JP2006261756A - Color conversion using color conversion lookup table with ununiform input grating point appropriate to photographing scene - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a technology capable of carrying out color conversion by using a color conversion lookup table appropriate to an individual image being a processing object. <P>SOLUTION: A color conversion apparatus prepares a plurality of color conversion lookup tables in cross-reference with a plurality of photographing scenes, discriminates a photographing scene on the basis of an image data file including image data, and selects a proper color conversion lookup table in response to the discriminated photographing scene. Then the apparatus converts the image data into ink amount data by using the selected color conversion lookup table. A plurality of the color conversion lookup tables cross-referenced with the photographing scenes respectively include ununiform input grating points at ununiform intervals associated with pixel values of the image data. Further, the color conversion lookup tables cross-referenced with the different photographing scenes have ununiform input grating points different from one another. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a technique for converting image data into ink amount data.

  When converting image data (for example, RGB values or Lab values) into ink amount data (for example, CMYK values), a color conversion lookup table is generally used. A normal color conversion lookup table stores only ink amount data corresponding to a small number of RGB values selected in advance among all the RGB values that can be taken by the input pixel value. This is because storing the ink amount data for all the RGB values results in a huge table capacity. Note that the points in the RGB color space defined by the RGB input values of the color conversion lookup table are also called “input grid points” of the color conversion lookup table.

  As a method for performing color conversion using a color conversion lookup table, for example, those disclosed in Patent Documents 1 and 2 are known.

JP 2001-285663 A Japanese Patent Laid-Open No. 10-136216

  Patent Document 2 discloses a technique for accurately reproducing the skin color by making the unit cube in the input grid of the color region corresponding to the skin color smaller than the unit cube in the input grid of the other color region.

  However, the range of colors that are desired to be reproduced with high accuracy may vary from image to image. Conventionally, the fact that color conversion is performed using a color conversion look-up table suitable for individual images to be specified has not been sufficiently devised.

  It is an object of the present invention to provide a technique capable of performing color conversion using a color conversion lookup table suitable for individual images to be processed.

A color conversion device according to the present invention is a color conversion device that converts image data into ink amount data,
A table storage unit for storing a plurality of color conversion lookup tables associated with a plurality of shooting scenes;
A scene determination unit that determines a shooting scene based on an image data file including image data;
A table selection unit that selects an appropriate color conversion lookup table from the plurality of color conversion lookup tables according to the determined shooting scene;
A color conversion execution unit that converts the image data into ink amount data using the selected color conversion lookup table;
With
The plurality of color conversion lookup tables each have unequal input grid points having unequal intervals with respect to pixel values of the image data, and color conversion lookup tables associated with different shooting scenes. Have different unequal input grid points.

  According to this apparatus, since a table suitable for an image capturing scene is selected from a plurality of color conversion lookup tables, color conversion can be performed using a color conversion lookup table suitable for each image. And preferable color reproduction can be realized. In particular, since the color conversion lookup table has unequal input grid points, accurate color conversion suitable for individual images can be realized without excessively increasing the data amount of each table. it can.

The table storage unit is further a plurality of input value conversion tables associated with the plurality of shooting scenes, and the pixel values of the image data are any of unequal input grid points of a color conversion lookup table. A plurality of input value conversion tables for converting into grid point input values corresponding to
The table selection unit further selects an appropriate input value conversion table from the plurality of input value conversion tables according to the determined shooting scene,
The color conversion execution unit
Using the selected input value conversion table, the pixel value of the image data is probabilistically converted to a grid point input value corresponding to one of the non-uniform grid points of the selected color conversion lookup table. 1 conversion unit;
A second converter that converts the grid point input value into ink amount data using the selected color conversion lookup table;
You may make it have.

  In this configuration, since the input value conversion table used for the first conversion and the color conversion lookup table used for the second conversion are selected according to the shooting scene, color conversion suitable for each image can be performed.

Each input value conversion table is a table that outputs a grid point input value corresponding to one of the input grid points of the color conversion lookup table and an additional grid point determination value according to the input pixel value. Yes,
The first conversion unit includes:
By referring to the selected input value conversion table using the pixel value of the image data, a grid point input value corresponding to any of the non-uniform grid points of the selected color conversion lookup table, and A table reference unit for acquiring additional lattice point determination values;
A random noise adding unit that adds random noise to the lattice point determination value;
A lattice point determination unit that determines a lattice point input value to be input to the color conversion lookup table by adjusting the lattice point input value according to a lattice point determination value to which the random noise is added;
You may make it provide.

  In this configuration, the first conversion unit can convert the pixel value of the image data to a lattice point input value corresponding to one of the unequal lattice points of the color conversion lookup table.

The scene determination unit
A hue determination unit that analyzes the image data and determines a characteristic hue that characterizes the image data;
A scene determination unit that determines the shooting scene using the determined characteristic hue;
You may make it provide.

  According to this configuration, the shooting scene can be determined by analyzing the image data.

The scene determination unit further includes:
An area specifying unit for specifying an area corresponding to the determined characteristic hue in the image represented by the image data;
A texture extraction unit for extracting the texture of the specified region;
With
The scene determination unit may determine the shooting scene using the extracted texture in addition to the determined characteristic hue.

  According to this configuration, since both the characteristic hue and the texture are used, more accurate scene determination can be executed.

The image data file has additional information including a shooting scene setting value set when the image represented by the image data is shot.
The scene determination unit may determine the shooting scene from the shooting scene setting value in the additional information.

  Since the setting value at the time of shooting is high in reliability, a scene determination result with higher reliability can be obtained by using the shooting scene setting value of the additional information name.

  The present invention can be realized in various forms, for example, a color conversion method and apparatus, an image processing method and image processing apparatus including a color conversion process, a printing method and a printing apparatus, a printer driver, and a printing method. Alternatively, the present invention can be realized in the form of a computer program for realizing the functions of the printing apparatus, a recording medium on which the computer program is recorded, or the like.

Next, embodiments of the present invention will be described in the following order based on examples.
A. First embodiment:
B. Second embodiment:
C. Third embodiment:
D. How to determine the shooting scene:
E. Variations:

A. First embodiment:
FIG. 1 is an explanatory diagram showing a configuration of an image processing system 100 as an embodiment of the present invention. The system 100 includes a computer 200 as an image processing apparatus and a printer 300 as an image output apparatus.

  The computer 200 is installed with a printer driver 210 for generating print data based on the image data file IMG. The printer driver 210 includes a resolution conversion unit 220, a color conversion unit 230, a halftone processing unit 240, a data arrangement unit 250, and a scene determination unit 260. The resolution conversion unit 220 has a function of converting the resolution of the input color image data IMG into a resolution suitable for processing after the color conversion unit 230. The color conversion unit 230 converts the color image data IMG (usually RGB data) into ink amount data of a plurality of inks. The ink amount data is data representing the ink amount of each ink per unit area.

  The scene determination unit 260 determines an image shooting scene based on the image data file IMG. The color conversion unit 230 selects a color conversion lookup table suitable for the determined shooting scene, and performs color conversion using the selected table. The method for determining the shooting scene will be described in detail in section D.

  The halftone processing unit 240 generates dot data indicating the ink dot formation state for each print pixel by executing halftone processing based on the ink amount data of each ink. The data arrangement unit 250 arranges these dot data and outputs them as print data PD. The print data PD generated by the printer driver 210 is supplied to the printer 300. The print data PD includes dot data representing the ink dot recording state for each pixel on the main scan line having the print resolution, and sub-scan feed amount data specifying the sub-scan feed amount.

  The printer driver 210 is usually implemented as a program stored in a storage unit such as a hard disk in the computer. In this case, the print data PD created by the printer driver is supplied to an external printer. Alternatively, a printer driver may be installed in the printer. In this case, the print data PD created by the printer driver is supplied to a print execution unit (printing mechanism) inside the printer. Even in the case of a printer driver installed in a computer, an external printer can be called a “print execution unit”. Accordingly, the printer driver generally has a function of generating print data to be supplied to the print execution unit based on the color image data. Note that the resolution conversion unit 220 and the data arrangement unit 250 may be omitted from the printer driver. Also, part or all of the printer driver can be realized by hardware.

  FIG. 2 is a block diagram illustrating an internal configuration of the color conversion unit 230 in the first embodiment. The color conversion unit 230 includes a lookup table (LUT) reference unit 410, a lookup table selection unit 420, a plurality of color conversion lookup tables 430, and an interpolation execution unit 440.

  The LUT selection unit 420 selects an LUT suitable for an image shooting scene from a plurality of color conversion LUTs. The LUT reference unit 410 converts input image data IMG (for example, RGB data) into ink amount data INK with reference to the selected color conversion LUT. Further, the interpolation execution unit 440 uses a known interpolation method (for example, tetrahedral interpolation method), interpolates the ink amount data INK according to the pixel value (referred to as “input pixel value”) of the image data IMG, and inputs it. Final ink amount data FINK suitable for the pixel value is determined.

  The color conversion LUT 430 is a table that converts a set of three input pixel values (for example, RGB values) into ink amount data INK for a plurality of inks (for example, CMYK). In this embodiment, a plurality of color conversion LUTs associated with a plurality of shooting scenes are prepared in advance.

FIG. 3 is an explanatory diagram showing input grid points of a plurality of color conversion LUTs. FIG. 3A shows an example of a color conversion LUT having a uniform grid. The input grid points of the uniform grid color conversion LUT are defined in the RGB color space, and all grid points are arranged equally. That is, in the input grid of this color conversion LUT, a plurality of discrete input values selected from all possible values 0 to 255 of the pixel value are set on each of the R axis, the G axis, and the B axis. The difference between adjacent input values is constant. For example, when nine discrete input values are set for each axis, the number of input grid points is 9 ³ = 729. Further, when 17 discrete input values are set for each axis, the number of input grid points is 17 ³ = 4913. The input grid point of R = G = B = 0 is a black point BK representing black, and the input grid point of R = G = B = 255 (maximum value) is a white point W representing white.

  FIG. 3B shows an example of a color conversion LUT 430a for night scenes. In the night view color conversion LUT 430a, each axis is divided into a dense lattice FL and a coarse lattice CL. The dense lattice FL is set in the range where the value of each axis is the smallest (most shadow side), and the coarse lattice CL is set on the highlight side. The dense lattice FL is a difference between input values (that is, a lattice having a relatively small lattice pitch, and the coarse lattice CL is a lattice having a relatively large difference between input values. For example, a difference between input values in the dense lattice FL. Can be set to 8, and the difference between the input values in the coarse lattice CL can be set to 16 which is twice that of the dense lattice FL.

  In the example of FIG. 3B, the same division is applied to each of the R, G, and B axes as the division of the dense lattice FL and the coarse lattice CL, but different divisions may be applied to the three axes. Good. Further, the lattices having different lattice pitches are not limited to the two types of dense lattice FL and coarse lattice CL, and three or more types of lattice pitches may be used. In this specification, an input grid point composed of such an unequal grid is referred to as an “unequal input grid point” or an “non-uniform input grid point”.

  FIG. 3C shows an example of a color conversion LUT 430b for portrait (for portraits). In the portrait color conversion LUT 430b, each axis is divided into a dense lattice FL and a coarse lattice CL. The dense grid FL is set slightly on the highlight side than the night scene color conversion LUT 430a, and the coarse grid CL is set on both sides of the dense grid FL. FIG. 3D shows an example of the color conversion LUT 430c for the sky blue scene. In the sky color scene color conversion LUT 430b, each axis is also divided into a dense lattice FL and a coarse lattice CL, and the dense lattice FL is set further on the highlight side than the portrait color conversion LUT 430b.

  The night scene color conversion LUT 430a can reproduce the dark color of the night scene with high accuracy. The portrait color conversion LUT 430b can reproduce the skin color with high accuracy. Also, the sky color scene color conversion LUT 430b can reproduce bright sky colors with high accuracy. Note that a color conversion LUT suitable for scenes other than these three shooting scenes may be used. As described above, if a plurality of color conversion LUTs 430a to 430c suitable for a photography scene are prepared in advance and an appropriate color conversion LUT is selected in accordance with the image photography scene, the color of each image can be selected. Can be accurately reproduced.

  In addition, instead of using a non-uniform color conversion LUT having a dense lattice FL and a coarse lattice CL, if all use a uniform color conversion LUT having a fine lattice pitch equivalent to the dense lattice FL, the color reproduction accuracy is improved. To do. However, such a dense grid uniform color conversion LUT has a problem that the number of input grid points is enormous and the required memory capacity is enormous. On the other hand, if an uneven color conversion LUT suitable for a shooting scene is used, it is possible to improve the color reproduction accuracy without excessively increasing the memory capacity.

  The plurality of color conversion LUTs 430a to 430c associated with the plurality of shooting scenes can be installed in the computer in various modes. For example, if a plurality of color conversion LUTs are stored in advance in the installer file of the printer driver 210, they can be installed on the computer by copying the files. Alternatively, only the uniform grid color conversion LUT (FIG. 3A) is stored in the installer file of the printer driver 210, and a plurality of grid points of the uniform grid color conversion LUT are interpolated when the printer driver 210 is installed. The color conversion LUTs 430a to 430c may be created. In the latter method, the capacity of the installation file for the printer driver 210 can be reduced.

  As described above, in the first embodiment, an appropriate color conversion LUT is selected from a plurality of color conversion LUTs according to the shooting scene of the image determined from the image data file IMG, and the selected color conversion LUT is selected. Since color conversion is executed using the color, it is possible to accurately reproduce colors suitable for individual images. In particular, since a table having unequal input grid points is used as the color conversion LUT suitable for the shooting scene, accurate color conversion can be realized without excessively increasing the data amount of the table. Further, since printing is executed using the ink amount data obtained by this color conversion, a printed matter having a preferable color reproduction suitable for each image can be created.

B. Second embodiment:
FIG. 4 is a block diagram showing an internal configuration of the color conversion unit 230a in the second embodiment. The color conversion unit 230a includes a pre-conversion unit 500 and a post-conversion unit 600. The post conversion unit 600 has a configuration in which the interpolation execution unit 440 is omitted from the color conversion unit 230 of the first embodiment shown in FIG.

  The pre-conversion unit 500 includes a table reference unit 510, a table selection unit 520, a plurality of pre-conversion tables 530, a random noise addition unit 540, and a grid value determination unit 550. As will be described below, the pre-conversion unit 500 has a function of probabilistically converting the pixel value of the input image data IMG into one of a plurality of grid point input values of the color conversion LUT.

  The table selection unit 520 of the pre-conversion unit 500 selects a table suitable for the shooting scene from the plurality of pre-conversion tables 530. The table reference unit 510 converts the pixel value of the input image data IMG (for example, RGB data) into a grid point input value LI suitable for the color conversion look step table with reference to the selected pre-conversion table. However, the output of the pre-conversion table 530 includes the lattice point input value LI (also referred to as “temporary lattice point input value”) and the additional decimal point AF. The random noise adding unit 540 adds a random noise RN to the additional decimal point AF. The grid value determination unit 550 determines the final grid point input value LI depending on whether or not the added value (AF + RN) of the additional decimal number AF and the random noise RN is 1.0 or more. That is, when the addition value (AF + RN) is 1.0 or more, a value obtained by adding 1 to the temporary grid point input value LI is adopted as the final grid point input value FLI, while the addition value (AF + RN) is 1. When it is less than 0, the temporary grid point input value LI is directly adopted as the final grid point input value FLI.

  Note that the plurality of pre-conversion tables 530 and the plurality of color conversion LUTs 430 can be considered to be stored in a common table storage unit.

  FIG. 5 is an explanatory diagram showing an outline of pre-conversion and post-conversion. Grid points Pa to Ph indicated by white circles in FIG. 5A are input grid points of the color conversion LUT, and black circles are points Px represented by pixel values (Rx, Gx, Bx) of the image data IMG. In the pre-transformation, the input pixel value (Rx, Gx, Bx) is probabilistically assigned to any of the eight nearby input grid points Pa to Ph. The lower part of FIG. 5 (A) shows pre-transformation relating to the R component. First, the input pixel value Rx is converted into the grid point input value LI and the additional decimal AF by the pre-conversion table. The value of the grid point input value LI is the grid point input value R1 closest to the input pixel value Rx among the grid point input values equal to or less than the input pixel value Rx. The additional decimal point AF is set to a value obtained by dividing the difference (Rx−R1) between the input pixel value Rx and the grid point input value LI (= R1) by the difference (R2−R1) between the two grid point input values R1 and R2. Is done. For example, when R1 = 8, R2 = 16, and Rx = 10, AF = (10−8) / (16−8) = 0.25. A random noise RN in the range of 0 to 1 is added to the additional decimal point AF. As described above, the final grid point input value FLI for the input pixel value Rx is set to one of the two grid point input values R1 and R2 according to the addition result (AF + RN).

  As described above, the final grid point input value FLI obtained by the pre-transformation is set to one of the two values R1 and R2 according to the value of the input pixel value Rx. It depends on the random noise RN. Therefore, even in a uniform image region in which pixels having the same input pixel value Rx are gathered, the result of the pre-transformation is a result of randomly distributing the two values R1 and R2. Further, the average value of the grid point input value FLI of the pixels in the uniform image region is equal to the input pixel value Rx.

  As shown in FIG. 5B, in the post-conversion, the final grid point input value FLI determined by the pre-conversion is used for referring to the color conversion LUT 430. Therefore, in the post conversion, an interpolation operation for the ink amount data INK obtained by the color conversion LUT 430 is not necessary. This point is a great difference from the point that the interpolation calculation is necessary in the first embodiment.

  FIG. 6 shows an example of a pre-conversion table 530a for night scenes and a color conversion LUT 430a for night scenes. The input of the pre-conversion table 530a is 256 pixel values from 0 to 255. The entire range of the input value is divided into a range 0 to 32 of the dense lattice FL and a range 33 to 255 of the coarse lattice CL. There are two outputs of the pre-conversion table 530a: the grid point input value LI and the additional decimal number AF. The difference between the lattice point input values LI is 8 in the range of the dense lattice FL and 16 in the range of the coarse lattice CL. The value of the additional decimal AF depends on the difference between the grid point input values LI. That is, in the example of FIG. 6A, when the input pixel value increases by 1 in the range of the dense grid FL, the additional fractional AF increases by 0.125, whereas in the range of the coarse grid CL, the input pixel value increases by 1. Then, the additional fractional AF increases by about 0.06. The final grid point input value FLI obtained by the pre-conversion using the pre-conversion table 530a is any one of 0, 8, 16, 24, 32, 48, 64... 240, 255. Note that the pre-conversion table 530a is used in common for the three color components RGB.

  The color conversion LUT 430a for night view shown in FIG. 6B is only one of the final grid point input values FLI values 0, 8, 16, 24, 32, 48, 64... 240, 255 obtained by the pre-conversion. Are input values of RGB components. The color conversion LUT 430a corresponds to that shown in FIG. As can be understood from the examples of FIGS. 6A and 6B, the classification of the dense grid FL and the coarse grid CL in the color conversion LUT and the classification of the dense grid FL and the backward CL in the pre-conversion table are mutually Set to match. Accordingly, a pre-conversion table suitable for a plurality of shooting scenes (that is, a table corresponding to a plurality of uneven color conversion LUTs) is prepared in advance.

  FIG. 7 shows an example of pre-conversion using the pre-conversion table 530a of FIG. As shown in FIG. 7A, it is assumed here that the R component Rx of the input pixel value in the 4 × 4 pixel region is a constant value (= 10). At this time, referring to the pre-conversion table 530a in FIG. 6A, as shown in FIG. 7B, a temporary grid point input value NI = 8 and additional decimal number AF = 0.25 are obtained for each pixel. . FIG. 7C shows the result of adding random noise RN to the additional decimal point AF of each pixel. Among the 4 × 4 pixels in FIG. 7C, hatched pixels have an addition result of the additional decimal number AF and the random noise RN of 1.0 or more. Therefore, for these pixels, as shown in FIG. 7D, the final grid point input value FLI is set to a value (= 16) next to the temporary grid point input value LI = 8. It can be understood that the average value of the final grid point input value FLI is about 10, which is substantially equal to the input pixel value Rx.

  FIG. 7E shows a final grid point input value as a comparative example when a pre-conversion table having only a coarse grid CL is used instead of the night view pre-conversion table 530a shown in FIG. ing. The pre-conversion table used here is a table in which the final grid point input value FLI takes any value of 0, 16, 32, 48, 64... 240, 255. When pre-conversion is performed on the input pixel value Rx in FIG. 7A using such a pre-conversion table, the final lattice point input value of each pixel becomes either 0 or 16. Note that the average value of the final grid point input values in FIG. 7E is substantially equal to the input pixel value Rx (= 10), which is the same as in FIG. 7D.

  In the result of the second embodiment shown in FIG. 7D, the final grid point input value FLI of each pixel is either 8 or 16, and the difference between the values of the pixels is small. On the other hand, in the result of the comparative example shown in FIG. 7E, the final grid point input value of each pixel is either 0 or 16, and the difference between the values of the pixels is large. When the difference between the values of the pixels is large, the local variation of the ink amount in the printed material becomes large, and this variation may be seen as image noise when the printed material is observed. This noise is due to pre-post conversion. On the other hand, in the result of the second embodiment shown in FIG. 7D, since the local fluctuation of the final grid point input value is small, the generation of noise due to such pre-post conversion is reduced. Is possible.

  A plurality of pre-conversion tables associated with a plurality of shooting scenes can also be stored in advance in an installer file for installing the printer driver 210 in the computer, as in the color conversion LUT. Alternatively, only a uniform pre-conversion table is stored in the installer file of the printer driver 210, and a pre-conversion table for a plurality of shooting scenes is obtained by interpolating the grid points of the uniform pre-conversion table when the printer driver 210 is installed. You may make it create.

  As described above, in the second embodiment, an appropriate pre-conversion table and color conversion LUT are selected according to the shooting scene of the image determined from the image data file IMG, and the selected pre-conversion table and color conversion LUT are selected. Since color conversion is executed using the color, it is possible to accurately reproduce colors suitable for individual images. In addition, since a table having unequal lattice points including the dense lattice FL and the coarse lattice CL is used as the pre-conversion table, the generation of noise due to the pre-post conversion is reduced in the dense lattice FL portion. Is possible.

C. Third embodiment:
FIG. 8 shows an example of the night scene pre-conversion table 530a and the night scene color conversion LUT 430a in the third embodiment. In the third embodiment, the apparatus configuration and processing flow are the same as those in the second embodiment, and only the configurations of the pre-conversion table and the color conversion LUT are slightly different from those in the second embodiment.

  As shown in FIG. 8A, the pre-conversion table 530a used in the third embodiment outputs a grid point number LN instead of the grid point input value LI (FIG. 6A). The grid point number LN is a continuous number starting from 0, and takes 20 values from 0 to 19 in FIG. In the parentheses in the column of the grid point number LN, the correspondence between the grid point number LN and the grid point input value NI used in the second embodiment is shown for convenience of explanation. FIG. 8B shows a color conversion LUT 430a used in the third embodiment. The color conversion LUT 430a receives a grid point number LN instead of the grid point input value LI (FIG. 6B).

  9A to 9D show an example of pre-conversion using the pre-conversion table 530a in FIG. 8A. As can be understood by comparing with FIGS. 7A to 7D, in the third embodiment, the grid point number LN is obtained instead of the grid point input value LI by referring to the pre-conversion table. Only the point differs from the second embodiment. As shown in FIG. 8B, the input value of the color conversion LUT 430a is the grid point number LN, but the value of the ink amount data as the output is the same as in the second embodiment (FIG. 6B). The same. Therefore, the ink amount data obtained by the pre-post conversion in the third embodiment is the same as that obtained in the second embodiment.

  As described above, in the third embodiment, the grid point number LN, which is a serial number, is used instead of the grid point input value LI that takes values in the same range (0 to 255) as the input pixel value. Also in this case, the same effect as the second embodiment can be obtained. Further, since the grid point number LN has a smaller number of bits than the grid point input value LI, there is an advantage that the data capacity of the pre-conversion table and the color conversion look table can be reduced.

  The grid point number LN used in the third embodiment can also be considered as a grid point input value in a broad sense. That is, the “grid point input value” can be defined as an input value that matches the value of the input grid point of the color conversion lookup table.

D. How to determine the shooting scene:
The scene determination process executed in the scene determination unit 260 will be described with reference to FIGS. FIG. 10 is an explanatory diagram showing functions realized by the scene determination unit 260 in a block diagram. FIG. 11 is a flowchart showing a processing routine for scene determination processing. FIG. 12 is an explanatory diagram conceptually showing an example of a reduced image data buffer for image data centered on the sky. FIG. 13 is an explanatory diagram showing an example of a graph obtained as a result of executing the two-dimensional Fourier transform on the reduced image data buffer shown in FIG. FIG. 14 is an explanatory diagram conceptually showing an example of a reduced image data buffer for image data centered on the sea. FIG. 15 is an explanatory diagram showing an example of a graph obtained as a result of executing the two-dimensional Fourier transform on the reduced image data buffer shown in FIG. FIG. 16 is an explanatory diagram showing an example of a map for determining a photographic scene using the characteristic hue of image data and the texture of the characteristic hue area.

  First, functions realized by the scene determination unit 260 will be described with reference to FIG. The reduced image data generation unit for analysis 301 generates reduced image data for data analysis from the input image data. The image data analysis unit 302 analyzes the generated reduced data for analysis in units of one pixel, and acquires RGB components for each pixel constituting the reduced image data for analysis (image data). The characteristic hue determination unit 303 determines a characteristic hue that is a hue that characterizes the image data, using the analysis result of the analysis reduced image data. The texture extraction unit 304 determines a pixel region corresponding to the characteristic hue determined by the characteristic hue determination unit 303, performs frequency analysis of the reduced image data for analysis by two-dimensional Fourier transform or the like, and sets the characteristic hue. The texture of the corresponding pixel area is extracted. The scene determination unit 305 determines a photographic scene of the photographic image (image data) based on the obtained characteristic hue and the texture of the pixel area that forms the characteristic hue.

  The shooting scene determination process will be described with reference to FIG. The reduced image data generation unit for analysis 301 generates a reduced image data buffer for image data analysis from the input image data (step S110). The image data analysis unit 302 scans the generated reduced image data buffer in the n direction (lateral direction in FIGS. 12 and 14), and acquires the hue information of the image data (step S120). Specifically, the image data analysis unit 302 acquires RGB values in association with the coordinate position (n, m) for each pixel.

  The characteristic hue determination unit 303 determines the characteristic hue of the image data that characterizes the captured image using the acquired hue information (step S130). Specifically, the characteristic hue determination unit 303 determines the target pixel when the target pixel and the adjacent pixel (n direction) have the same hue with respect to all the pixels constituting the reduced image data buffer. Is correlated with the hue, and when the pixel of interest and the adjacent pixel (n direction) do not have the same hue, a process of not counting the pixel of interest is executed. In this embodiment, the hues of blue, green, skin color, and red are determined, and whether the hues are the same is not determined based on the strict identity of the RGB values, but the RGB values are within a predetermined range. If it has the value of, it is determined that the hue is the same.

  The characteristic hue determination unit 303 obtains the ratio of the number of pixels counted for each hue with respect to the total number of pixels constituting the reduced image data buffer, and determines the characteristic hue of the image data using a map prepared in advance. . In the map, for example, the ratio of the number of pixels constituting each hue and the characteristic hue are described in association with each other.

  In the example of FIGS. 12 and 14, since the image data of the landscape image centered on the sky and the image data of the landscape image centered on the sea, the characteristic hue is blue. However, it is unclear whether the blue hue represents the sky or the ocean.

  Therefore, the texture extraction unit 304 acquires the texture of the determined characteristic hue area (step S140). Specifically, the texture extraction unit 304 identifies a pixel area that forms a characteristic hue, and performs frequency analysis on the identified pixel area. The pixel area constituting the characteristic hue is specified based on the hue and coordinate position information of each pixel included in the hue information. The frequency analysis for the identified pixel region is executed in the n direction and m direction of the analysis reduced image data buffer using a two-dimensional Fourier transform equation.

  The relationship between the results obtained by the two-dimensional Fourier transform and the texture will be described with reference to FIGS. The reduced image data buffer for analysis shown in FIG. 12 corresponds to an image obtained by photographing a landscape centering on the sky as described above, and the characteristic hue is blue. When the two-dimensional Fourier transform is performed on the analysis reduced image data buffer, a graph shown in FIG. 13 is obtained. In general, an image corresponding to the sky is smooth and often has no pattern. Therefore, there are many low frequency components and few high frequency components.

  On the other hand, the reduced image data buffer for analysis shown in FIG. 14 corresponds to an image of a landscape centered on the sea as described above, and the characteristic hue is blue. When the two-dimensional Fourier transform is performed on the analysis reduced image data buffer, a graph shown in FIG. 15 is obtained. Generally, an image corresponding to the sea has many patterns caused by ripples and the like. Therefore, there are few low frequency components and many high frequency components tend to appear.

  Considering these tendencies, even if the characteristic hues are the same, it is possible to further identify the image features (subjects) by determining the texture of the area that forms the characteristic hues. It is. Therefore, the shooting scene can be specified with high accuracy.

  The scene determination unit 305 determines a shooting scene using the acquired characteristic hue and the texture of the pixel region corresponding to the characteristic hue (step S150). For example, a map shown in FIG. 16 is used to determine the shooting scene.

(1) When the characteristic hue is green and the extracted texture has a high frequency, it is determined (determined) to be a landscape centered on green (green scene) such as a mountain or a plain.
(2) When the characteristic hue is blue and the extracted texture has a low frequency, it is determined that the scene is a sky-centered landscape (sky-blue scene).
(3) When the characteristic hue is blue and the extracted texture has a high frequency, it is determined that the scene is a landscape centered on the sea (sea color scene).
(4) When the characteristic hue is skin color and the extracted texture has a low frequency, it is determined that the scene is a person-captured scene (portrait) centered on a person.
(5) When the characteristic hue is skin color and the extracted texture has a high frequency, it is determined to be a beach or a similar landscape (beach scene).
(6) If the characteristic hue is gray and the extracted texture has a low frequency, it is determined to be a night scene.

  Since six shooting scenes are determined in the example of FIG. 16, color conversion LUTs (color conversion LUTs and pre-conversion tables in the second and third embodiments) respectively associated with these six shooting scenes are determined. It is preferable to prepare.

  As described above, the scene determination unit 260 determines the shooting scene using the characteristic hue obtained by analyzing the image data and the texture of the region constituting the characteristic hue. Therefore, based on the analysis of the image data. Accurate shooting scene determination.

  That is, since there is a hue that characterizes the shooting scene in each shooting scene, the shooting scene can be specified from the image data by comparing the characteristic hue that characterizes the image data with the hue that characterizes each shooting scene. In addition, since the texture is extracted by performing frequency analysis on the constituent pixels constituting the region formed by the characteristic hue of the image data, a plurality of shooting scenes have the same hue as the characteristic hue. Also, the shooting scene can be determined appropriately.

  Therefore, even if the image data file IMG (FIG. 1) does not have information related to the shooting scene, it is possible to specify the shooting scene from only the image data, and the image data can be specified without bothering the user. Thus, color conversion corresponding to the shooting scene can be executed.

  In the above embodiment, the shooting scene is determined using the characteristic hue of the image data and the texture of the characteristic hue area. However, the shooting scene is also detected using additional information such as shooting time information describing shooting conditions at the time of shooting. You may judge. Such shooting time information can describe a shooting scene set in the digital still camera. Therefore, by using one or both of the shooting scene set in the shooting time information and the shooting scene obtained by analyzing the image data, the shooting scene is specified, and color conversion suitable for the shooting scene is performed. Can be executed.

  In the above embodiment, the analysis reduced image data buffer is used when obtaining the hue and frequency of the image data, but original image data (non-reduced image data) may be used. In such a case, more accurate analysis can be performed.

  In the above embodiment, the shooting scene is specified using the characteristic hue and texture. However, even if the shooting scene is determined using only the characteristic hue or using the characteristic hue and shooting time information. good.

  In the scene determination of the portrait image centering on the person in the above embodiment, pattern matching regarding the edges of the entire face part, pattern matching of each face part, and determination using a learned neural network may be performed. In such a case, it can be more accurately determined whether or not the image data to be processed is image data corresponding to a photographed image centered on a person. Further, since the face area can be extracted from the person area, it is determined that the face area is dark when the face area is dark, and correction (gamma correction) that makes the dark part brighter than normal brightness correction may be executed. .

  In the above embodiment, two-dimensional Fourier transform is used for texture extraction, but wavelet transform or the like may be used in addition to this. It suffices if frequency analysis (texture extraction) of image data can be executed.

  In the above embodiment, the characteristic hue of the image data is obtained, and texture extraction (frequency analysis) is performed on the area formed by the characteristic hue (the constituent pixels constituting the area). Alternatively, texture extraction may be performed. In such a case, the texture of the region constituting the characteristic hue is acquired by associating the result of texture extraction with the region (position) information on the image data and comparing it with the region information of the characteristic hue. be able to.

E. Variations:
The present invention is not limited to the above-described examples and embodiments, and can be implemented in various modes without departing from the gist thereof. For example, the following modifications are possible.

E1. Modification 1:
In the second and third embodiments, when determining the final grid point input value FLI from the temporary grid point input value LI obtained by referring to the pre-conversion table, the random noise RN is added to the additional decimal point AF. However, the final lattice point input value may be determined using another additional value instead of the additional decimal AF. For example, as an additional value, an integer value of a predetermined number of bits is given according to the input pixel value, and a final lattice is obtained by comparing the addition result obtained by adding random noise to a predetermined threshold value. The point input value FLI may be determined. In this specification, an additional value for determining the final grid point input value FLI is referred to as an “addition determination value”.

E2. Modification 2:
Instead of the pre-conversion method used in the second and third embodiments, other pre-conversion methods can be used. For example, in the second and third embodiments, noise is randomly generated. However, noise may be generated according to a predetermined noise pattern such as a dither pattern. As pre-conversion (first conversion), the pixel value of the input image data is stochastically converted to a grid point input value corresponding to one of the non-uniform grid points of the color conversion lookup table suitable for the shooting scene. It is possible to adopt a conversion method.

E3. Modification 3:
Although the case where printing is performed using four types of CMYK inks has been described in the above embodiment, the present invention can be applied to the case where any type of ink is used.

E4. Modification 4:
In the above embodiment, an example in which the present invention is applied to an ink jet printer has been described. However, the present invention can also be applied to other types of printers such as a laser printer, and can also be applied to a facsimile machine and a copier. . When applied to a printer using toner, the toner corresponds to “ink”. The term “ink” in this specification includes narrowly defined ink and toner, and generally means an image forming agent (dot forming agent) for forming dots.

It is explanatory drawing which shows the structure of the image processing system as one Example of this invention. It is a block diagram which shows the internal structure of the color conversion part in 1st Example. It is explanatory drawing which shows the input lattice point of several color conversion LUT. It is a block diagram which shows the internal structure of the color conversion part in 2nd Example. It is explanatory drawing which shows the outline | summary of pre-conversion and post-conversion. It is explanatory drawing which shows an example of the pre-conversion table for night views and the color conversion LUT for night views in 2nd Example. It is explanatory drawing which shows the specific example of the preconversion in 2nd Example. It is explanatory drawing which shows an example of the pre-conversion table for night views and the color conversion LUT for night views in 3rd Example. It is explanatory drawing which shows the specific example of the preconversion in 3rd Example. It is explanatory drawing which shows the function implement | achieved by the scene determination part 260 with a block diagram. It is a flowchart which shows the process routine of a scene determination process. It is explanatory drawing which shows notionally an example of the reduction image data buffer of the image data centering on the sky. It is explanatory drawing which shows an example of the graph obtained as a result of performing two-dimensional Fourier-transform with respect to the reduced image data buffer shown in FIG. It is explanatory drawing which shows notionally an example of the reduction image data buffer of the image data centering on the sea. It is explanatory drawing which shows an example of the graph obtained as a result of performing two-dimensional Fourier-transform with respect to the reduction image data buffer shown in FIG. It is explanatory drawing which shows an example of the map for determining a picked-up scene using the characteristic hue of image data, and the texture of a characteristic hue area | region.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 ... Image processing system 200 ... Computer 210 ... Printer driver 220 ... Resolution conversion part 230 ... Color conversion part 240 ... Halftone processing part 250 ... Data arrangement part 260 ... Scene determination part 300 ... Printer 301 ... Reduction image data generation part for analysis 302 ... Image data analysis unit 303 ... Characteristic hue determination unit 304 ... Texture extraction unit 305 ... Scene determination unit 420 ... Look-up table selection unit 430 ... Color conversion lookup table 440 ... Interpolation execution unit 500 ... Pre-conversion unit 510 ... Table Reference unit 520 ... Table selection unit 530 ... Pre-conversion table 540 ... Random noise addition unit 550 ... Grid value determination unit 600 ... Post-conversion unit

Claims (8)

A color conversion device that converts image data into ink amount data,
A table storage unit for storing a plurality of color conversion lookup tables associated with a plurality of shooting scenes;
A scene determination unit that determines a shooting scene based on an image data file including image data;
A table selection unit that selects an appropriate color conversion lookup table from the plurality of color conversion lookup tables according to the determined shooting scene;
A color conversion execution unit that converts the image data into ink amount data using the selected color conversion lookup table;
With
The plurality of color conversion lookup tables each have unequal input grid points having unequal intervals with respect to pixel values of the image data, and color conversion lookup tables associated with different shooting scenes. Have different unequal input grid points. The color conversion device according to claim 1,
The table storage unit is further a plurality of input value conversion tables associated with the plurality of shooting scenes, and the pixel values of the image data are any of unequal input grid points of a color conversion lookup table. A plurality of input value conversion tables for converting into grid point input values corresponding to
The table selection unit further selects an appropriate input value conversion table from the plurality of input value conversion tables according to the determined shooting scene,
The color conversion execution unit
Using the selected input value conversion table, the pixel value of the image data is probabilistically converted to a grid point input value corresponding to one of the non-uniform grid points of the selected color conversion lookup table. 1 conversion unit;
A second converter that converts the grid point input value into ink amount data using the selected color conversion lookup table;
A color conversion device. The color conversion device according to claim 2,
Each input value conversion table is a table that outputs a grid point input value corresponding to one of the input grid points of the color conversion lookup table and an additional grid point determination value according to the input pixel value. Yes,
The first conversion unit includes:
By referring to the selected input value conversion table using the pixel value of the image data, a grid point input value corresponding to any of the non-uniform grid points of the selected color conversion lookup table, and A table reference unit for acquiring additional lattice point determination values;
A random noise adding unit that adds random noise to the lattice point determination value;
A lattice point determination unit that determines a lattice point input value to be input to the color conversion lookup table by adjusting the lattice point input value according to a lattice point determination value to which the random noise is added;
A color conversion device comprising: The color conversion device according to any one of claims 1 to 3,
The scene determination unit
A hue determination unit that analyzes the image data and determines a characteristic hue that characterizes the image data;
A scene determination unit that determines the shooting scene using the determined characteristic hue;
A color conversion device comprising: The color conversion device according to claim 4,
The scene determination unit further includes:
An area specifying unit for specifying an area corresponding to the determined characteristic hue in the image represented by the image data;
A texture extraction unit for extracting the texture of the specified region;
With
The scene determination unit is a color conversion device that determines the photographic scene using the extracted texture in addition to the determined characteristic hue.
Color conversion device. The color conversion device according to any one of claims 1 to 3,
The image data file has additional information including a shooting scene setting value set when the image represented by the image data is shot.
The scene determination unit is a color conversion device that determines the shooting scene from the shooting scene setting value in the additional information. A method for converting image data into ink amount data,
(A) determining a shooting scene based on an image data file including image data;
(B) selecting an appropriate color conversion lookup table from a plurality of color conversion lookup tables associated with a plurality of shooting scenes according to the determined shooting scene;
(C) converting the image data into ink amount data using the selected color conversion lookup table;
With
The plurality of color conversion lookup tables each have unequal input grid points having unequal intervals with respect to pixel values of the image data, and color conversion lookup tables associated with different shooting scenes. Have different unequal input grid points, and a color conversion method. A printer driver that creates print data for a printer that performs printing by recording ink dots,
A color converter for converting image data into ink amount data;
Based on the ink amount data, a halftone processing unit that generates dot data representing an ink dot formation state;
With
The color converter is
A table storage unit for storing a plurality of color conversion lookup tables associated with a plurality of shooting scenes;
A scene determination unit that determines a shooting scene based on an image data file including image data;
A table selection unit that selects an appropriate color conversion lookup table from the plurality of color conversion lookup tables according to the determined shooting scene;
A color conversion execution unit that converts the image data into ink amount data using the selected color conversion lookup table;
With
The plurality of color conversion lookup tables each have unequal input grid points having unequal intervals with respect to pixel values of the image data, and color conversion lookup tables associated with different shooting scenes. Has different unequal input grid points.

Images