the physical address.

.. Interpolation calculation When an interpolation calculation start request rgb_en is input from the data interface block 62 to the unit address generation block 66, each component R _in [7: 0], G _{in of the} original color data representing the color of one pixel of interest The lower 4 bits (r [3: 0], g [3: 0], b [3: 0]) of [7: 0] and B _in [7: 0] and unit data are unit address generation block 66. Is output to the interpolation calculation block 68. Specifically, unit data is input to the interpolation calculation block 68 for each grid correction color data, and Pr, Pg, and Pb that are R, G, and B components of one grid correction color data are simultaneously input to the interpolation calculation block 68. Entered.

FIG. 14 is a circuit diagram showing a configuration of an interpolation calculation block 68 as an interpolation calculation means. The interpolation calculation block 68 includes the component R _in [7: 0], G _in [7: 0], and B _in [7: 0] of the unit data input from the unit address generation block 66 and the original color data of the target pixel. Correction color data is sequentially derived by pipeline processing for four pixels of interest by linear interpolation using the lower 4 bits (r [3: 0], g [3: 0], b [3: 0]). . The interpolation calculation block 68 includes a coefficient calculation block 102 corresponding to a coefficient generation circuit and a cumulative addition block 104 corresponding to a cumulative addition circuit. The coefficient calculation block 102 applies the lower 4 bits (r, g, b) of each component R _in [7: 0], G _in [7: 0], B _in [7: 0] of the original color data of the target pixel. Based on this, a coefficient to be multiplied with each of the eight grid correction color data is derived. The cumulative addition block 104 multiplies the coefficient output from the coefficient calculation block 102 by each of the eight grid correction color data, and cumulatively adds the multiplication results. Although the shift calculation circuit is incorporated in each of the coefficient calculation block 102 and the cumulative addition block 104, the shift calculation may be executed at any stage.

... Linear interpolation calculation in hexahedral interpolation calculation mode In hexahedral interpolation calculation, the components R _out , G _out , and B _out of the correction color data are obtained in principle by the following calculation. Incidentally, Pn _r, Pn _g, Pn _b is, R component of the correction color data corresponding to the grid color Pn respectively, G component, and B component.
R _out = (15−r) × (15−g) × (15−b) × P0 _r / 15 ³
+ (15-r) × (15-g) × b × P1 _r / 15 ³
+ (15-r) × g × (15-b) × P2 _r / 15 ³
+ (15-r) × g × b × P3 _r / 15 ³
+ R × (15−g) × (15−b) × P4 _r / 15 ³
+ R × (15−g) × b × P5 _r / 15 ³
+ R × g × (15−b) × P6 _r / 15 ³
+ R × g × b × P7 _r / 15 ³
G _out = (15−r) × (15−g) × (15−b) × P0 _g / 15 ³
+ (15-r) × (15-g) × b × P1 _g / 15 ³
+ (15-r) × g × (15-b) × P2 _g / 15 ³
+ (15-r) × g × b × P3 _g / 15 ³
+ R × (15−g) × (15−b) × P4 _g / 15 ³
+ R × (15−g) × b × P5 _g / 15 ³
+ R * g * (15-b) * P6 _g / 15 < ³ >
+ R × g × b × P7 _g / 15 ³
B _out = (15−r) × (15−g) × (15−b) × P0 _b / 15 ³
+ (15-r) × (15-g) × b × P1 _b / 15 ³
+ (15-r) × g × (15-b) × P2 _b / 15 ³
+ (15-r) × g × b × P3 _b / 15 ³
+ R × (15−g) × (15−b) × P4 _b / 15 ³
+ R × (15−g) × b × P5 _b / 15 ³
+ R × g × (15−b) × P6 _b / 15 ³
+ R × g × b × P7 _b / 15 ³

However, since the division circuit that divides by 15 ³ has a large number of stages, in this embodiment, division by 15 ³ is replaced with a shift operation. Specifically, each component R _out [7: 0], G _out [7: 0], B _{out of the} correction color data is obtained by a combination of the following calculation and dither processing (fixed value addition of 0, 1, 2, 3). _out [7: 0] is obtained. In the following formula, >> n means a shift operation that lowers n bits.
R _out = (64−mR) × (64−mG) × (64−mB) × P0 _r >> 16
+ (64-mR) × (64-mG) × mB × P1 _r >> 16
+ (64-mR) × mG × (64-mB) × P2 _r >> 16
+ (64-mR) × mG × mB × P3 _r >> 16
+ MR × (64−mG) × (64−mB) × P4 _r >> 16
+ MR × (64−mG) × mB × P5 _r >> 16
+ MR × mG × (64−mB) × P6 _r >> 16
+ MR × mG × mB × P7 _r >> 16
G _out = {(64-mR) × (64-mG) × (64-mB) × P0 _g >> 16
+ (64-mR) × (64-mG) × mB × P1 _g >> 16
+ (64-mR) × mG × (64-mB) × P2 _g >> 16
+ (64-mR) × mG × mB × P3 _g >> 16
+ MR × (64-mG) × (64-mB) × P4 _g >> 16
+ MR × (64-mG) × mB × P5 _g >> 16
+ MR × mG × (64−mB) × P6 _g >> 16
+ MR × mG × mB × P7 _g >> 16
B _out = {(64-mR) × (64-mG) × (64-mB) × P0 _b >> 16
+ (64-mR) × (64-mG) × mB × P1 _b >> 16
+ (64-mR) × mG × (64-mB) × P2 _b >> 16
+ (64-mR) × mG × mB × P3 _b >> 16
+ MR × (64-mG) × (64-mB) × P4 _b >> 16
+ MR × (64−mG) × mB × P5 _b >> 16
+ MR × mG × (64−mB) × P6 _b >> 16
+ MR × mG × mB × P7 _b >> 16

The circuit 70 outputs the lower 4 bits (r [3: 0], g [3: 0] of each component R _in [7: 0], G _in [7: 0], B _in [7: 0] of the original color data. ], B [3: 0]) are converted into numbers mR, mG, and mB that are nearly four times the number. The correspondence between r [3: 0] and mR [6: 0] is, for example, as shown in FIG. The circuit 74 inputs 64 and 0 to the selector 78. The circuit 76 is a division circuit that outputs (64-mR), (64-mG), and (64-mB), respectively. The selector 78 selects any three of (64-mR), (64-mG), (64-mB), mR, mG, and mB for each grid correction color data and outputs the selected data to the flip-flop 80. It is. The correspondence between the grid correction color data and the three data selected by the selector 78 corresponds to the above equation. For example, (64-mR), (64-mG), and (64-mB) are selected for grid correction color data corresponding to P0, and (64-mR) for grid correction color data corresponding to P1. ), (64-mG), mB are selected. The multiplication circuit 82 multiplies any three of (64-mR), (64-mG), (64-mB), mR, mG, and mB selected by the selector 78. The circuit 84 is a shift arithmetic circuit that shifts the value output from the multiplication circuit 82 by 5 bits to divide by 32.

The multiplication circuits 90, 103, and 112 receive Pr, Pg, and Pb that are R, G, and B components of the grid correction color data via the flip-flops 86 and 88. The multiplication circuit 90 multiplies Pr, which is the R component of the grid correction color data, and the value output from the coefficient calculation block 102. Eight components sequentially output from the coefficient calculation block 102 and P0r, P1r, P2r, P3r, P4r, P5r, P6r, P7r, which are R components of the eight grid correction color data sequentially output from the unit address generation block 66. Eight multiplication results multiplied by the value are sequentially output from the multiplication circuit 90. The multiplication circuits 103 and 112 execute the same operation as the multiplication circuit 90 for Pg and Pb which are the G and B components of the grid correction color data.
Circuits 92, 105, and 114 are shift operation circuits for dividing the values output from the multiplier circuit 90, the multiplier circuit 103, and the multiplier circuit 112 by 2 ¹¹ , respectively.

By the above calculation, the following eight values are sequentially output from the circuit 92 for the R component of one target pixel. The adder circuit 94 and the flip-flop 96 cumulatively add the following eight values.
(64-mR) × (64-mG) × (64-mB) × P0 _r >> 16
(64-mR) × (64-mG) × mB × P1 _r >> 16
(64-mR) × mG × (64-mB) × P2 _r >> 16
(64-mR) × mG × mB × P3 _r >> 16
mR × (64-mG) × (64-mB) × P4 _r >> 16
mR × (64-mG) × mB × P5 _r >> 16
mR × mG × (64−mB) × P6 _r >> 16
mR × mG × mB × P7 _r >> 16

Further, the following eight values are sequentially output from the circuit 105 for the G component of one target pixel by the above-described calculation. The adder circuit 106 and the flip-flop 108 cumulatively add the following eight values.
(64-mR) × (64-mG) × (64-mB) × P0 _g >> 16
(64-mR) × (64-mG) × mB × P1 _g >> 16
(64-mR) × mG × (64-mB) × P2 _g >> 16
(64-mR) × mG × mB × P3 _g >> 16
mR × (64-mG) × (64-mB) × P4 _g >> 16
mR × (64-mG) × mB × P5 _g >> 16
mR × mG × (64-mB) × P6 _g >> 16
mR × mG × mB × P7 _g >> 16

Further, the following eight values are sequentially output from the circuit 114 for the B component of one target pixel by the above-described calculation. The adder circuit 116 and the flip-flop 118 cumulatively add the following eight values.
(64-mR) × (64-mG) × (64-mB) × P0 _b >> 16
(64-mR) × (64-mG) × mB × P1 _b >> 16
(64-mR) × mG × (64-mB) × P2 _b >> 16
(64-mR) × mG × mB × P3 _b >> 16
mR × (64-mG) × (64-mB) × P4 _b >> 16
mR × (64-mG) × mB × P5 _b >> 16
mR × mG × (64-mB) × P6 _b >> 16
mR × mG × mB × P7 _b >> 16

Circuit 98,110,120 correction color data by 2 bits shift operation for dividing by dithering and 4 each R component R _out [7: 0 ", G component G _out [7: 0", B component B _out [ 7: 0 "is generated. In the dither processing, one of 0, 1, 2, and 3 is added according to the position of the target pixel. FIG. 16 illustrates a dither pattern. As a result, R _out [7: 0], G _out [7: 0], and B _out [7: 0] with 8-bit precision have area gradations with 10-bit precision. The components R _out [7: 0], G _out [7: 0], and B _out [7: 0] of the correction color data are simultaneously output from the cumulative addition block 104 via the flip-flop 100.

... Coefficient calculation in tetrahedral interpolation calculation mode As shown in FIG. 17, one unit is divided into six tetrahedron regions by a diagonal line connecting P0 and P7. This tetrahedral region is called a subunit. In the tetrahedral interpolation calculation, four grid correction color data corresponding to the vertices of the subunits including the color of the target pixel are required. However, in this embodiment, the operation of the color conversion block 17 is the same except for the selector 78 in both the tetrahedral interpolation calculation mode and the hexahedral interpolation calculation mode. Therefore, in the tetrahedral interpolation calculation mode, the coefficient calculation block 102 sets the coefficient to be multiplied to four of the eight grid correction color data constituting the unit data to zero. That is, the coefficients corresponding to the four grid correction color data that do not correspond to the vertices of the subunits including the color of the target pixel are set to 0. Specifically, in the tetrahedral interpolation calculation mode, a determination for specifying a subunit including the color of the pixel of interest is performed by the circuit 72 and the selector 78 as a comparison circuit, and the selector 78 determines according to the calculation result. 64, 0, (64-mR), (64-mG), (64-mB), mR, mG, mB, (mR-mG), (mR-mB), (mG-mR) as coefficient elements Any three of (mG-mB), (mB-mR), and (mB-mG) are selected and output to the flip-flop 80. The circuit 72 outputs the lower 4 bits (r [3: 0], g [3: 0] of each component R _in [7: 0], G _in [7: 0], B _in [7: 0] of the original color data. ], B [3: 0]) is a subtraction circuit for clarifying the magnitude relationship.

The subunits specified by the circuit 72 and the selector 78 are as shown in Table 1 below.

The correspondence relationships between the three coefficient elements selected by the selector 78 in the tetrahedral interpolation mode and the grid conversion color data are as shown in Tables 2 to 7 below. Note that when the color of the target pixel is located on a diagonal line of a unit shared by the six subunits, mR-mG, mG-mB, and mB-mR are all 0. In this case, the same calculation result can be obtained regardless of the interpolation calculation for any subunit. In the present embodiment, when mR-mG, mG-mB, and mB-mR are all positive or 0, interpolation calculation is performed for subunit A, and mR-mG, mG-mB, and mB-mR are all If negative, an interpolation operation is performed for subunit F.

.. Data output The correction color data output from the cumulative addition block 104 by the above calculation is output to the data interface block 62 (see FIG. 11). When the interpolation calculation is completed for the four target pixels, an interpolation calculation end notification out_en is output from the interpolation calculation block 68 to the data interface block 62, and four pixels worth of data are output from the data interface block 62 according to the input of the interpolation calculation end notification out_en. The corrected color data is output from the terminal sd_d_out. At this time, the address signal req_sw corresponding to the pixel whose color is represented by the correction color data for four pixels (four pixels of the correction image corresponding to the four target pixels) is output, and the correction color data for four pixels is stored in the RAM 14. A write request signal req_sw for writing to is output. When the interpolation calculation end notification out_en is input, the data interface block 62 repeats the data input described above for the next four pixels.

[Flow of image correction]
FIG. 1 is a flowchart showing the flow of color correction for automatic image quality correction of a photographic image by the printer 1. The process shown in FIG. 1 is executed by the CPU 22 that executes the control program described above except for step S114.
In step S100, the print setting parameter is analyzed, and a control value for changing the correction parameter is set. Specifically, the print setting parameter passed from the print control module 40 (see FIG. 4) is analyzed by the color correction control module 42, and the control value of the correction parameter is set according to the print setting parameter. The print setting parameters that affect image correction are, for example, scenes, brightness operation values, and contrast operation values that are set for a photographic image to be printed. The scene settings include standard, person, landscape, night view, backlight, and the like. A control value for changing the vertical width of the control point represented by the correction parameter is set according to the scene setting, the brightness operation value, the contrast operation value, and the like.

  In step S102, the original image to be printed is analyzed. That is, the correction parameter generation module 48 samples an RGB tone original image, and derives a statistical value such as a QVGA size RGB tone image generated by the sampling. When the print target is the JPEG format, an RGB tone original image is generated by decoding before analysis. When the print target is RAW data, an RGB tone original image is generated by demosaic processing before analysis. The details are as described above for generating the three-dimensional color conversion table.

  In step S104, correction parameters are set. In other words, the correction parameter generation module 48 sets a correction parameter according to the analysis result of the original image and the control value of the correction parameter. The details are as described above for generating the three-dimensional color conversion table.

  When the vertical width of the control point of the tone curve determined by the correction parameter increases, the radius of curvature of the tone curve decreases on average, so each component of the correction color data determined by the one-dimensional color conversion table that defines the tone curve and linear interpolation A difference (see ΔB shown in FIG. 18) from each component of the correction color data obtained by the calculation becomes large. When this difference increases, the smoothness of the gradation of the corrected image is impaired.

  Therefore, in step S106, the correction method determination module 44 determines whether or not the verification result of the three-dimensional color conversion table satisfies a parameter criterion that is a first criterion for determining a color correction possible range by hardware. Specifically, for example, the determination is made as follows based on a table shown in FIG. 19 that holds the lower limit value and the upper limit value of each correction parameter as parameter references. This table is included in the control program. If parameter 1 (for example, a highlight parameter), which is one of the correction parameters, is −50 or less that is a reference lower limit value or +50 or more that is a reference upper limit value, it is determined that the parameter criterion is not satisfied. In this case, the continuous mode is set, and the process proceeds to step S116. If there is even one correction parameter that is equal to or less than the reference lower limit value or greater than the reference upper limit value, it is determined that the parameter criterion is not satisfied. The parameter criterion is determined as follows, for example. First, arbitrary correction parameters are set for a large number of photographic images, and a three-dimensional color conversion table is generated according to the set correction parameters. Next, a large number of corrected images are generated by linear interpolation using the generated three-dimensional color conversion table. By evaluating a large number of correction images generated in this way, it is possible to specify a correction parameter for obtaining a correction image with sufficient image quality for any correction image.

  If the parameter criterion is satisfied, the table generation module 50 generates a three-dimensional color conversion table (step S108). Since the 3D color conversion table is not generated unless the parameter criterion is satisfied, the 3D color conversion table that cannot generate a corrected image with sufficient image quality is not generated wastefully. First, a one-dimensional color conversion table corresponding to the tone curves of the R, G, and B components is generated based on the correction parameters. In addition, the start address of the three-dimensional color conversion table is set. Output gradation corresponding to each component of an arbitrary grid color is acquired as each component of grid correction color data from the three one-dimensional color conversion tables. Each acquired grid correction color data is stored at an address corresponding to the corresponding grid color. These details are as described above. Actually, if color correction by the S / W color correction module 52 is performed using an image in which each pixel has a grid color as an original image, an image in which each pixel is composed of grid correction color data is generated. The pixel value of each pixel of the generated image may be sequentially stored at a predetermined address as grid correction color data.

In step S110, the three-dimensional color conversion table is verified by the correction method determination module 44. The accuracy of the linear interpolation calculation is statistically worst when the color represented by the original color data is located at the center of gravity of the unit. Therefore, the correction color data (second correction color data) obtained as a result of the linear interpolation calculation by executing the linear interpolation calculation by the color conversion block 17 on the original color data representing the color (determination color) corresponding to the center of gravity of the unit. And the accuracy of the linear interpolation operation is determined by comparing the correction color data (first correction color data) obtained by software processing using the one-dimensional color conversion table corresponding to the R, G, and B tone curves. Can do. Here, the correction color data obtained by the linear interpolation calculation by the color conversion block 17 is (R _h , G _h , B _h ), and software processing using a one-dimensional color conversion table corresponding to the R, G, and B tone curves. The correction color data obtained by (color correction processing by the S / W color correction module 52) is assumed to be (R _s , G _s , B _s ) (see FIG. 20). In step S110, the two correction color data (R _h , G _h , B _h ) corresponding to the center of gravity of the unit and the square of the distance Δd between the two points in the color space indicated by (R _s , G _s , B _s ) Is required.
Δd ² = (Rh−Rs) ² + (Gh−Gs) ² + (Rh−Rs) ²

The verification of the three-dimensional color conversion table may be executed for all units, or may be executed for some units. For example, since the color twist and gradation jump are particularly noticeable on the gray axis where R _in = G _in = B _in , only the unit through which the gray axis penetrates, or only the unit through which the gray axis penetrates and the adjacent unit are targeted. Verification of the three-dimensional color conversion table may be performed. By executing the verification of the three-dimensional color conversion table only for some units, the time required for the verification can be shortened.

Specifically, for example, the three-dimensional color conversion table is verified as follows. First, the S / W color correction module 52 performs color correction using a one-dimensional color conversion table on the sample image in which the color of each pixel is represented by the original color data corresponding to the center of gravity of each unit to be verified. As a result, an S / W corrected image is generated. Further, a linear interpolation operation by the color conversion block 17 is executed on the same sample image, and as a result, an H / W corrected image is generated. Then, Δd ² is derived for all the pixels of the S / W corrected image and the H / W corrected image. Finally, statistical values such as an average value, a maximum value, and a median value are derived using Δd ² derived for all pixels as a population.

In step S112, the correction method determination module 44 determines whether or not the verification result of the three-dimensional color conversion table satisfies a table criterion that is a second criterion for determining a color correction possible range by hardware. Specifically, for example, statistical values such as an average value, a maximum value, and a median value derived as a population from Δd ² obtained from the S / W correction image and the H / W correction image are determined in advance as a table reference. If it is less than the fixed value, it is determined that the table criterion is satisfied. By determining the possible range of color correction by hardware through such quantitative analysis, the target image quality of the corrected image can be accurately managed regardless of the image quality of the original image.

If both the parameter criterion and the table criterion are satisfied, color correction of the image by hardware is executed in step S114. In other words, the linear interpolation operation as described above is executed by the color conversion block 17, and the conversion data derived by the color conversion block 17 is sequentially stored in the RAM 14 from the start address of the correction image set by the H / W color correction module 54. Stored. As a result, a corrected image in which the color of each pixel is represented by the conversion data is generated.
If either the parameter criterion or the table criterion is not satisfied, image correction by software is executed in step S116. That is, the CPU 22 that executes the S / W color correction module 52 performs the following processing.

FIG. 21 is a flowchart showing a flow of image correction processing by software.
In step S200, one of the pixels of the original image is selected as the target pixel.
In step S202, original color data representing the color of the target pixel is read from the RAM 14 into the CPU 22.
In step S204, the R component of the original color data is converted into an address of a one-dimensional conversion table corresponding to the tone curve of the R component, and the address of the one-dimensional conversion table corresponding to the G component of the original color data corresponds to the tone curve of the G component. The B component of the original color data is converted into an address of a one-dimensional conversion table corresponding to the tone curve of the B component. The R, G, and B components of the correction color data are sequentially read into the CPU 22 from the three generated addresses.
In step S206, the correction color data is stored in the pixel corresponding to the target pixel. That is, the correction color data is stored at the address of the RAM 14 corresponding to the target pixel.
In step S208, it is determined whether the target pixel is the last pixel of the original image. The processing from step S200 to step S206 is repeated up to the last pixel of the original image.

In the color correction by software in the continuous mode described above, reading of the original color data, generation of the address of the R component one-dimensional color conversion table, reading of the R component of the correction color data, generation of the address of the G component one-dimensional color conversion table The CPU 22 repeats a series of processes of reading the G component of the correction color data, generating the address of the one-dimensional color conversion table of the B component, reading the B component of the correction color data, and writing the correction color data for each pixel. The time required for color correction is longer than that for hardware color correction. On the other hand, in the color correction by software, 256 for nonlinear color conversion by a one-dimensional color conversion table holding the correction color data corresponding to all the colors represented by the ^three colors RGB color space is executed, three-dimensional color conversion More precise color conversion is possible than color conversion by linear interpolation using a table. In the present embodiment, since the color correction by software and the color correction by hardware are switched according to the correction amount determined by the correction parameter, the image quality of the corrected image is not substantially deteriorated by the color correction by hardware. As long as color correction is performed by hardware, color correction is performed at high speed. In most cases, a suitable image quality can be obtained by color correction with a small correction amount for a photographic image generated by a digital camera under an automatically set imaging condition. Therefore, the image quality of photographic images can be improved by color correction by hardware at a considerably high rate in practice.

[Other Embodiments]
In the embodiment described above, the three-dimensional color conversion table is generated based on the analysis result of the photographic image. However, the three-dimensional color conversion table may be generated without analyzing the photographic image. For example, when a control value for which a correction parameter is uniquely determined is attached to an image to be printed, an appropriate three-dimensional color conversion table for the photographic image can be obtained without analyzing the photographic image. Can be generated. In this embodiment, in the color correction by software processing, the correction color data is derived by tone curve conversion using a one-dimensional color conversion table. However, the correction color data is derived by function calculation with four arithmetic operations for each pixel. May be. Information (program code) defining such a function operation and a one-dimensional color conversion table correspond to the function information. Further, the present invention can be applied to various image processing apparatuses other than a printer. Examples of the image processing apparatus to which the present invention can be applied include a projector, a digital camera, an image scanner, an MFP (Multi Function Printer), a laser copying machine, and a photographic image dedicated printing system.

The flowchart which concerns on one Embodiment of this invention. The block diagram which concerns on one Embodiment of this invention. The block diagram which concerns on one Embodiment of this invention. The block diagram which concerns on one Embodiment of this invention. The schematic diagram which concerns on one Embodiment of this invention. The figure which shows the tone curve which concerns on one Embodiment of this invention. The schematic diagram which concerns on one Embodiment of this invention. The schematic diagram which concerns on one Embodiment of this invention. The schematic diagram which concerns on one Embodiment of this invention. The memory map which concerns on one Embodiment of this invention. The block diagram which concerns on one Embodiment of this invention. The timing diagram which concerns on one Embodiment of this invention. The table explaining the address signal concerning one embodiment of the present invention. The block diagram which concerns on one Embodiment of this invention. The schematic diagram which concerns on one Embodiment of this invention. The schematic diagram which concerns on one Embodiment of this invention. The schematic diagram which concerns on one Embodiment of this invention. The figure which shows the tone curve which concerns on one Embodiment of this invention. The table which shows the parameter standard concerning one embodiment of the present invention. The schematic diagram which concerns on one Embodiment of this invention. The flowchart which concerns on one Embodiment of this invention.

Explanation of symbols

1: printer (image processing apparatus), 17: color conversion block (dedicated processor), 22: CPU 22 (general-purpose processor), 42: color correction control module (mode setting means), 44: correction method determination module (parameter determination means, Table setting means), 48: correction parameter generation module (analysis means), 50: table generation module (table generation means), 52: S / W color correction module (continuous correction means), 62: data interface block (data acquisition means) ), 64: Arbiter block (data acquisition means), 66: Unit address generation block (data acquisition means), 68: Interpolation calculation block (interpolation calculation means), 72: Circuit (comparison circuit), 78: Selector (comparison circuit) , 102: Coefficient calculation block (coefficient generation circuit), 104: Cumulative addition Lock (cumulative addition circuit)

Claims (9)

An image processing apparatus for correcting the color of an image represented in a digital color space having N colors,
Memory,
A three-dimensional color conversion table for deriving n grid correction color data representing correction colors corresponding to n (n <N) grid colors and holding the grid correction color data for each grid color is stored in the memory. A general-purpose processor that generates
A unit configured to acquire original color data representing the color of a target pixel of an original image stored in the memory and to include a plurality of grid correction color data corresponding to the plurality of grid colors close to the color of the target pixel Data is acquired from the three-dimensional color conversion table based on the original color data, and corrected color data corresponding to the original color data is derived by linear interpolation using the unit data acquired from the three-dimensional color conversion table. A dedicated processor that stores the correction color data in the memory for the pixel of interest, and generates a corrected image in which the color of each pixel is represented by the correction color data by repeating a series of processes;
An image processing apparatus comprising: The general-purpose processor analyzes the original image and derives each grid correction color data corresponding to each grid color based on an analysis result of the original image.
The image processing apparatus according to claim 1. The general-purpose processor analyzes the original image in a plurality of color spaces;
The image processing apparatus according to claim 2. The dedicated processor derives the correction color data by the hexahedral interpolation calculation as the linear interpolation calculation in the hexahedral interpolation calculation mode, and the correction color by the tetrahedral interpolation calculation as the linear interpolation calculation in the tetrahedral interpolation calculation mode. Deriving data The general-purpose processor controls switching between the hexahedral interpolation calculation mode and the tetrahedral interpolation calculation mode.
The image processing apparatus according to claim 1. The general-purpose processor stores the three-dimensional color conversion table by storing the unit data composed of the grid correction color data corresponding to the eight grid colors adjacent to each other at different consecutive addresses in the memory. Generate
The dedicated processor includes data acquisition means for simultaneously acquiring a plurality of grid correction color data corresponding to one unit data from a plurality of consecutive addresses in the memory, and in the hexahedral interpolation calculation mode, Interpolation calculation means for deriving the correction color data by hexahedral interpolation calculation,
The image processing apparatus according to claim 1. The interpolation calculation means sequentially acquires the eight grid correction color data constituting the unit data from the data acquisition means in the hexahedral interpolation calculation mode and the tetrahedral interpolation calculation mode, and the data acquisition means A coefficient corresponding to the grid correction color data acquired from the above is generated based on the original color data, and the eight grid correction color data constituting the unit data are weighted by the corresponding coefficient, The coefficient corresponding to the four grid correction color data determined by the original color data among the obtained unit data in the tetrahedral interpolation calculation mode is derived by accumulatively adding each component. Produces 0 as
The general-purpose processor controls switching between the hexahedral interpolation calculation mode and the tetrahedral interpolation calculation mode;
The image processing apparatus according to claim 5. The interpolation calculation unit weights the grid correction color data acquired from the data acquisition unit with the corresponding coefficient, and cumulatively adds the components for each component to derive the correction color data; and Based on the comparison circuit that compares the magnitude relationship of the lower bit pattern of each component of the original color data that represents the color of the target pixel, and the lower bit pattern of each component of the original color data that represents the color of the target pixel, A coefficient generation circuit that generates the coefficient corresponding to the grid correction color data acquired by the cumulative addition circuit;
In the hexahedral interpolation calculation mode, the coefficient generation circuit performs all the acquisition by the cumulative addition circuit regardless of the magnitude relation of the lower bit pattern of each color component of the original color data representing the color of the target pixel. The coefficient corresponding to the grid correction color data is generated, and in the tetrahedral interpolation calculation mode, the original color data representing the color of the target pixel among the one unit data acquired by the cumulative addition circuit. Generating 0 as the coefficient corresponding to the four pieces of grid correction color data determined according to the magnitude relationship of the lower bit pattern of each color component;
The image processing apparatus according to claim 6. The dedicated processor operates in wired logic,
The image processing apparatus according to claim 1. The dedicated processor derives an approximate value of a division result by a combination of a shift operation circuit and a multiplication circuit.
The image processing apparatus according to claim 1.

Images