JP2014192859A - Color correction method, program, and device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a color correction method, etc., having excellent reproducibility of colors closer to actual colors.SOLUTION: The color correction method includes an image acquisition step S10, a tone curve generation step S11, and a correction step S12. The tone curve generation step S11 includes: a first sub-step S20 for plotting, for each of RGB color elements on two-dimensional coordinates, a set of pixel value and ideal value of a pixel targeted for a patch of multiple monochrome gradations on a color chart included in a first image and calculating an approximate linear line in the plot, and a second sub-step S21, etc., for setting a range including all sets in the plot by moving the calculated approximate linear line in parallel and calculating an approximate cubic curve using a set of pixel value and ideal value of a pixel targeted for a patch of multiple color gradations in the color chart so that the approximate cubic curve fits within the set range.

Description

  The present invention relates to a color correction method, program, and apparatus for color digital images expressed by RGB (red, green, and blue) color elements.

  Color digital images represented by RGB color elements are obtained by photographing with a digital camera or the like. However, when shooting with a camera, recordings different from the actual color are recorded due to differences in light sources such as sunlight and LED light, and changes in color due to the characteristics of the camera lens. Therefore, color correction is performed on the digital image obtained by photographing.

  The color correction for the digital image can be manually adjusted in consideration of the color balance and the like using photo editing software such as Adobe Photoshop (registered trademark of Adobe Systems) depending on the subjectivity of the editor. Such manual color correction lacks objectivity and reproducibility to actual colors.

  Therefore, conventionally, when a subject is photographed, a color chart (color sample) is also photographed, and the pixel values in the color chart included in the obtained digital image are used to automatically correct the color of the digital image. The technique which performs is proposed (for example, refer patent document 1, 2).

  In Patent Document 1, a color chart and a subject are photographed on a film. Then, a coefficient for displaying the color of each color chart of the color chart in a correct color is calculated. Finally, the image data of the subject is corrected using this coefficient. Thus, an image with a correct color can always be reproduced regardless of shooting conditions.

  In Patent Document 2, an ideal RGB value for each patch of the color chart is set. Then, first to third tone curves representing the R value, G value, and B value of the original image of the photographed color chart are set. Furthermore, an inflection point is set on each tone curve, the optimal value derivation method is repeatedly used, the tone curve is deformed by displacing the inflection point, and the target image is colored based on the tone curve that satisfies a certain condition. Convert. Thereby, the color of the image of the same subject photographed by the digital camera can be reproduced to an approximate color regardless of the photographing condition and the type of camera.

Japanese Patent Laid-Open No. 6-46448 JP 2002-281521 A

  However, although the color correction described in Patent Documents 1 and 2 can reproduce the target image with an approximate color regardless of the shooting conditions and the type of camera, it is not always sufficient in reproducibility to an actual color. it is conceivable that. In Patent Document 1, the gray scale is not distinguished as another color, and in Patent Document 2, emphasis is placed on two colors that are considered to have the largest variation in numerical values even within the same color due to white and black skipping or blackout. This is because correction is performed in the above.

  Accordingly, the present invention has been made in view of such problems, and an object thereof is to provide a color correction method, program, and apparatus that are more excellent in reproducibility to actual colors.

  In order to achieve the above object, one aspect of the color correction method according to the present invention is a color correction method for a color digital image expressed by RGB (red, green, and blue) color elements. An image acquisition step of acquiring a first image which is a digital image generated by photographing a subject together with a color chart composed of a plurality of gradation patches, and a color chart included in the acquired first image A tone curve for generating a first tone curve used for correcting the color of the first image for each of the RGB color elements, using the pixel value of the pixel constituting the patch and the ideal value of the pixel. A generation step; and a correction step for correcting the color of the first image using the generated first tone curve, wherein the tone curve generation step includes: For each color element of GB, a set of pixel values and ideal values of pixels is plotted on a two-dimensional coordinate for a plurality of monochrome patches in the color chart included in the first image, The first sub-step for calculating the approximate primary line in the plot and the calculated approximate primary line are translated to set a range that includes all the sets in the plot, and is approximated within the set range. A second sub-step of calculating an approximate cubic curve using a set of pixel values and ideal values of pixels for a plurality of color patches in the color chart so that a cubic curve is included; And a third sub-step of outputting the calculated approximate cubic curve as the first tone curve.

  As a result, the first processing unit calculates an approximate linear line using the gray scale in the color chart, so that it is possible to perform color correction that adjusts the overall brightness while adjusting the color balance in the gray scale. Become. Then, the second processing unit uses color points (a pair of pixel values and ideal values) so that a curve enters a range that includes all the points of gray scale (a pair of pixel values and ideal values). Since the approximated cubic curve is calculated, it is possible to perform color correction that reduces the error while maintaining the color balance adjusted by the first processing unit 14a. By creating a tone curve in two steps, brightness adjustment using only gray scale and fine adjustment using color while maintaining the result, the reproducibility to actual colors is better than before. Color correction is possible.

  Here, in the image acquisition step, at least one second image, which is a digital image generated by shooting a subject under the same shooting conditions as the shooting in the generation of the first image, is acquired, and the tone curve generation step Then, using the color chart included in the first image, a second tone curve used for correcting the color of the second image is generated for each of the color elements. The second image may be color corrected using the second tone curve.

  This makes it possible to perform similar color correction even for digital images that have not been photographed together with a color chart.

  In the third sub-step, for each of the RGB color elements, the approximate cubic is such that the ratio of whiteout and blackout in the approximate cubic curve calculated in the second substep is small. A curve may be shifted in the axial direction of the ideal value in the two-dimensional coordinates, and the obtained approximate cubic curve may be generated as the first tone curve. More specifically, in the third sub-step, each of the RGB color elements is set such that the sum of the ratios of whiteout and blackout in the approximate cubic curve with respect to each of the RGB color elements is minimized. The approximate cubic curve for can be shifted in the same way.

  As a result, in the case where there are whiteout and blackout in the first tone curve, the color information of the bright part and dark part of the digital image is lost, and the problem that the actual color cannot be reproduced by color correction is avoided. . That is, loss of color information can be prevented by adjusting the overall brightness while maintaining the color balance.

  Further, the image processing method further includes a color chart position recognition step of acquiring the position of the color chart in the first image from a user, and the pixel values of the pixels constituting the patch of the color chart included in the first image are The position of the color chart in the first image is specified based on the position acquired in the color chart position recognition step, the position of the specified color chart, and the predetermined position of the color chart and the patch The position of the patch may be specified by using the relationship, and the pixel value of the pixel constituting the patch may be specified based on the specified position of the patch.

  As a result, the position of each patch is recognized only by indicating the position of the color chart, and thus the process of acquiring the pixel value of each patch is automated.

  The present invention can be realized not only as a color correction method as described above, but also as a program for causing a computer to execute the color correction method, or a computer-readable recording medium such as a CD-ROM storing the program. Or a color correction device realized by a program and a CPU that executes the program.

  The color correction method, program, and apparatus according to the present invention automatically and logically and objectively perform color correction on a color digital image obtained by photographing, so that an actual color can be obtained with higher accuracy than before. It is reproduced.

  Therefore, the significance of the present invention in the present day when high definition of digital images has progressed is extremely high.

The figure which shows the example of the color chart used by embodiment of this invention The figure which shows the example of the digital image (photograph) used as the object of the color correction in embodiment of this invention A plot of the correspondence (points) between pixel values and ideal values of each patch in the color chart shown in the photograph of FIG. The figure which shows the difference between the pixel value of the 24 color patch of the color chart reflected in the photograph of FIG. 2, and an ideal value The figure which shows the color balance in the gray scale of the color chart reflected in the photograph of FIG. The block diagram which shows the function structure of the color correction apparatus in embodiment of this invention The figure explaining the position recognition process of the color chart and patch by the same color correction apparatus Flow chart showing the operation of the same color correction device FIG. 8A is a flowchart showing details of the tone curve generation step. The figure which shows the example of the approximate primary straight line computed by the 1st process part of the same color correction apparatus The figure which shows the example of the approximate cubic curve calculated by the 2nd process part of the same color correction apparatus The figure which shows the example of the process by the 3rd process part of the same color correction apparatus The figure which shows the analysis result of the image correct | amended by the same color correction apparatus

(Knowledge that became the basis of the present invention)
Before describing specific aspects of the color correction method, program, and apparatus according to the present invention, first, knowledge that is the basis of the present invention will be described.

  The inventor of the present invention uses a digital camera that captures a color chart shown in FIG. 1 together with a subject to obtain a 40-megapixel digital image (hereinafter, the digital image is also referred to as a “photograph”). The difference between the actual RGB value (ideal value) and the RGB value (pixel value) of the photograph was analyzed. The photograph used for the analysis is, for example, a photograph as shown in FIG. In this specification, “actual color (RGB value)” refers to the color (RBG value) of an actual subject and a color chart, that is, an ideal value. Here, the RGB value (or pixel value) is the brightness of each color, and is expressed in 256 levels from 0 to 255 in the present embodiment. For example, if all three colors are 0, it is black, and if all three colors are 255, it is white.

  In addition, as shown in FIG. 1, the color chart used here is a patch of gray color with 6 gradations and 18 color patches made by combining 6 gradations of R, G, and B (a total of 24 patches). X-Rite Corporation. The actual RGB values (ideal values) in the color chart were the values described in the data sheet attached to the color chart. As a digital camera, PENTAX 645D was used. For the analysis, numerical calculation software MATLAB (registered trademark of MathWorks) was used.

  FIG. 3 shows each patch of the color chart shown in the photograph for each RGB color element ((a) in FIG. 3 is red, (b) in FIG. 3 is green, (c) blue in FIG. 3)). It is the figure which plotted correspondence (point) with the pixel value of a pixel to perform, and the ideal value of the pixel concerned. Here, the X axis is the pixel value, and the Y axis is the ideal value. For each RGB color element, 24 points corresponding to a total of 24 patches are plotted. In addition, the pixel value of the pixel which comprises one patch is a representative value, for example, is the average of the pixel value of the some pixel which belongs to the said patch. Hereinafter, “a pixel value of a pixel constituting one patch (or simply“ a pixel value of one patch ”)” represents a representative value of the pixel value of the patch. The straight lines shown in each of FIGS. 3A to 3C are reference straight lines, and are a set of points where the X coordinate value and the Y coordinate value are the same (hereinafter, this straight line is referred to as a straight line). Called "ideal straight line"). If the color of the photograph is the same as the color of the actual color chart, the points will be plotted on this straight line.

  From this figure, it can be seen that the pixel value of the photograph is lower than the ideal value in any element of RGB, and is dark overall (low brightness). Also, comparing the three colors, the position of each point in blue is closer to the ideal straight line, so the difference between the red and green elements as a whole may be a factor that increases the difference from the actual color. It is done. However, while the red and green dots are aligned in a straight line, the positions of the blue dots vary, indicating that blue is particularly difficult to correct.

  By the way, the color balance of the three elements of RGB can be evaluated by the gray scale (monochrome gradation). The gray scale is a color obtained by mixing white and black (including white and black itself) and is synonymous with an achromatic color. An achromatic color has a saturation of 0, and when expressed in RGB, the values of each element must all be the same (R value = G value = B value). Table 1 below shows the color balance seen in the six-step gray scale in the above picture. That is, for each of the six types of patches in monochrome of the color chart shown in the photograph, the RGB value (pixel value) and the difference between the maximum value and the minimum value of the RBG value in the patch are shown.

  In Table 1, the value of each element is different for any color, with red being low and blue being high. It can also be seen that the difference decreases as the brightness decreases (from white to black). From this, it can be said that the photograph is bluish as a whole, and the tendency is more prominent in the bright part.

  FIG. 4 shows each color element of RGB (FIGS. 4A and 4D are red, FIGS. 4B and 4E are green, and FIGS. 4C and 4F are blue). FIG. 10 is a diagram illustrating a difference between a pixel value and an ideal value in a patch of 24 colors of the color chart shown in the photograph. That is, (a) to (c) in FIG. 4 correspond to (a) to (c) in FIG. 3, respectively. However, in FIGS. 4A to 4C, in addition to the ideal straight line in FIG. 3, approximate linear lines determined by the least square method or the like from 24 points parallel to the ideal straight line are also drawn. . Also, (d) to (f) in FIG. 4 are histograms showing the frequency distribution (100 fraction) for each pixel value. The horizontal axis indicates the pixel value, and the vertical axis indicates the 100 fraction.

  From (a) to (c) of FIG. Since the approximate primary straight line and the ideal straight line in FIGS. 4A to 4C are both straight lines through which the pixel value and the ideal value pass the same value, the distance between these two straight lines is calculated. It is possible to compare the color balance according to lightness and the lightness between actual and photograph. Since all of the three elements of RGB have an approximate primary line above the ideal line, it can be seen that the picture is darker than the actual one. The degree is expressed as an average value of the difference between actual and photograph. There are three colors, about 60 for red, about 50 for green, and about 27 for blue. When attention is paid to the relationship between the position of the 24 points and the position of the approximate primary line, in all three elements of the RBG, many points are below the approximate primary line in the low pixel value (lightness) range. There are many points above the approximate linear line in the high pixel value range. From this, it can be said that the dark part of the photograph is darker than the actual part, and the bright part is brighter than the actual part.

  On the other hand, the following can be understood from FIGS. Several peaks are observed in all three elements of RGB, but the ratio (100 fraction) of the number of pixel values in the low value range and the high value range is almost zero. Here, since the color correction is performed based on the color in the color chart, the ratio occupied by the color and the amount of error as a whole from the ratio are expressed numerically as follows. That is, the proportion of the color chart color is about 11% for red, 15% for green, and about 12% for blue. The errors seen in the entire photo were 3.7 for red, 1.5 for green, and 6.5 for blue. The errors shown in FIGS. 4A to 4C are about 10.5 for red, about 3.2 for green, and about 4.6 for blue. Therefore, for red and green, the errors in (d) and (e) in FIG. 4 are about half smaller than the errors in (a) and (b) in FIG. It can be said that the error is large when the ratio is small and the error is small when the ratio is large. On the other hand, for blue, the error of (f) of FIG. 4 is 1.4 times larger than the error of (c) of FIG. It is considered large.

  FIG. 5 is a diagram showing the color balance in the gray scale (monochrome six-color patch) of the color chart shown in the photograph. Here, the upper stage ((a) to (f) in FIG. 5), the middle stage ((g) to (l) in FIG. 5), and the lower stage ((m) to (r) in FIG. 5) are R, G and B values, which are separated from left to right and corresponding to six gray scale colors from white to black. In each figure, the number of pixels of the value indicated by each color is displayed as a histogram (the horizontal axis is the pixel value, and the vertical axis is the frequency (100 fraction)). As can be seen from the figure, the width of the value is about 20 for each color element, and its peak is in the form of a normal distribution exceeding 15%. This confirms that there is little variation in the same color and that one point of each color can be set as a representative value. In addition, when looking at each color, the value of each element is different in any color, with red being low and blue being high. The result is the same as the result shown in the average value of the difference between the actual image and the photograph shown in FIG. It can also be seen that the difference decreases as the brightness decreases (from white to black).

  From the above analysis results, when editing a photo, color correction to approximate the values of the actual color chart is as follows: (1) Overall brightness (all color elements are brightened, but the extent is red, green, blue (2) The fine adjustment to reduce the error as much as possible while maintaining the color balance caused by (1) (particularly in R and G, the error in the portion that occupied a large proportion was small) It is understood that the following two are necessary: maintaining the value and reducing the error with respect to the value with a small proportion of the whole).

  Based on the above knowledge, the present inventor performs color correction automatically and logically and objectively on color digital images, and a color correction method and program capable of reproducing actual colors with higher accuracy than before. In the following, embodiments of the present invention will be described with reference to the drawings. Each figure is a schematic diagram and is not necessarily illustrated exactly.

  The embodiment described below shows a specific example of the color correction method, program, and apparatus according to the present invention. The numerical values, components, arrangement positions and connection forms of the components shown in the following embodiments are examples, and are not intended to limit the color correction method, program, and apparatus according to the present invention. In addition, among the constituent elements in the following embodiments, constituent elements that are not described in the independent claims are described as arbitrary constituent elements.

  FIG. 6 is a block diagram showing a functional configuration of the color correction apparatus 10 according to the embodiment of the present invention. The color correction device 10 is a device that performs color correction on a color digital image (photograph) expressed by RGB color elements, and includes an input unit 11, a control unit 12, a display unit 16, and a storage unit. 17. In the figure, a digital camera 20 is also shown as an example of a data source from which a photograph is acquired.

  The input unit 11 is an input device that acquires an instruction from the user, and is, for example, a mouse, a keyboard, or the like.

  The display unit 16 is a display that displays photographs before and after correction, analysis results obtained by the color correction apparatus 10, and the like, for example, an LCD.

  The storage unit 17 is a memory that stores photographs before and after correction, data that is temporarily generated, and the like, for example, a nonvolatile memory such as an HDD (Hard Disk Drive).

  The control unit 12 is a processing unit that performs color correction on a photograph acquired from the digital camera 20, and includes an image acquisition unit 13, a tone curve generation unit 14, and a correction unit 15. The control unit 12 is realized by, for example, a personal computer (PC) having a control program and a CPU that executes the control program. Specifically, it is realized by creating a color correction program using the above-mentioned MATLAB (registered trademark of MathWorks) and executing the color correction program on a computer on which MATLAB (registered trademark of MathWorks) is installed. Alternatively, the created color correction program is converted into an independent application program using a MATLAB (registered trademark of MathWorks), and the converted application program is executed on a general computer. May be.

  The image acquisition unit 13 acquires a first image that is a digital image (photograph) generated by photographing a subject together with a color chart composed of a plurality of color and monochrome patches. Here, the image acquisition unit 13 acquires a photo from the digital camera 20.

  The tone curve generation unit 14 uses the pixel values of the pixels constituting the patch of the color chart included in the first image acquired by the image acquisition unit 13 and the ideal values of the pixels for each of the RGB color elements. A processing unit that generates a first tone curve used to correct the color of the first image, and includes a first processing unit 14a, a second processing unit 14b, and a third processing unit 14c.

  The first processing unit 14a, for each of the RGB color elements, has a plurality of monochrome gradation patches (here, six gradation patches) in the color chart included in the first image acquired by the image acquisition unit 13. A set of pixel values and ideal values of pixels is plotted on a two-dimensional coordinate, and an approximate linear line in the plot is calculated.

  The second processing unit 14b translates the approximate linear straight line calculated by the first processing unit 14a for each of the RGB color elements, so that all the plots by the first processing unit 14a for monochrome are processed. A plurality of color patches in the color chart included in the first image (six-tone patches in each of RGB, that is, a patch of each of RGB, that is, an approximate cubic curve is set within the set range. , An approximate cubic curve using a set of pixel values and ideal values of pixels for a total of 18 patches) is calculated.

  The third processing unit 14c outputs the approximate cubic curve calculated by the second processing unit 14b as a first tone curve. At this time, as an optional process, the third processing unit 14c is configured to reduce the ratio of overexposure and underexposure in the approximate cubic curve calculated by the second processing unit 14b for each of the RGB color elements. The approximate cubic curve is shifted in the direction of the ideal value axis (in this case, the Y axis) in the two-dimensional coordinates, and the obtained approximate cubic curve is generated as the first tone curve. Specifically, the third processing unit 14c is configured so that the sum of the ratio of overexposure and underexposure in the approximate cubic curve for each of the RGB color elements calculated by the second processing unit 14b is minimized. The first tone curve is generated by shifting the three approximate cubic curves in the same manner.

  Note that “out-of-white” means that the value before correction (pixel value) is less than the maximum value (255) in the tone curve, but the value after correction (correction value) is the maximum value (255). An area that ends up. Further, “blackout” means that the value before correction (pixel value) in the tone curve is larger than the minimum value (0), but the value after correction (correction value) becomes the minimum value (0). An area.

  The correction unit 15 corrects the color of the first image using the first tone curve generated by the tone curve generation unit 14.

  Note that the image acquisition unit 13 may acquire at least one second image that is a digital image (photograph) generated by shooting a subject under the same shooting conditions as the shooting in the generation of the first image. This can be selected multiple times. At this time, the tone curve generation unit 14 generates a second tone curve used to correct the color of the second image for each of the RGB color elements, using the color chart included in the first image. . Then, the correction unit 15 corrects the color of the second image using the second tone curve generated by the tone curve generation unit 14. As a result, color correction is possible even for a photograph in which a color chart is not shown. Here, the second tone curve is basically the same as the first tone curve. However, it is necessary to determine the necessity and the degree of shift for an image without a color chart only for the tone curve shift processing to prevent overexposure and blackout. That is, the first tone curve and the second tone curve are different only when the tone curve needs to be shifted. For this purpose, a step of generating a second tone curve by adding a constant to the first tone curve may be added. When no shift is necessary, the constant is 0 and the first tone curve and the second tone curve are the same.

  Note that the position of the color chart in the first image is recognized based on a user instruction acquired from the input unit 11. At this time, the pixel values of the pixels constituting the patch of the color chart included in the first image are first determined based on the position instructed by the user in the first image of the color chart (here, the positions of the four corners). ) Is specified, and then the position of each patch is specified by using the position of the specified color chart and the positional relationship between the predetermined color chart and each patch. Based on the position of each patch, the pixel values of the pixels constituting each patch are specified.

  Specifically, as shown in the color chart display example of FIG. 7, the user clicks the crosses at the four corners of the color chart shown in the photograph with the mouse. Thereby, the position of the color chart is recognized. Since pixel values vary around the boundary of each patch, clipping is performed slightly inside (cutting out within a rectangular area surrounded by four straight lines in each patch in FIG. 7), and RGB values of each color are unified. Instead of being represented by one pixel, the average of the pixel values of all the pixels in each rectangular area that is divided into squares is taken, and that value is used as the RGB value (pixel value) of each patch. Thereby, the color information of the color chart can be obtained more accurately.

  Next, the operation (color correction method according to the present invention) of the color correction apparatus 10 according to the present embodiment configured as described above will be described.

  FIG. 8A is a flowchart showing the operation of the color correction apparatus 10 in the present embodiment.

  First, the image acquisition unit 13 acquires a photograph as a first image (image acquisition step S10). Specifically, the image acquisition unit 13 acquires an image file such as JPEG generated by the digital camera 20 via a USB cable or a memory card, and a photograph to be corrected (the storage unit 17 in FIG. 6). Is stored in the storage unit 17 as “original image” shown in FIG.

  Then, the tone curve generation unit 14 uses the pixel values of the pixels constituting the patch of the color chart included in the first image acquired by the image acquisition unit 13 and the ideal values of the pixels to calculate the RGB color elements. For each, a first tone curve used to correct the color of the first image is generated (tone curve generation step S11). Specifically, the tone curve generation unit 14 generates a first tone curve using a color chart shown in a photograph (“original image”) stored in the storage unit 17 and stores the first tone curve in the storage unit 17. ("Tone curve" shown in the storage unit 17 in FIG. 6).

  Finally, the correction unit 15 performs color correction on the first image using the first tone curve generated by the tone curve generation unit 14 (correction step S12). Specifically, the correction unit 15 includes a photograph stored in the storage unit 17 (“original image” shown in the storage unit 17 in FIG. 6) and a first tone curve (“shown in the storage unit 17 in FIG. 6”. The color correction is performed on the photograph using the “tone curve”), and the corrected photograph is stored in the storage unit 17 as the “corrected image” and displayed on the display unit 16.

  FIG. 8B is a flowchart showing a detailed procedure of the tone curve generation step S11 in FIG. 8A.

  In the generation of the first tone curve, first, the first processing unit 14a, for each of the RGB color elements, patches of a plurality of monochrome gradations in the color chart included in the first image acquired by the image acquisition unit 13. A set of pixel values and ideal values of pixels is plotted on a two-dimensional coordinate target (here, a patch of 6 gradations), and an approximate linear line in the plot is calculated (first sub-step S20). Specifically, the first processing unit 14 a reads a photograph (“original image”) from the storage unit 17, and targets six patches in the gray scale of the color chart shown in the photograph as two-dimensional coordinates ( The X-axis is a coordinate value and the Y-axis is an ideal value), and an approximate linear line of the plot is calculated using a least square method or the like.

  FIGS. 9A to 9C are diagrams showing examples of approximate primary lines calculated by the first processing unit 14a for RGB color elements. Note that the approximate linear line shown in this figure is calculated for only the six gray scale patches described above, but the points plotted are 18 for color patches for reference. A total of 24 points including the points are shown.

  Since the ideal value of grayscale is the same for all three elements of RGB, the color balance in grayscale is adjusted by deriving an approximate linear line using only grayscale in the first sub-step S20. However, a tone curve (approximate linear line) for adjusting the overall brightness is generated.

  Next, the second processing unit 14b translates the approximate linear line calculated by the first processing unit 14a for each of the RGB color elements, thereby plotting by the first processing unit 14a for monochrome. A range including all pairs in the color chart is set, and a plurality of color patches (sixth floor in each of RGB) in the color chart included in the first image so that an approximate cubic curve is included in the set range. An approximate cubic curve using a set of pixel values and ideal values of pixels targeted for a tone patch (that is, a total of 18 patches) is calculated (second sub-step S21).

  Specifically, the second processing unit 14b creates an approximate curve (here, an approximate cubic curve) within a range of 6 points on the gray scale so as not to disturb the gray scale balance. For this purpose, as shown in FIG. 10, first, the second processing unit 14b translates the approximate linear line calculated by the first processing unit 14a to set a range that includes all six gray scale points. (Dotted line in FIG. 10). Then, the second processing unit 14b draws an approximate cubic curve using the least squares method or the like at the remaining 18 points of the color so that the curve is included therein (enclosed by the two dotted lines in FIG. 10). (Approximate cubic curve drawn within the range). 10A to 10C are diagrams illustrating examples of approximate cubic curves calculated by the second processing unit 14b for RGB color elements. Here, of the points plotted, only 6 points on the gray scale are shown. In addition, an approximate linear line and an approximate cubic curve are shown in a range surrounded by two dotted lines in each of FIGS.

  In order to further reduce the error of the approximate curve, it is possible to draw a higher order curve, but a cubic curve is preferable in order to prevent unnaturalness due to a sudden change in value.

  As described above, in the second sub-step S21, a tone curve (approximate cubic curve) is generated that reduces the error (error of the color between the actual image and the photograph) while maintaining the color balance adjusted in the first sub-step S20. The

  Finally, the third processing unit 14c is configured so that the sum of the ratios of whiteout and blackout in the approximate cubic curve for each of the RGB color elements calculated by the second processing unit 14b is minimized. The first tone curve is generated by shifting the approximate cubic curve in the same manner (third sub-step S22).

  When the tone curve starts from 0 or more with a photograph value, or when the correction value is 255 or more, a phenomenon of whiteout or blackout occurs. In such a case, the color information of the bright part and dark part of the photograph is lost, and the photograph cannot be reproduced with an actual color by color correction. In such a case, it is necessary to prevent loss of color information by adjusting the overall brightness while maintaining the color balance. For this purpose, the third processing unit 14c causes the three approximate cubic curves calculated by the second processing unit 14b to rise and fall in the same manner for the three elements R, G, and B, resulting in overexposure and underexposure. The position where the sum of the ratios becomes the smallest is calculated, the approximate cubic curve is shifted to that position, and the obtained approximate cubic curve is generated as the final tone curve (first tone curve).

  (A) to (c) of FIG. 11 are diagrams illustrating examples of processing by the third processing unit 14c for RGB color elements. Here, the three approximate cubic curves are shifted downward so that the sum of the ratios of whiteout and blackout in the three approximate cubic curves calculated by the second processing unit 14b is minimized. The situation is shown.

  By such processing in the third sub-step S22, while maintaining the color balance, the ratio of overexposure and underexposure in the tone curve (approximate cubic curve) is reduced, and the tone curve (the first curve) for appropriately adjusting the brightness. 1 tone curve) is generated.

  FIG. 12 is a diagram showing the analysis result of the image corrected by the color correction apparatus 10 in the present embodiment, and corresponds to FIG. 4 showing the analysis result before correction. That is, (a) to (c) of FIG. 12 show the correspondence between the pixel values of the pixels constituting each patch of the color chart shown in the corrected photograph and the ideal values of the pixels for each RGB color element. FIG. 12 is a plot of (24 points), and (d) to (f) of FIG. 12 are histograms showing the distribution (100 fraction) for each pixel value.

  12A to 12C, an ideal straight line and an approximate linear line of 24 points parallel to the ideal straight line are drawn, as in FIGS. 4A to 4C. However, these ideal straight lines and the approximate primary straight lines almost overlap. From this, it can be seen that the color chart values of the actual and corrected photographs are almost the same.

  Table 2 below shows the analysis results of the photographs before and after the correction in terms of numerical values.

  Table 2 shows three values: "average value of difference between actual and photograph", "difference (actual) between actual and photograph considering overall brightness", and "error seen in the entire photograph". Regarding the evaluation, the values in the photograph before and after the correction are shown. Here, “average value of difference between actual and photograph” is obtained by subtracting the RGB value of the color chart in the photograph from the RGB value of the actual color chart at 24 points, and dividing the total by 24. Is the average. The “difference (error) between the actual image and the photo in consideration of the overall brightness” is a value calculated by the following procedure. That is, first, an approximate linear curve (straight line) is derived at 24 points parallel to a straight line in which the RGB value of the color chart in the photograph is equal to the RGB value of the actual color chart. This is a set of theoretical values considering the overall brightness of the RGB values of the photograph. Then, using this straight line, at each of the 24 points, the RGB value of the color chart of the photograph is taken from the theoretical value, and the total is divided by 24 and averaged. Further, the “error in the entire photograph” is a value calculated by the following procedure. That is, first, at each of the 24 points, the difference between the RGB values of the color chart of the photograph is taken from the theoretical value, similarly to the calculation procedure of “difference (error) between the actual and the photograph considering the overall brightness”. . At the same time, the number of pixels showing the same value as the RGB value of the color chart in the entire photo is counted, and the ratio of the value to the number of pixels in the entire photo is calculated for all 24 colors. These two elements are multiplied and totaled. Since the value obtained in this way is a value of only the pixel showing the same value as the RGB value of the color chart, it is considered that an error in the entire photo is expressed by converting to the entire photo (100%). .

  As can be seen from Table 2, with respect to the “average value of the difference between actual and photograph”, all the color elements are remarkably reduced by the color correction of the present embodiment. Also, “difference (error) between actual and photograph taking the overall brightness into consideration” is also reduced to about one third of red and green by the color correction of the present embodiment. As for blue, there was variation in the points, and not much improvement. Regarding the “error seen in the entire photograph”, since the error of the portion where the ratio of the entire image before the correction is large was originally small, no significant change due to the color correction of the present embodiment was seen in the numerical value.

  Table 3 below shows the balance of colors seen on a gray scale in six levels in the photo after correction, and corresponds to Table 1 showing similar results for the photo before correction.

  In Table 3, as compared with Table 1, the range of values for all colors is remarkably smaller, and the value for white is 0. From this, it can be seen that the color balance is greatly improved by the color correction of the present embodiment.

  In this embodiment, the analysis result for one photo is shown, but other photos are also corrected to colors that are close to the actual color by the color correction of this embodiment, and the color expression is greatly improved. Make sure it will be.

  As described above, according to the present embodiment, it is possible to automatically perform color correction on all color digital images photographed with a color chart. In addition, by photographing under the same photographing conditions, it is possible to simultaneously perform color correction on a plurality of images on which a color chart is not captured, using the same tone curve. In addition, displaying the analysis result based on the color chart may serve as a standard for knowing the characteristics of the color and the error tendency of the image. Furthermore, the brightness can be adjusted while maintaining the color balance so that important information is not lost.

  More specifically, according to the color correction apparatus 10 in the present embodiment, the first processing unit 14a calculates an approximate linear line using the gray scale in the color chart, so that the color balance in the gray scale is adjusted. Color correction that adjusts the overall brightness while trimming is possible. Then, the second processing unit 14b calculates an approximate cubic curve using 18 color points so that the curve falls within a range including all 6 gray scale points. It is possible to perform color correction that reduces the error while maintaining the specified color balance. Furthermore, the approximate cubic curve is shifted by the third processing unit 14c so that the ratio of whiteout and blackout in the approximate cubic curve is reduced while maintaining the color balance in the approximate cubic curve. It is possible to perform color correction that appropriately adjusts brightness while maintaining balance.

  Thereby, color correction is automatically performed logically and objectively with respect to a color digital image obtained by photographing, and an actual color is reproduced with higher accuracy than before.

  As described above, the embodiments of the color correction method, the program, and the apparatus according to the present invention have been described based on the embodiments. However, the present invention is not limited to the embodiments. Unless it deviates from the meaning of the present invention, various modifications made to the embodiments or other forms constructed by combining the components in the embodiments are also included in the present invention.

  For example, in the above embodiment, the color correction apparatus 10 includes the third processing unit 14c that corrects overexposure and blackout, but it is not always necessary to include a processing unit having such a function. Depending on the photograph obtained by shooting, the approximate cubic curve calculated by the second processing unit 14b may not have whiteout and blackout, or the ratio of whiteout and blackout may be negligibly small. Because there is. Therefore, the third processing unit 14c specifies the amount of overexposure and blackout in the approximate cubic curve calculated by the second processing unit 14b. The approximate cubic curve calculated by the second processing unit 14b may be output as it is as the final tone curve for correction without performing the shift process of the approximate cubic curve described in the above.

  In the above-described embodiment, the color correction apparatus 10 is realized as software by a program and a processor that executes the program, but may be realized as hardware by a dedicated electronic circuit such as an LSI.

  The present invention relates to a color correction method, program, and apparatus for color digital images expressed by RGB (red, green, and blue) color elements, and in particular, a color correction method that performs color correction automatically logically and objectively. For example, it can be used as a color correction program for photographs in digital archives that digitize and store information for long-term storage, such as library and art museum originals and official documents. .

DESCRIPTION OF SYMBOLS 10 Color correction apparatus 11 Input part 12 Control part 13 Image acquisition part 14 Tone curve production | generation part 14a 1st process part 14b 2nd process part 14c 3rd process part 15 Correction part 16 Display part 17 Storage part 20 Digital camera

Claims (7)

A color correction method for digital images of colors expressed by RGB (red, green and blue) color elements,
An image acquisition step of acquiring a first image which is a digital image generated by photographing a subject together with a color chart composed of patches of a plurality of color and monochrome gradations;
The color of the first image is corrected for each of the RGB color elements using the pixel value of the pixel constituting the patch of the color chart included in the acquired first image and the ideal value of the pixel. A tone curve generating step for generating a first tone curve used for
A correction step of color correcting the first image using the generated first tone curve,
In the tone curve generation step, for each of the RGB color elements,
A set of pixel values and ideal values of pixels is plotted on two-dimensional coordinates for a plurality of monochrome gradation patches in the color chart included in the first image, and an approximate linear line in the plot is calculated. A first sub-step,
By moving the calculated approximate linear line in parallel, a range including all the sets in the plot is set, and a plurality of colors in the color chart are set so that an approximate cubic curve falls within the set range. A second sub-step of calculating an approximate cubic curve using a set of pixel values and ideal values of pixels for a patch of gray scale;
And a third substep of outputting the calculated approximate cubic curve as the first tone curve. The image acquisition step further acquires at least one second image, which is a digital image generated by shooting a subject under the same shooting conditions as the shooting in the generation of the first image,
In the tone curve generation step, using a color chart included in the first image, a second tone curve used to correct the color of the second image is generated for each of the color elements,
The color correction method according to claim 1, wherein in the correction step, the second image is color-corrected using the generated second tone curve. In the third sub-step, for each of the RGB color elements, the approximate cubic curve is calculated so that the ratio of whiteout and blackout in the approximate cubic curve calculated in the second substep is small. The color correction method according to claim 1, wherein the approximate cubic curve obtained by shifting in the axial direction of the ideal value in the two-dimensional coordinates is generated as the first tone curve. In the third sub-step, the approximate cubic curve for each of the RGB color elements is such that the sum of the ratio of overexposure and blackout in the approximate cubic curve for each of the RGB color elements is minimized. The color correction method according to claim 3, wherein the colors are shifted in the same manner. And a color chart position recognizing step of acquiring a position of the color chart in the first image from a user,
For the pixel values of the pixels constituting the patch of the color chart included in the first image, the position of the color chart in the first image is specified based on the position acquired in the color chart position recognition step, The position of the patch is specified by using the position of the specified color chart and the predetermined positional relationship between the color chart and the patch, and the patch is determined based on the position of the specified patch. The color correction method according to claim 1, wherein a pixel value of a constituent pixel is specified. A program for color correction of a digital image of a color expressed by RGB (red, green and blue) color elements,
The program which makes a computer perform the step of Claim 1. A color correction device that performs color correction on a color digital image expressed by RGB (red, green, and blue) color elements,
An image acquisition unit that acquires a first image that is a digital image generated by photographing a subject together with a color chart composed of patches of a plurality of color and monochrome gradations;
The color of the first image is corrected for each of the RGB color elements using the pixel value of the pixel constituting the patch of the color chart included in the acquired first image and the ideal value of the pixel. A tone curve generator for generating a first tone curve used for
A correction unit that performs color correction on the first image using the generated first tone curve;
The tone curve generation unit, for each of the RGB color elements,
A set of pixel values and ideal values of pixels is plotted on two-dimensional coordinates for a plurality of monochrome gradation patches in the color chart included in the first image, and an approximate linear line in the plot is calculated. A first processing unit,
By moving the calculated approximate linear line in parallel, a range including all the sets in the plot is set, and a plurality of colors in the color chart are set so that an approximate cubic curve falls within the set range. A second processing unit that calculates an approximate cubic curve using a set of pixel values and ideal values of pixels for a patch of the gray scale;
And a third processing unit that outputs the calculated approximate cubic curve as the first tone curve.