JP2012154711A - Optical spectrum estimation device and learning spectrum generation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To efficiently acquire learning spectrum data for optical spectrum estimation.SOLUTION: A learning data generation data retention part 132 retains a spectral reflectance of a substrate and an optical spectrum of multiple monochrome coloring materials with 100% of dot ratio on the substrate. A learning data generation part 122 generates multiple optical spectra to be determined as learning data, using: a first expression regulating a relationship between a spectral reflectance of a coloring material and a tristimulus value, on the premise of a predetermined standard light source and a predetermined color-matching function; a second expression regulating a relationship between the tristimulus value and a dot ratio of the coloring material; a third expression regulating a relationship between a reflectance of a coloring material layer on the substrate and a value of S/K (S: scattering coefficient, K: absorption coefficient) thereof; and a fourth expression regulating a relationship between a value of S/K of a mixed-color coloring material, which is obtained by adding a value of S/K of the substrate to a value of S/K of the monochrome coloring materials on the substrate, and the reflectance thereof.

Description

  The present invention relates to a spectral spectrum estimation apparatus that estimates a spectral spectrum of a printed material and the like, and a learning spectrum generation method that generates learning spectrum data necessary for the estimation.

  In recent years, a spectrum estimation technique has attracted attention in many fields. Spectral spectrum estimation is a technique for estimating spectral spectrum characteristics in increments of several nm from multiband information in the order of tens to hundreds of nm. Spectral spectral characteristics are generally measured using a spectroscope, but by using a spectral spectrum estimation technique, it is possible to measure with a device that is smaller, lighter, and less expensive than the spectroscope.

  There are cameras that can capture spectral images having spectral information for each pixel. For example, a multi-band camera that combines a solid-state image sensor such as a CCD (Charge Coupled Device) image sensor or CMOS (Complementary Metal Oxide Semiconductor) and a variable band filter such as multiple movable filters, liquid crystal tunable filters, and acousto-optic filters. Spectral images can be taken using. However, such a camera is very expensive. On the other hand, an inexpensive apparatus can be substituted by obtaining a spectral image by combining information of an image photographed with a normal color camera detecting with 3 bands or a camera with a band number of about 6 bands with a spectral spectrum estimation technique.

  A spectroscope can only obtain an average value of colors in a partial region, but by using a spectroscopic image, a spectrum in a wide range can be observed and captured as a distribution. Therefore, a method for estimating spectral spectral characteristics from a multiband image is applied to remote sensing, medical care, cosmetics, analytical inspection, foreign matter inspection, sugar content and fat content measurement, color management, and the like.

  Examples of techniques for estimating spectral spectral characteristics from a low-dimensional image include Wiener estimation, multiple regression analysis, Markov estimation, principal component analysis, and the like. Wiener estimation is formulated as an estimation process that minimizes the error between the spectrum estimated from the low-dimensional camera detection value and the spectrum true value, taking into account the statistical properties of the observation target and the characteristics of the observation noise. is there. In order to realize such estimation, learning data is required to generate an estimation matrix used for estimation. The learning data is known data of the low-dimensional camera detection value and spectrum, and is configured as a data set including a large number of data. The learning data may be any object as long as the low-dimensional camera detection value and the spectrum are correct values, but the following is desired in order to improve the estimation accuracy. That is, it is possible to estimate with higher accuracy that the learning data includes a lot of spectral spectra having the same characteristics as the spectral spectral characteristics of the measurement object. For this reason, there has been proposed a method for achieving high accuracy by selecting and estimating learning spectrum data classified by similar features (see, for example, Patent Documents 1 and 2).

  Such technology can be applied as follows. As part of color management, color management using spectral reflectance measurement is performed in color printers and color copiers. Spectral reflectance can be used to measure the absolute value of an object color (ie, the color that appears when an object is illuminated by light). Since the tristimulus values of any light source / field of view can be calculated from the spectral reflectance, the application range is wide and it is optimal for color difference management of printed matter. The color difference between different media can be managed including conditional colors (metamerism).

  However, a spectroscope is expensive to attach to a printing press, and if the measurement wavelength range is wide, it takes time (for example, several minutes) to measure. Moreover, only the color of a limited area of about 5 × 5 mm can be observed. Therefore, we propose a color matching system that predicts the print result so that the desired print result is finished and controls the amount of ink using a method that estimates the spectral spectrum characteristics in steps of several nanometers from the multiband image captured by the camera. Has been. Management using spectra using a low-dimensional camera and spectral estimation is suitable for management using spectral data in such a wide measurement wavelength range.

  The color management of printing using the spectrum is more effective when the special color ink is used as the ink. In process printing, special color inks are often used to reproduce colors that cannot be represented by the basic inks cyan, magenta, yellow, and black (hereinafter referred to as C, M, Y, and K). As for special color inks, ink companies such as DIC Corporation, Toyo Ink Manufacturing Co., Ltd., and PANTONE Corporation have prepared various colors, and there are about 2000 colors in the DIC Color Guide (registered trademark).

  When implementing color management using spectra, if it is normal CMYK printing, the spectral spectrum characteristics in increments of several nanometers can be used to predict printing results using statistics such as dot gain and mixed color space of ink. It can also be reflected and can be predicted with relatively high accuracy. On the other hand, in printing using special color ink, the number of past data is limited, and it takes time and labor to acquire data necessary for prediction. For this reason, the management which detected the spectrum is desirable.

  When using “small” or “not” ink that has been used so far, a large number of learning data is generated by calculation using a small number of measured spectrum data, and an estimation matrix using the learning data, etc. The spectral spectrum may be estimated at By adopting a method of generating a large number of learning data by calculation, it is not necessary to generate and measure a printed matter that is actually printed based on the actually measured spectrum data. Color management using a spectroscopic spectrum is considered the most efficient and desirable at low cost.

JP 2008-304205 A JP 2009-247518 A JP 2010-156612 A

  As described above, a technique for estimating a spectrum using a low-dimensional camera detection value is a very effective technique. In order to perform high-accuracy spectral spectrum estimation, it is necessary to acquire several tens to several hundreds of learning spectral data having the spectral spectral characteristics of the measurement object. It will take a lot of effort.

  The present invention has been made in view of such circumstances, and an object thereof is to provide a technique for efficiently generating learning spectrum data for spectral spectrum estimation by a computational method. That is, an object of the present invention is to easily generate a large amount of such learning data. For example, the present invention provides a technique for easily generating a large amount of learning data that sufficiently reflects the characteristics of the spot color ink as described above.

  A spectral spectrum estimation apparatus according to an aspect of the present invention includes a spectral reflectance of a base material, a holding unit that holds spectral spectra of a plurality of 100% monochrome color materials on the base material, and a predetermined standard light source. Assuming a predetermined color matching function, the first equation defining the relationship between the spectral reflectance of the color material and the tristimulus value, and the second equation defining the relationship between the tristimulus value and the halftone dot rate of the color material, , The third equation defining the relationship between the reflectance of the color material layer on the substrate and the S / K value (S is the scattering coefficient, K is the absorption coefficient), and the S / K value of the substrate and the substrate By using the fourth formula that defines the relationship between the S / K value and the reflectance of the mixed color material obtained by adding the S / K values of a plurality of single color materials, a plurality of learning data should be used. Measured by a learning data generation unit that generates a spectrum, a multiband measurement unit that measures low-dimensional information of the color material to be observed, and a low-dimensional measurement unit Comprising a low-dimensional information of the timber, and a spectrum estimating unit for estimating a spectral reflection spectrum of the color material from the generated learning data by the learning data generating unit. The learning data generation unit calculates the spectral reflectance of the mixed color material obtained by superimposing the plurality of 100% halftone dot color materials held in the holding unit from the third and fourth equations. Set the spectral reflectance of single color material and mixed color material, set multiple types of halftone dot ratio of single color material, calculate multiple tristimulus values, and set the multiple tristimulus values in the first formula Thus, a plurality of spectral reflectances to be used as learning data are calculated.

  Another embodiment of the present invention is also a spectral spectrum estimation apparatus. This device is premised on a spectral spectrum of a base material, a holding unit that holds a spectral spectrum of a substance having a content of 100% on the base material, a predetermined standard light source, and a predetermined color matching function. The first equation defining the relationship between the tristimulus value and the tristimulus value, the second equation defining the relationship between the tristimulus value and the halftone dot ratio of the color material, the reflectance of the color material layer on the substrate, and S / Add the third equation that defines the relationship with the K value (S is the scattering coefficient, K is the absorption coefficient), the S / K value of the base material, and the S / K values of a plurality of single color materials on the base material. A learning data generation unit for generating a plurality of spectral spectra to be used as learning data by using the fourth formula that defines the relationship between the S / K value and the reflectance of the mixed color material obtained in the above; A low-dimensional measurement unit that measures low-dimensional information of the substance to be measured, low-dimensional information of the substance measured by the low-dimensional measurement unit, and a learning data generation unit And a spectrum estimation unit that estimates a spectrum of the made learning data material. A substance is a substance that has a specific absorption wavelength or emission wavelength, and the peak position of the specific absorption wavelength or emission wavelength does not change with the change in the content of the substance. Calculates a spectral spectrum in a state in which a plurality of substances with a content of 100% held in the holding unit are superimposed from the third and fourth formulas, and the second formula includes the non-overlapping substance and the overlapping substance. Set a spectral spectrum, set multiple types of non-overlapping substance content, calculate a plurality of tristimulus values, set the plurality of tristimulus values in the first equation, A spectrum is calculated.

  Yet another embodiment of the present invention is a learning spectrum generation method. This method presupposes a predetermined standard light source and a predetermined color matching function, defines the relationship between the spectral reflectance of the color material and the tristimulus value, the tristimulus value and the dot rate of the color material, The second equation that defines the relationship of the above, the third equation that defines the relationship between the reflectance of the color material layer on the substrate and the S / K value (S is the scattering coefficient, K is the absorption coefficient), Using the fourth formula that defines the relationship between the S / K value and the reflectance of the mixed color material obtained by adding the S / K value and the S / K values of a plurality of single color materials on the substrate. And a learning data generation step of generating a plurality of spectral spectra to be learned data. The learning data generation step includes a step of acquiring the spectral reflectance of the base material and the spectral reflectance of a single color material having a halftone dot rate of 100% on the base material, and a plurality of meshes from the third and fourth formulas. Calculating a spectral reflection spectrum of a mixed color material in which a single color material with a dot rate of 100% is superimposed, and setting the spectral reflectance of the single color material and the mixed color material in the second formula, Setting a plurality of types of rates, calculating a plurality of tristimulus values, and setting the plurality of tristimulus values in the first equation to calculate a plurality of spectral reflectances to be learned data.

  Yet another embodiment of the present invention is also a learning spectrum generation method. This method presupposes a predetermined standard light source and a predetermined color matching function, defines the relationship between the spectral reflectance of the color material and the tristimulus value, the tristimulus value and the dot rate of the color material, The second equation that defines the relationship of the above, the third equation that defines the relationship between the reflectance of the color material layer on the substrate and the S / K value (S is the scattering coefficient, K is the absorption coefficient), The fourth formula that defines the relationship between the S / K value and the reflectance of the mixed color material obtained by adding the S / K value and the S / K values of a plurality of single color materials on the substrate is diverted. Then, a learning data generation step for generating a plurality of spectral spectra to be used as learning data is provided. A substance is a substance that has a specific absorption wavelength or emission wavelength, and the peak position of the specific absorption wavelength or emission wavelength does not change with the change in the content of the substance. In the step, a spectral spectrum of the base material, a spectral spectrum of a substance having a plurality of contents of 100% on the base material, and a substance having a plurality of contents of 100% are superimposed from the third and fourth formulas. Calculating the spectrum of the non-overlapping substance and the overlapping substance in the second equation, setting a plurality of types of non-overlapping substance content, and setting a plurality of tristimulus values. Calculating and setting a plurality of tristimulus values in the first equation to calculate a plurality of spectral spectra to be used as learning data.

  Note that any combination of the above-described components and a representation of the present invention converted between a method, an apparatus, a system, a recording medium, a computer program, and the like are also effective as an aspect of the present invention.

  According to the present invention, learning spectrum data for spectral spectrum estimation can be efficiently acquired.

It is a figure which shows an example of the printed matter P by which the halftone printing was carried out by CMY3 color. Is a diagram showing the spectrum of D ₆₅ standard light source. It is a figure which shows CIE (1964) X10Y10Z10Color Matching Function. It is a figure for demonstrating the overlapping pattern of the halftone dot in CMY. It is a figure for demonstrating Kubelka-Munk theory. It is a figure which shows the figure which actually performed the simulation by making into an example the case where a total of 5 colors which added CMYK and special color ink DIC589 were used. Using the spectral reflectance spectrum of CMYKS 100% shown in FIG. 6, all combinations (35 = 243) when the halftone ratios are 0.3, 0.6, and 0.9, respectively. It is a figure which shows the calculated result. FIG. 8A is a diagram showing measured data and estimated data of the spectral reflection spectrum of a printed matter printed with a dot ratio of 100% of the special color ink DIC589. FIG. 8B is a diagram showing measured data and estimated data of the spectral reflection spectrum of a printed material that is overprinted with a dot ratio of 40% for cyan ink C and a dot ratio of 100% for spot color ink DIC589. It is a figure which shows the structure of the spectral-spectrum estimation apparatus which concerns on embodiment of this invention. It is a figure which shows the structure of the spectral-spectrum estimation system which concerns on the modification of embodiment of this invention. It is a flowchart for demonstrating the operation example of the spectral-spectrum estimation apparatus which concerns on embodiment of this invention.

  FIG. 1 is a diagram illustrating an example of a printed material P that is dot-printed in three colors of CMY. FIG. 1 shows an example of process printing by CMY. The enlarged view on the right side of FIG. 1 is an enlarged view of a part of the drawing area D of the printed matter P.

  In the embodiment of the present invention, the printed matter P halftone-printed in CMY three colors is photographed with a low-dimensional camera, and the spectral reflection spectrum is estimated. A method for generating learning data necessary for spectral reflection spectrum estimation will be described below. CMY is an example of colors, and is not limited to these colors.

Figure 2 is a diagram showing the spectrum of D ₆₅ standard light source. The horizontal axis of the graph shown in FIG. 2 indicates wavelength [nm], and the vertical axis indicates radiant energy. FIG. 3 is a diagram showing the CIE (1964) X ₁₀ Y ₁₀ Z ₁₀ Color Matching Function. In the graph shown in FIG. 3, the horizontal axis represents wavelength [nm], and the vertical axis represents sensitivity.

The following (Expression 1) is an expression that prescribes a relationship between the spectral reflection spectrum of the colorant and the tristimulus values on the premise of a predetermined standard light source and a predetermined color matching function. Here, the light source is a _D65 standard light source (see FIG. 2), and the color matching function is CIE (1964) X ₁₀ Y ₁₀ Z ₁₀ Color Matching Function (see FIG. 3).

  Spectral reflectance Reflectance (λ) in the visible light region of the base material (paper in this embodiment), and the visible light region of each color of CMY printed on the base material with a dot ratio of 100% (that is, solid printing) The spectral reflectance Reflectance (λ) at is measured. For example, it is measured with a spectroscope. Then, the spectral reflectance Reflectance (λ) of each color is substituted into the following (Equation 1). Thereby, the tristimulus value XYZ of each color of CMY printed on the base material and the base material at a dot ratio of 100% can be calculated.

  Next, the Neugebauer equation will be described. The Neugebauer equation is a linear equation that expresses the relationship between the tristimulus value and the halftone dot area ratio (hereinafter simply referred to as the halftone dot ratio) of the color material. An example is given below.

FIG. 4 is a diagram for explaining a CMY overlapping pattern that can occur in a printed dot product. There are 8 overlapping patterns shown by 0 to 7 below. In the dot print shown in FIG.
“0” is a piece of paper (white (hereinafter referred to as W)) only “1” is a portion of C only “2” is a portion of M only “3” is a portion of Y only “4” is C and M "5" where C and Y overlap, "6" where C and Y overlap, and "7" where M and Y overlap each other, where C, M and Y overlap. Each part 0-7 has a different tristimulus value. The tristimulus values of each region are shown by paying attention to the region where such dots are gathered.

  Next, consider the tristimulus values of the printed material. Let us consider the tristimulus values of the region of interest. This area is composed of a large number of dots, and the number of dots is determined by the dot ratio of each of the C, M, and Y inks. The tristimulus value of the region of interest can be obtained by weighting and adding the tristimulus values of the overlapping portions of the respective inks with the area ratio in the region of interest. The area ratio of the overlapping portions of the inks in the region of interest can be calculated from the dot ratios c, m, and y of C, M, and Y.

Here, the left sides X, Y, and Z are tristimulus values of the target object. Xx, Yx, and Zx on the right side are tristimulus values in the x portion, and the lowercase letter x indicates the overlapping region described above. For example, Xw indicates the X stimulation value of the W (paper) area, and similarly, Xc, Xm, Xy, Xcm, Xmy, Xcy, and Xcmy are the C, M, Y, CM, MY, CY, and CMY areas. The X stimulation value is shown. C, m, and y indicate halftone dot ratios.

  The tristimulus values XYZ of the secondary and tertiary colors “4” to “7” where the inks are overlapped may be obtained from actual measurement by performing overprinting of the CMY primary color 100%, but the CMY primary color 100% Using the Kubelka-Munk theory and the color mixing theory, the following (Equation 3) and (Equation 4) can be used.

  FIG. 5 is a diagram for explaining the Kubelka-Munk theory. As shown in FIG. 5, a uniform color material layer 12 having a certain scattering coefficient S and absorption coefficient K and having a thickness X is placed in close contact (optical contact) on a base 11 having reflectance Rg. In this case, the reflectance R is calculated. That is, the following (Equation 3) is an equation that defines the relationship between the reflectance of the color material layer on the substrate and the S / K value (S is a scattering coefficient, K is an absorption coefficient). Is an equation that defines the relationship between the S / K value and the reflectance of a color mixing material obtained by adding the S / K value of the substrate and the S / K values of a plurality of single color materials on the substrate. It is.

Here, the subscripts a, b, and w indicate the ink a, the ink b, and the paper that are printed with a halftone dot ratio of 100%, respectively, and mix indicates a mixed color. The mixed color can be formed by superposition printing shown in FIG.

Once R _{mix at} each wavelength is obtained, the tristimulus values XYZ of “4” to “7” can be obtained by substituting them into Reflectance (λ) in the above (Equation 1). Furthermore, the tristimulus value XYZ of a printed matter printed at a certain dot ratio can be obtained from the above (Equation 2), and the spectral reflection spectrum can be obtained again using the above (Equation 1).

  According to this method, the spectral reflection spectrum needs to be actually measured only for a base material portion such as paper and a printed portion having a halftone dot rate of 100% using the ink used. %, The spectral reflectance spectrum when the halftone dot ratio is changed can be obtained by calculation.

  In the embodiment of the present invention, the print result is not predicted from such a dot ratio, but is measured with a low-dimensional camera without measuring the spectral spectrum, and the spectral spectrum is obtained by using the estimation method described later. Estimated by calculation. At this time, the generation of several tens to several hundreds of learning spectrum data necessary for spectral spectrum estimation is generated by changing the dot ratio of the above (Equation 2).

  Management of printing using spectra is effective and useful for all color prints, but is particularly useful when spot inks are used. In the following embodiment, a case where spot color ink is used will be described. When process printing is performed using a special color ink or the like, in particular, in the case of a special color ink including fluorescence, many of them have a characteristic spectral shape that CMYK inks do not have. Therefore, it is difficult to accurately estimate the spectral reflection spectrum with the learning data for CMYK printing. Therefore, in order to perform highly accurate estimation, learning spectrum data having the characteristics of special color ink is necessary.

  For example, it takes time and labor to print and measure with a plurality of halftone dots even with a total of five colors including one special color ink added to CMYK. Therefore, it is required to generate learning spectrum data from minimum measured spectrum data. Here, the minimum data is actual measurement data of the spectral reflection spectrum of a printed matter printed at a dot ratio of 100% for all paper and ink used. Based on these actually measured data, the spectral reflection spectrum when the halftone dot ratio is changed in the above (Formula 2) is obtained using the above (Formula 1) to (Formula 4) to generate learning spectrum data. However, in the above (Equation 2), the matrix calculation is increased by the combination of the number of inks used. In this way, if the spectral data of the printed matter can be estimated with high accuracy only by measuring several bands, color management using the spectral spectrum can be realized at low cost.

  FIG. 6 is a diagram in which a simulation is actually performed by using as an example a case where a total of five colors including CMYK and special color ink DIC589 are used. The horizontal axis of the graph shown in FIG. 6 indicates the wavelength [nm], and the vertical axis indicates the reflectance [%]. In FIG. 6, the spectral reflection spectrum of the special color ink DIC589 is denoted by S. FIG. 6 shows a spectral reflection spectrum of CMYKS with a dot ratio of 100% (solid printing).

7 uses the spectral reflection spectrum of CMYKS shown in FIG. 6 with a dot ratio of 100%, and the combinations when the dot ratios are 0.3, 0.6, and 0.9 are all (3 ⁵ = 243)) It is a figure which shows the calculation result. That is, 243 kinds of learning spectrum data are generated. Note that the combination of halftone dot ratios is not limited to 0.3, 0.6, and 0.9. For example, it may be 0.25, 0.5, or 0.75. Also, the types of halftone dots are not limited to three, and the types may be further increased.

  Here, a method for estimating a spectral reflection spectrum of each point of an object (for example, a printed material) from a low-dimensional image (for example, a 3-band image or a 6-band image) will be described. As the estimation method, Wiener estimation, multiple regression analysis, Markov estimation, principal component analysis, and the like can be used. An example using Wiener estimation will be described below.

When an object is photographed through a wide band filter of a plurality of bands with a low-dimensional camera, the spectral distribution of light incident on an image sensor (for example, a CCD element) corresponding to the coordinates (x, y) of the image is t (λ) E. (Λ) r (x, y; λ). Here, t (λ) is the spectral transmittance of the i-th filter, E (λ) is the spectral radiance of illumination, and r (x, y; λ) is the image coordinates (ie, pixel position) (x , Y) shows the spectral reflectance of the object. Also, S (λ) is the total spectral performance that combines the spectral transmittance of the lens and the spectral sensitivity of the image sensor. At this time, the sensor response v _i (x, y) obtained in each element is given by integrating the incident light tE (λ) r (x, y; λ) and the spectral (λ) in the wavelength region. Therefore, it is represented by the following (formula 5).

Here, m represents the number of bands of the low-dimensional camera. The spectral product S (λ) was assumed to be zero outside the visible light range of wavelength 400 to 700 nm. When m is set to a large number, it may be called a multiband camera. There is no numerical value that defines the number of multibands for m or more, and there are widths depending on the method and application for realizing multibands, such as 9 bands or more or 1024 bands or more. In this embodiment, the case where m is a single-digit numerical value is called a low-dimensional camera.

Next, in order to simplify mathematical handling, the spectral distribution is discretized and expressed using vectors and matrices. When v is a row vector having m elements representing sensor responses of m bands and r is a row vector composed of l elements representing the spectral reflectance of the object, the above (formula 5) is It is expressed using a vector matrix as shown below (Formula 6).

Here, the coordinates (x, y) of the image are omitted. The matrix F is a matrix T (see (Equation 7) below) in which row vectors t representing the spectral transmittance of the i-th filter are combined, and an l × l diagonal matrix corresponding to the spectral sensitivity of the illumination and the camera. Using the matrices E and S, it is defined as (Equation 8) below.

  The Wiener estimation is a technique for minimizing the mean square error E between the spectral reflectance r of a plurality of learning data samples and the estimated spectral reflectance (r tilde). expressed.

<> Represents an ensemble average for a spectral reflectance sample.

From the following (Equation 10), an estimation matrix G for estimating the spectral reflectance from the sensor response vector is considered.

At this time, an estimation matrix that minimizes the mean square error E expressed by the above (formula 9) is given by the following (formula 11).
Here, R _rv (see (Expression 12) below) represents the cross-correlation matrix of r and v related to the sample, and R _vv (see (Expression 13) below) represents the autocorrelation matrix of v. R _rr (see (Expression 14) below) represents an autocorrelation matrix of r.

Therefore, the above (formula 11) can be rewritten as the following (formula 15).

When the sensor response includes noise n, the above (formula 6) becomes the following (formula 16).

The autocorrelation matrix R _nn of noise n is expressed as (Equation 17) below.

Thus, the Wiener estimation gives an estimation matrix that minimizes the mean square error of the estimated value by a simple linear operation when the statistics of the signal and noise are known. Here, if the input spectrum and the noise spectrum are uncorrelated, an estimation matrix for minimizing the mean square error is given by (Equation 18) below.

As a result, when the estimation matrix G is obtained, the spectral reflectance of the object in the detection value (x, y) of the image obtained by photographing the measurement object with the low-dimensional camera is set as v, and the estimated spectral reflection is calculated from the above equation (10). The rate (r tilde) can be determined. For example, when v is an RGB value (3 bands) and spectral reflectance data is obtained every 2 nm from 400 nm to 700 nm, it is expressed as (Equation 19) below.

  FIG. 8A is a diagram showing measured data and estimated data of the spectral reflection spectrum of a printed matter printed with a dot ratio of 100% of the special color ink DIC589. In FIG. 8A, the spectral reflection spectrum A indicates actually measured data. The spectral reflection spectrum E1 indicates estimation data using the learning data shown in FIG. 7, and the spectral reflection spectrum E2 indicates estimation data created using JIS ink data (not including spot color ink).

  FIG. 8B is a diagram showing measured data and estimated data of the spectral reflection spectrum of a printed material that is overprinted with a dot ratio of 40% for cyan ink C and a dot ratio of 100% for spot color ink DIC589. In FIG. 8B, the spectral reflection spectrum A indicates measured data. The spectral reflection spectrum E1 indicates estimation data using the learning data shown in FIG. 7, and the spectral reflection spectrum E2 indicates estimation data created using JIS ink data (not including spot color ink).

  8A and 8B, it can be seen that the estimated data using the learning data in FIG. 7 has a higher degree of coincidence with the actually measured data than the estimated data using JIS ink data.

  FIG. 9 is a diagram showing a configuration of the spectral spectrum estimation apparatus 100 according to the embodiment of the present invention. The spectral spectrum estimation apparatus 100 according to the embodiment includes a low-dimensional measurement unit 110, a control unit 120, a storage unit 130, an input unit 140, and an output unit 150.

  The low-dimensional measurement unit 110 acquires low-dimensional information of the measurement object. For the low-dimensional measurement unit 110, for example, a low-dimensional camera can be employed. The low-dimensional measuring unit 110 may be other than a camera as long as it can measure low-dimensional information. Here, low-dimensional information refers to the reflectance in several arbitrary wavelength regions. In the following description, it is assumed that low-dimensional information is acquired by irradiating a measurement target with a certain light source (for example, a white light source) and that there is no influence of external light.

  The low-dimensional measurement unit 110 includes an image sensor 111. As the image sensor 111, a CCD image sensor or a CMOS image sensor can be used. An image sensor 111 mounted on a general camera is equipped with a three primary color filter, and the image sensor 111 outputs three primary color signals of red, green, and blue (hereinafter referred to as R, G, and B). To do. That is, RGB 3-band information is output.

  In addition, by attaching any other color filter to the image sensor 111, the image sensor 111 can output luminance information in any other wavelength region. For example, two types of filters having different spectral sensitivity characteristics may be used. That is, by using an RGB filter and an R′G′B ′ filter having spectral sensitivity characteristics that interpolate the dead band of the RGB filter, the image sensor 111 can output a total of 6 band information. If any image information is not required, a method of irradiating a light source having an arbitrary wavelength without using the image sensor 111 and measuring the reflectance with a photodiode (not shown) may be used.

  In the following description, it is assumed that the low-dimensional information measured by the low-dimensional measuring unit 110 is a low-dimensional image. The low-dimensional image acquired by the low-dimensional measurement unit 110 is supplied to the control unit 120. The control unit 120 controls the entire spectral spectrum estimation apparatus 100 in an integrated manner. More specifically, the control unit 120 includes a trimming & pixel value averaging processing unit 121, a learning data generation unit 122, and a spectral spectrum estimation unit 123.

  These configurations can be realized by an arbitrary processor, memory, or other LSI in terms of hardware, and can be realized by a program loaded in the memory in terms of software, but here by their cooperation. Draw functional blocks. Therefore, those skilled in the art will understand that these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof.

  The storage unit 130 includes a measurement low-dimensional information holding unit 131, a learning data generating data holding unit 132, a learning spectrum data holding unit 133, a light source spectrum characteristic holding unit 134, and a camera spectral sensitivity characteristic holding unit 135.

  The measurement low-dimensional information holding unit 131 temporarily holds low-dimensional information measured by the low-dimensional measurement unit 110. The learning data generation data holding unit 132 is the minimum measured spectrum data necessary for generating learning data (in this embodiment, halftone dots of the base material (paper) and the color materials (C, M, Y, K)). 100% spectral reflectance spectrum data) is retained. The measured spectrum data is measured in advance by a spectroscope or the like, for example.

  The light source spectrum characteristic holding unit 134 holds light source spectrum data necessary for spectral reflection spectrum estimation. The camera spectral sensitivity characteristic holding unit 135 holds the spectral sensitivity characteristic of the camera necessary for spectral reflection spectrum estimation. The learning spectrum data holding unit 133 holds the learning spectrum data generated by the learning data generation unit 122.

  The trimming & pixel value averaging processing unit 121 averages the pixel values in the region in which the measurement object is captured for each of the multiband images acquired by the low-dimensional measurement unit 110. At that time, the region is trimmed. That is, the pixel values in the range inside several pixels from the boundary of the region are averaged. Since the boundary portion is likely to cause color mixing due to printing misalignment or ink bleeding, the pixel values in the range inside the boundary may be averaged. Note that a median value may be used instead of the average value. For example, when an RGB three-band image is supplied from the low-dimensional measurement unit 110, three values to be the basis of spectral reflection spectrum estimation are generated.

  The learning data generation unit 122 generates a plurality of spectral spectra to be used as learning data using the above (Expression 1) to (Expression 4). The learning data generation unit 122 calculates the spectral reflectance of the mixed color material obtained by superimposing a plurality of 100% halftone color materials from the above (Expression 3) and (Expression 4). Then, the spectral reflectances of the single color material and the mixed color material are set in the above (Equation 2), a plurality of types of halftone dot ratios of the single color material are set, and a plurality of tristimulus values are calculated. In the example described above, three types of 0.3, 0.6, and 0.9 are set as the halftone dot ratio of the single color material. These dot ratios are values set from the input unit 140 due to user settings. The learning data generation unit 122 sets a plurality of tristimulus values calculated by the above (Equation 2) in the above (Equation 1), calculates a plurality of spectral reflectances to be used as learning data, and a learning spectrum data holding unit 133. To store.

  The spectral spectrum estimation unit 123 includes a multiband image of a color material to be observed held in the measurement low-dimensional information holding unit 131, and a learning spectrum generated by the learning data generation unit 122 and held by the learning spectrum data holding unit 133. The spectral reflection spectrum of the colorant is estimated from the data. The above-described Wiener estimation, multiple regression analysis, Markov estimation, principal component analysis, and the like can be used. The output unit 150 outputs the spectral reflection spectrum estimated by the spectral spectrum estimation unit 123. For example, it is displayed on a display unit (not shown) or transmitted to another device (not shown).

  FIG. 10 is a diagram showing a configuration of a spectral spectrum estimation system 300 according to a modification of the embodiment of the present invention. The spectral spectrum estimation system 300 includes a spectral spectrum estimation device 100 and an external calculation device 200. This modification is an example in which the above-described learning spectrum data generation processing is performed by the external arithmetic device 200 (for example, a PC or a server). The spectral spectrum estimation apparatus 100 and the external calculation apparatus 200 can exchange data via a network (for example, the Internet or a dedicated line).

  Compared with the spectral spectrum estimation apparatus 100 shown in FIG. 9, the spectral spectrum estimation apparatus 100 according to the present modification has a configuration in which the learning data generation data holding unit 132 and the learning data generation unit 122 are removed. In this modification, the external arithmetic device 200 includes a learning data generation data holding unit 232 and a learning data generation unit 222 corresponding to them. The learning data generation unit 222 of the external arithmetic device 200 calculates the learning spectrum data and transmits it to the learning spectrum data holding unit 133 of the spectral spectrum estimation device 100.

  FIG. 11 is a flowchart for explaining an operation example of the spectral spectrum estimation apparatus 100 according to the embodiment of the present invention. FIG. 11 illustrates an example of estimating the spectral reflection spectrum of a printed material with three colors of CMY. First, using a spectroscope or the like, the spectral reflectance of the base material and CMY single color with a dot ratio of 100% is measured (S10). The learning data generation unit 122 calculates the spectral reflectance of the overlapping portion of CMY from the Kubelka-Munk formula (S12). Next, a plurality of tristimulus values XYZ are calculated from the Neugebauer equation using the dot ratio of CMY as a parameter (S14). Next, a spectral spectrum is calculated from the plurality of tristimulus values XYZ from the above (Equation 1) and used as learning data (S16).

  The spectral spectrum estimation unit 123 calculates an estimation matrix from the Wiener estimation (S18). Finally, the spectral reflectance is estimated by calculation from the multiband measurement values and the estimation matrix (S20).

  As described above, according to the present embodiment, by applying a measured value of the minimum spectral spectrum to the Neugebauer equation and changing the dot ratio of the Neugebauer equation to perform a spectral spectrum simulation, Learning spectrum data for spectrum estimation can be generated efficiently. In addition, highly accurate learning spectrum data having the characteristics of the spectrum of the measurement object can be generated. And the spectral spectrum of a measurement object can be estimated with high precision using this learning spectrum data.

  Generally, the learning spectrum data needs several tens to several hundreds, but according to the present embodiment, it is sufficient to measure only the spectral spectrum of the base material and the monochrome color material having a dot ratio of 100%. Since the remainder can be generated by the above-described calculation, time and labor can be greatly reduced.

  The present invention has been described based on some embodiments. It is understood by those skilled in the art that these embodiments are exemplifications, and that various modifications can be made to combinations of the respective constituent elements and processing processes, and such modifications are also within the scope of the present invention. By the way.

  The learning spectrum data generation method according to the above-described embodiment can be applied to other than printed matter. That is, the present invention can be applied to all substances having a certain characteristic peak, the spectral change is only the luminance change, and the peak wavelength is not shifted. More specifically, a substance having a specific absorption wavelength or emission wavelength, which can be widely applied as long as the peak position at the specific wavelength does not change and only the peak luminance increases or decreases according to the change in the content. It is.

  For example, in a pathological diagnosis by staining, the content of a certain substance is measured from a fluorescence spectrum observed with a microscope. This is a technique in which the content (that is, the concentration) of the substance corresponds to the luminance at a specific peak wavelength in the fluorescence spectrum, and the concentration is obtained from the luminance. Specifically, a spectrum with a density of 100% may be applied instead of a spectrum with an ink amount of 100% (that is, a halftone dot ratio of 100%). However, since this method is only related to a method for generating learning spectrum data having spectral features, for example, it corresponds to a spectral spectrum when the estimation result set to 60% concentration is substantially 60% concentration. Do not mean.

  It can also be applied to freshness management of meat. The concentration of substances contained in meat changes with time.

  DESCRIPTION OF SYMBOLS 100 Spectral spectrum estimation apparatus, 110 Low-dimensional measurement part, 111 Image sensor, 120 Control part, 121 Trimming & pixel value averaging process part, 122 Learning data generation part, 123 Spectral spectrum estimation part, 130 Storage part, 131 Measurement low-dimensional information Holding unit, 132 learning data generation data holding unit, 133 learning spectrum data holding unit, 134 light source spectrum characteristic holding unit, 135 camera spectral sensitivity characteristic holding unit, 140 input unit, 150 output unit, 200 external computing device, 232 learning data Data generation unit for generation, 222 Learning data generation unit, 240 input unit, 300 spectral spectrum estimation system.

Claims (5)

A holding unit for holding the spectral reflectance of the base material and the spectral spectrum of a single color material having a plurality of dot ratios of 100% on the base material;
Based on the premise of a predetermined standard light source and a predetermined color matching function, the first equation that defines the relationship between the spectral reflectance of the color material and the tristimulus value and the relationship between the tristimulus value and the halftone dot rate of the color material are defined. The second equation, the third equation defining the relationship between the reflectance of the color material layer on the substrate and the S / K value (S is the scattering coefficient, K is the absorption coefficient), and the S / K value of the substrate Learning data using the fourth formula that defines the relationship between the S / K value and the reflectance of the color mixing material obtained by adding the S / K values of a plurality of single color materials on the substrate. A learning data generation unit for generating a plurality of spectral spectra to be
A low-dimensional measuring unit that measures low-dimensional information of the colorant to be observed;
Low-dimensional information of the color material measured by the low-dimensional measurement unit, and a spectral spectrum estimation unit that estimates a spectral reflection spectrum of the color material from learning data generated by the learning data generation unit,
The learning data generation unit calculates a spectral reflectance of a mixed color material in which a plurality of 100% halftone dot color materials held in the holding unit are overlapped from the third and fourth formulas, Spectral reflectances of the single color material and mixed color material are set in the second equation, a plurality of dot ratios of the single color material are set, a plurality of tristimulus values are calculated, and the plurality of tristimulus values are calculated. A spectral spectrum estimation apparatus characterized in that a plurality of spectral reflectances to be used as the learning data are calculated by setting a value in the first equation. A holding part for holding the spectral spectrum of the substrate and the spectral spectrum of the substance having a content of 100% on the substrate;
Based on the premise of a predetermined standard light source and a predetermined color matching function, the first equation that defines the relationship between the spectral reflectance of the color material and the tristimulus value and the relationship between the tristimulus value and the halftone dot rate of the color material are defined. The second equation, the third equation defining the relationship between the reflectance of the color material layer on the substrate and the S / K value (S is the scattering coefficient, K is the absorption coefficient), and the S / K value of the substrate Using the fourth formula that defines the relationship between the S / K value and the reflectance of the mixed color material obtained by adding the S / K values of a plurality of single color materials on the base material and learning A learning data generation unit for generating a plurality of spectral spectra to be data;
A low-dimensional measuring unit that measures low-dimensional information of the substance to be observed;
Low-dimensional information of the substance measured by the low-dimensional measurement unit, and a spectral spectrum estimation unit that estimates a spectral spectrum of the substance from learning data generated by the learning data generation unit,
The substance is a substance having a specific absorption wavelength or emission wavelength, and the peak luminance changes without changing the peak position of the specific absorption wavelength or emission wavelength according to the change in the content thereof,
The learning data generation unit calculates a spectral spectrum in a state in which a plurality of substances having a content rate of 100% held in the holding unit are overlapped from the third formula and the fourth formula, Set the spectral spectrum of the overlapping material and the overlapping material, set multiple types of content of the non-overlapping material, calculate a plurality of tristimulus values, and set the plurality of tristimulus values in the first equation A spectral spectrum estimation apparatus that calculates a plurality of spectral spectra to be set and used as the learning data.   The spectral spectrum estimation apparatus according to claim 1 or 2, wherein the second equation is a Neugebauer equation. Based on the premise of a predetermined standard light source and a predetermined color matching function, the first equation that defines the relationship between the spectral reflectance of the color material and the tristimulus value and the relationship between the tristimulus value and the halftone dot rate of the color material are defined. The second equation, the third equation defining the relationship between the reflectance of the color material layer on the substrate and the S / K value (S is the scattering coefficient, K is the absorption coefficient), and the S / K value of the substrate Learning data using the fourth formula that defines the relationship between the S / K value and the reflectance of the color mixing material obtained by adding the S / K values of a plurality of single color materials on the substrate. A learning data generation step for generating a plurality of spectral spectra to be
The learning data generation step includes
Obtaining the spectral reflectance of the base material and the spectral reflectance of a single color material having a plurality of dot ratios of 100% on the base material;
Calculating a spectral reflection spectrum of a mixed color material obtained by superimposing the plurality of 100% monochrome color materials from the third and fourth formulas;
Spectral reflectances of the single color material and mixed color material are set in the second equation, a plurality of dot ratios of the single color material are set, a plurality of tristimulus values are calculated, and the plurality of tristimulus values are calculated. And a step of calculating a plurality of spectral reflectances to be used as the learning data by setting a value to the first equation. Based on the premise of a predetermined standard light source and a predetermined color matching function, the first equation that defines the relationship between the spectral reflectance of the colorant and the tristimulus value and the relationship between the tristimulus value and the halftone dot rate of the colorant The second equation, the third equation defining the relationship between the reflectance of the color material layer on the substrate and the S / K value (S is the scattering coefficient, K is the absorption coefficient), and the S / K value of the substrate Using the fourth formula that defines the relationship between the S / K value and the reflectance of the mixed color material obtained by adding the S / K values of a plurality of single color materials on the base material and learning A learning data generation step for generating a plurality of spectral spectra to be data;
The substance to be measured is a substance having a specific absorption wavelength or emission wavelength, and the peak luminance changes without changing the peak position of the specific absorption wavelength or emission wavelength according to the change in the content thereof,
The learning data generation step includes
Obtaining a spectral spectrum of the substrate and a plurality of 100% content substances on the substrate;
Calculating a spectroscopic spectrum in a state where the plurality of substances having a content of 100% are overlapped from the third formula and the fourth formula;
A spectral spectrum of a non-overlapping substance and a superposed substance is set in the second equation, a plurality of types of contents of the non-overlapping substance are set, a plurality of tristimulus values are calculated, and the plurality of tristimulus values are calculated. And a step of calculating a plurality of spectral spectra to be used as the learning data by setting to the first equation.