JP2019028814A - Estimation device, estimation method, and program - Google Patents

Abstract

To provide an estimation device, an estimation method, and a program achieving an improvement in estimation accuracy upon estimating, from a plurality of filter images, the filter image with a wavelength section more subdivided than the plurality of filter images through estimation of a spectral reflectance by use of a basic function.SOLUTION: An estimation device 10 acquires, from a plurality of filter images obtained by photographing a predetermined subject by a photographing feature amount acquisition unit 111, a photographing feature amount of each of the plurality of filter images, and estimates, from the plurality of filter images based on the photographing feature amount acquired by an estimation unit 130, the filter image with a wavelength section more subdivided than the plurality of filter images.SELECTED DRAWING: Figure 2

Description

  The present invention relates to an estimation device, an estimation method, and a program.

  In recent years, as a spectroscopic technique that has made remarkable progress, there is a technique for acquiring / analyzing hyperspectral information. Here, the hyper spectrum means several hundred spectrum bands (spectral bands). Then, the spectral information of the object is acquired using several hundred spectral bands (generally, 100 or more) at a time.

  Further, hyperspectral imaging (hyperspectral camera) is known as an imaging device. Hyperspectral cameras have been used mainly in the field of remote sensing. However, in recent years, research in the medical field has also been conducted, and applications to tissue evaluation and pathological diagnosis combined with staining techniques have been performed.

  There are several types of hyperspectral cameras. For example, in the method using a line sensor, a horizontal line for a measurement visual field is vertically divided, and light passing through a slit is separated according to a wavelength by a spectroscopic element such as a diffraction grating or a prism. . The separated light for each wavelength is detected by a two-dimensional image sensor. However, this method can acquire only one line of spectral information at a time. Therefore, it is necessary to perform shooting while moving the camera or the object in the vertical direction.

  In the system using Fabry-Perot, light is introduced into a plane parallel plate and only light of a specific wavelength is extracted by multiple reflection and interference between the plane plates constituting the plane parallel plate. Specifically, the wavelength of the light to be extracted is changed by changing the distance between the two parallel flat plates. In addition to this, a method using an acousto-optic tunable filter (AOTF) and a method using a liquid crystal tunable filter have been proposed.

  However, both methods are common in that they use a two-dimensional image sensor to acquire information in the wavelength direction at the x and y coordinate positions. For this reason, several systems can acquire hundreds of spectral information. On the other hand, real-time imaging is difficult, which is a trade-off with resolution and wavelength resolution. In addition, hyperspectral cameras are expensive, are mainly used for research purposes, and are currently not widely used.

  Considering these points, research is also being conducted to estimate hyperspectral information using a multispectral camera that acquires several bands of information in real time at a relatively low cost. As a technique for estimating hyperspectral information from multispectral information obtained by a multispectral camera, a technique is known in which spectral reflectance is approximated by a low-dimensional linear sum of base spectrum groups (see, for example, Patent Document 1). .

JP 2009-175926 A

  The technique described in Patent Document 1 is a method of estimating hyperspectral information from multispectral information by estimating spectral reflectance using a basis function. However, the above-mentioned patent document 1 states that “the number of dimensions required for a low-dimensional linear model of spectral reflectance is about 3 for human skin and the like, and more generally about 5 to 8”. There is no mention of a method for obtaining the basis number of the optimum basis function. Therefore, there is a problem that the estimation accuracy varies greatly depending on the selection of the basis number of the basis function used to estimate the hyperspectral information from the multispectral information. Therefore, the above-described problem is given as an example of the problem to be solved by the present invention.

  The present invention has been made in view of the above-mentioned problem as an example. The spectral reflectance is estimated using a basis function, and is subdivided from a plurality of filter images than the plurality of filter images. It is a main object of the present invention to provide an estimation device that improves the estimation accuracy when estimating a filtered image of different wavelength sections.

  According to the first aspect of the present invention, an imaging feature amount acquisition unit that acquires an imaging feature amount for each of the plurality of filter images from a plurality of filter images obtained by imaging a predetermined subject, and the acquired imaging features And an estimation unit configured to estimate a filter image of a wavelength segment subdivided from the plurality of filter images from the plurality of filter images based on a quantity.

  According to an eighth aspect of the present invention, when the illumination light spectrum is known, an imaging feature amount for each of a plurality of filter images obtained by imaging a subject with a known spectral reflectance and a predetermined subject is acquired. An imaging feature amount acquisition unit, a spectral reflectance estimation unit that estimates a spectral reflectance spectrum of the predetermined subject based on the acquired imaging feature amount for each of the plurality of filter images, and the estimated spectral An error calculation unit that calculates an error from the difference between the reflectance spectrum and the known spectral reflectance spectrum, and the basis number of the basis function is determined based on the calculated error. Spectral reflectance for each of the plurality of filter images is estimated based on the basis number, and the filter image of the wavelength division subdivided from the plurality of filter images than the plurality of filter images. Characterized in that and a estimation unit that estimates a.

  The invention according to claim 10 is an estimation method in an estimation apparatus including an imaging feature quantity acquisition unit and an estimation unit, and the imaging feature quantity acquisition unit is obtained by imaging a predetermined subject. A first step of acquiring imaging feature amounts for each of the plurality of filter images from a plurality of filter images; and the estimation unit, based on the acquired imaging feature amounts, from the plurality of filter images, And a second step of estimating a filter image of a wavelength segment that is subdivided from the filter image.

  The invention according to claim 11 is a program for causing a computer to execute an estimation method in an estimation apparatus including an imaging feature quantity acquisition unit and an estimation unit, wherein the imaging feature quantity acquisition unit is a predetermined A first step of acquiring imaging feature values for each of the plurality of filter images from a plurality of filter images obtained by imaging a subject; and the estimation unit, based on the acquired imaging feature values, And a second step of estimating a filter image of a wavelength segment that is subdivided from the plurality of filter images.

In an embodiment of the present invention, it is a figure explaining change of the estimation accuracy by the number of bases used for estimation according to the presence or absence of noise of image data. It is an electrical block diagram which shows an example of the estimation apparatus which concerns on embodiment of this invention. It is a processing flowchart which shows an example of the estimation apparatus which concerns on embodiment of this invention. It is the figure which showed typically the creation method of the data table in the estimation apparatus which concerns on Example 1 of this invention. It is an electrical block diagram which shows an example of the estimation apparatus which concerns on Example 1 of this invention. It is a processing flowchart which shows an example of the estimation apparatus which concerns on Example 1 of this invention. It is an electrical block diagram which shows an example of the estimation apparatus which concerns on Example 2 of this invention. It is a processing flowchart which shows an example of the estimation apparatus which concerns on Example 2 of this invention. It is the figure which showed typically the process of the estimation apparatus which concerns on Example 3 of this invention. It is an electrical block diagram which shows an example of the estimation apparatus which concerns on Example 3 of this invention. It is a processing flowchart which shows an example of the estimation apparatus which concerns on Example 3 of this invention.

<Embodiment>
Hereinafter, embodiments of the present invention will be described with reference to FIGS. 1 to 3.

As described above, a technique for estimating a filter image of a wavelength segment that is subdivided from the plurality of filter images from a plurality of filter images obtained by imaging is known. For example, in the estimation method using the basis function, if the unknown number n is equal to or less than the filter number m, a filter image of a wavelength segment that is subdivided from the plurality of filter images obtained by imaging from the plurality of filter images. Can be estimated.
Here, the basis number of the basis function used in this estimation method (in the following description, simply expressed as a basis number unless otherwise noted) is generally 3 to 8.
Note that the plurality of filter images refers to images composed of different wavelength bands obtained by spectrally separating near-ultraviolet to near-infrared light incident on the imaging unit from a subject. In the following, a plurality of filter images refers to the above images.

However, past knowledge and various verifications have found that there is a significant relationship between the number of bases, the number of filters, and the S / N values of a plurality of filter images obtained by imaging. Hereinafter, the content will be specifically described.
In FIG. 1, the horizontal axis represents the wavelength λ, the vertical axis represents the spectral reflectance S (λ), and the spectral reflectance S (λ) with respect to the wavelength λ is represented for each basis number. FIG. 1 is a diagram showing a case where there is no noise in the filter image in the upper stage, and a case where there is noise in the filter image in the lower stage. Further, in FIG. 1, the true value of the spectral reflectance S (λ) is indicated by a solid line, and the measured value of the spectral reflectance S (λ) is indicated by a broken line.
Here, referring to FIG. 1, when using ideal data (data having a large S / N value) without noise in a plurality of filter images obtained by imaging, the base number to be used is brought close to the number of filters. It can be seen that the estimation accuracy is improved. On the other hand, when using data with noise (data with a small S / N value) in a plurality of filter images obtained by imaging, the knowledge that the estimation result tends to break down when the number of bases is increased Can be obtained.

That is, the estimation accuracy of the filter image varies depending on the S / N values of a plurality of filter images obtained by imaging. Specifically, when the S / N value is large (when the noise of the acquired image is small), the estimation accuracy of the filter image can be improved by bringing the number of bases used closer to the filter number m. On the other hand, when the S / N value is small (when the acquired image contains a lot of noise), the estimation accuracy of the filter image is improved by making the base number to be used smaller than the filter number m. be able to.
In the following, the present embodiment and examples described in detail are based on the above findings.

<Electrical configuration of estimation apparatus>
An electrical configuration of the estimation apparatus 10 according to the present embodiment will be described with reference to FIG.
As illustrated in FIG. 2, the estimation apparatus 10 according to the present embodiment includes an imaging unit 110, an imaging feature amount acquisition unit 111, an estimation unit 130, and a control unit 140.

The imaging unit 110 obtains a plurality of filter images by imaging a predetermined subject.
The imaging feature amount acquisition unit 111 acquires the imaging feature amount for each of the plurality of filter images from the plurality of filter images obtained from the imaging unit 110. Here, the subject may be an object, a human body, a living organism, or a tissue thereof. Further, the number of filters m for obtaining the filter image is not particularly limited, but considering the filter image estimation processing efficiency, the number of filters of the order of the multispectral camera is preferable. Further, examples of the imaging feature amount include a filter number m, a pixel value of each pixel, a luminance value, an S / N value, a camera gain, and the like. In the present embodiment, the imaging feature quantity acquisition unit 111 acquires the S / N value of the filter image as the imaging feature quantity. Here, the acquisition of the S / N value may be acquired by direct measurement, or an imaging feature amount having a significant relationship with the S / N value may be estimated as an index.

Based on the imaging feature amount acquired by the imaging feature amount acquisition unit 111, the estimation unit 130 estimates a filter image of a wavelength segment that is subdivided from the plurality of filter images from the plurality of filter images. In the present embodiment, it is assumed that either the illumination light spectrum or the spectral reflectance is known as a condition for estimation.
Specifically, when the illumination light spectrum is known, when the S / N value acquired as the imaging feature quantity by the imaging feature quantity acquisition unit 111 is larger than the predetermined S / N value. The estimation unit 130 brings the base number to be used closer to the filter number m. On the other hand, when the acquired S / N value is smaller than a predetermined S / N value, the estimation unit 130 makes the basis number to be used smaller than the filter number m. Then, the weighting coefficient of the basis function is calculated from the obtained basis number, and the spectral reflectance of the subject is estimated. As a result, a color signal spectrum for each pixel of the subject is derived, and a filter image of a wavelength segment subdivided from the plurality of filter images is estimated. The “predetermined S / N value” can be exemplified by, for example, the S / N value determined as having no noise shown in FIG.

  The control unit 140 controls the operation of the estimation device 10 according to the present embodiment based on a control program stored therein. For example, the control unit 140 instructs the imaging unit 110 to image a predetermined subject. In addition, the control unit 140 inputs imaging data from the imaging unit 110 and outputs the input imaging data to the imaging feature amount acquisition unit 111 in response to a request from the imaging feature amount acquisition unit 111. In addition, the control unit 140 inputs the imaging feature amount acquired by the imaging feature amount acquisition unit 111, and executes processing such as outputting the input imaging feature amount to the estimation unit 130 in response to a request from the estimation unit 130. .

<Processing of estimation device>
Processing of the estimation apparatus 10 according to the present embodiment will be described with reference to FIG.

  First, the imaging feature quantity acquisition unit 111 acquires imaging feature quantities for each of a plurality of filter images from a plurality of filter images obtained by the imaging unit 110 imaging a predetermined subject (S101). Then, the estimation unit 130 estimates a filter image of a wavelength segment that is subdivided from the plurality of filter images from the plurality of filter images based on the acquired imaging feature amount (S102).

As described above, according to the estimation apparatus 10 according to the present embodiment, for example, when the imaging feature quantity acquisition unit 111 acquires S / N values as imaging feature quantities, the S of the plurality of filter images is obtained. Based on the / N value, it is possible to determine the optimum basis number to be used for estimating the filter image of the wavelength segment subdivided from the plurality of acquired filter images.
Furthermore, the estimation apparatus 10 according to the present embodiment estimates the filter image of the wavelength segment subdivided from the plurality of acquired filter images using the determined optimum basis number. Therefore, according to the estimation apparatus 10 according to the present embodiment, it is possible to estimate a filter image with high estimation accuracy.

<Example>
A specific example corresponding to the above-described embodiment will be described with reference to FIGS.

Prior to the description of specific examples, an estimation method used in the following examples will be described.
The estimation method used as an example in the following embodiments is to estimate a filter image of further subdivided wavelength sections by approximation by combining a plurality of basis functions.

  First, the color signal of the subject is modeled. For example, when the wavelength is λ, the spectral reflectance of the surface of the subject is S (λ), the illumination light spectrum is E (λ), and the weighting coefficient at a certain position x is α (x), The color signal C (x, λ) can be modeled as shown in Equation 1 below. However, here, in order to simplify the description, it is assumed that the specular reflection component is “0”.

Next, the spectral reflectance S (λ) is approximated by a linear sum of n basis vectors and expressed as shown in Equation 2. Here, S _i is the i-th basis function, and σ _i is a weighting coefficient for the i-th basis function. In general, basis functions (basis vectors) are used in order from the highest contribution vector among the calculated eigenvectors by performing principal component analysis on a data group whose spectral reflectance is known.

In addition, a multispectral camera is exemplified as the imaging unit 110. Here, assuming that the _kth sensor output signal of the multispectral camera is ρ _k (x) and the sensor sensitivity of the _kth multispectral camera is R _k (λ), the kth sensor output signal ρ _k of the multispectral camera. (X) can be expressed as Equation 3 below.

Here, if the illumination light spectrum E (λ) is known, the spectral reflectance of one pixel can be estimated by estimating the weighting coefficient σ of the basis function if the number of filters m is greater than or equal to the unknown number n. . The unknown number n is the order of the weight vector of the basis function (basis vector), and varies depending on the basis number (order of the basis vector) used for estimation.
In the following example, the spectral reflectance S (λ) is estimated by determining the basis number, and the filter image of the wavelength segment subdivided from the plurality of acquired filter images is estimated. To do.

<Example 1>
Hereinafter, Example 1 is demonstrated using FIGS. 4-6.
In this embodiment, the average luminance value of the filter image is used as an index of the S / N value. For example, a data table having a plurality of data groups each consisting of an average luminance value in a filter image of a plurality of known subjects having different spectral reflectances and an optimum base number is prepared in advance.
Specifically, as shown in FIG. 4, the spectral reflectance of the subject A that has been measured in advance is assumed to be a true value. Then, an average luminance value of a plurality of filter images obtained by imaging the subject A is obtained, and the lowest average luminance value is specified. Furthermore, for the filter image having the lowest average luminance value, spectral reflectances at all base numbers are estimated, the estimation result is compared with the true value, an error is calculated, and the error is within the threshold value. Among the certain base numbers, the maximum base number is specified.
More specifically, the process described above is performed a plurality of times while changing the brightness of the illumination light to obtain a data set in which the lowest average luminance value is associated with the maximum basis number. Further, the same processing is performed for a plurality of subjects other than the subject A having different spectral reflectance spectra. Then, a data table including a plurality of data groups in which the maximum base number obtained from the above processing and the lowest average luminance value are paired is created.
When estimating the spectral reflectance of the filter image of a predetermined subject, the average luminance value of the filter image is calculated, and the basis number is determined with reference to the data table based on the average luminance value. Then, using the determined basis number, the spectral reflectance of a predetermined subject is obtained for all the pixels constituting the plurality of filter images. And the filter image of the wavelength division subdivided rather than the some filter image from the some filter image is estimated.
That is, in the present embodiment, the estimation process is performed on the entire filter image by applying the determined basis number to all the pixels of the filter image using the data table prepared in advance.
Hereinafter, a configuration for performing the above-described estimation process and detailed processing contents will be described.

<Electrical configuration of estimation apparatus>
The electrical configuration of the estimation apparatus 11 according to the present embodiment will be described with reference to FIG.
As shown in FIG. 5, the estimation apparatus 11 according to the present embodiment includes an imaging unit 110, an imaging feature amount acquisition unit 111, a first calculation unit 121, a second calculation unit 122, and a spectrum estimation unit 123. And an error calculation unit 124, a storage unit 125, an estimation unit 131, and a control unit 141.
In this embodiment, as an example, the second calculation unit 122, the spectrum estimation unit 123, the error calculation unit 124, the storage unit 125, and the control unit 141 function to create the data table. In the case where the imaging unit 110, the imaging feature quantity acquisition unit 111, the first calculation unit 121, the storage unit 125, the estimation unit 131, and the control unit 141 function for filter image estimation, ,explain.
In addition, in the following examples including the present example and excluding Example 4, it is assumed that the illumination light spectrum is known.

  The imaging unit 110 obtains a plurality of filter images by imaging a predetermined subject. In addition, although the present Example illustrated having the imaging part 110 in the estimation apparatus 11, the structure which acquires a filter image is not restricted to this. For example, a configuration may be adopted in which a filter image captured by an external imaging unit is acquired without including the imaging unit 110 in the estimation device 11. In addition, for example, a configuration may be employed in which the imaging unit 110 acquires a filter image that has already been captured from a storage unit in the estimation apparatus 11 that stores the filter image or an external server.

  The imaging feature amount acquisition unit 111 acquires the imaging feature amount for each of the plurality of filter images from the plurality of filter images obtained from the imaging unit 110.

  The first calculation unit 121 calculates a desired imaging feature amount of a predetermined subject from the imaging feature amount acquired by the imaging feature amount acquisition unit 111. In this embodiment, as an example, the average luminance value of each filter image is calculated from the luminance value acquired by the imaging feature amount acquisition unit 111. Then, it can be exemplified that the lowest average luminance value among the calculated average luminance values of each filter image is output to the estimation unit 131 described later.

The second calculation unit 122 changes the brightness of the illumination light, out of the average luminance value for each of the plurality of filter images obtained by imaging a plurality of subjects each having a different spectrum of known spectral reflectance. The lowest average luminance value is calculated as a desired imaging feature amount.
Then, the second calculation unit 122 outputs the calculated lowest average luminance value to the control unit 141 described later.

  The spectrum estimation unit 123 estimates the spectrum of spectral reflectance for each of a plurality of base numbers for the filter image for which the lowest average luminance value is calculated by the second calculation unit 122. Here, the basis numbers used for estimation are all the basis numbers less than or equal to the filter number m.

The error calculation unit 124 compares the spectrum of the known spectral reflectance with the spectrum of the spectral reflectance for each of the plurality of base numbers estimated by the spectrum estimation unit 123.
Then, the error calculation unit 124 calculates an error of the spectral reflectance spectrum for each of the estimated plurality of base numbers with respect to the known spectral reflectance spectrum, and outputs it to the control unit 141 described later.

The control unit 141 controls the operation of the estimation device 11 according to the present embodiment based on a control program stored therein.
Further, as an example, the control unit 141 includes a base whose error is within a predetermined threshold among the lowest average luminance value calculated by the second calculation unit 122 and the error calculated by the error calculation unit 124. A data table in which the largest base number among the numbers is associated is created and stored in the storage unit 125 described later.
In the above, the average luminance value is exemplified, but when the camera gain can be acquired, the camera gain value may be used as an index of the S / N value instead of the average luminance value. In this case, it is assumed that a table in which the camera gain value and the optimum base number are associated with each other is stored in the storage unit 125 described later.

The storage unit 125 stores a data table in which desired imaging feature amounts and basis numbers are associated with each other. In the present embodiment, as an example, storing the data table created by the control unit 141 can be exemplified. The storage unit 125 also stores spectral information (spectral data) of spectral reflectance of a subject whose spectral reflectance is known.
Moreover, in the present Example, although having illustrated having the memory | storage part 125 in the estimation apparatus 11 as mentioned above, the structure which memorize | stores a data table is not restricted to this. For example, the configuration may be such that the data table and data are stored in an external server without providing the storage unit 125 in the estimation device 11.

  The estimation unit 131 uses the lowest average luminance value among the average luminance values of the respective filter images calculated by the first calculation unit 121 as an S / N value index, and stores in the storage unit 125 based on the average luminance value. The base number is determined with reference to the information in the data table.

  Further, when the estimation unit 131 determines the number of bases used for the estimation process, the pixel of the pixel located at the corresponding coordinate in the m filter images (the pixel at the same coordinate position in the m filter images). A value is acquired, and the acquired pixel value is used as an input, and a weighting coefficient of a basis function is obtained by the above equation (7). Next, the estimation unit 131 estimates the spectral reflectance by approximation of the basis function according to Equation 2 above. Further, the estimation unit 131 substitutes the estimated spectral reflectance into the above formula 1, and estimates the color signal spectrum of the pixel located at the corresponding coordinate in the m filter images. And the estimation part 131 estimates the filter image of the wavelength division subdivided rather than the acquired several filter image by implementing such a process about all the pixels which comprise a several filter image. .

<Processing of estimation device>
The process of the estimation apparatus 11 which concerns on a present Example is demonstrated using FIG.
In the following description, in order to make the description more specific, the imaging unit 110 is exemplified as a multispectral camera, and an estimation process based on a filter image captured by the multispectral camera will be described.

  First, as shown in FIG. 6, m filter images corresponding to the number of filters m are acquired by a multispectral camera (S201). The first calculation unit 121 calculates the average luminance value of each acquired filter image (S202).

The control unit 141 determines whether or not the average luminance value is calculated by the first calculation unit 121 for the m filter images (S203), and determines that the average luminance value is not calculated (“No” in S203). ]), A calculation process is performed on the next filter image.
On the other hand, when it is determined that the average luminance value has already been calculated for the m filter images (“Yes” in S203), the control unit 141 advances the process to S204.

  In S <b> 204, the first calculation unit 121 determines the lowest average luminance value among the average luminance values calculated for the m filter images and outputs the determined average luminance value to the estimation unit 131. Based on the lowest average luminance value determined by the first calculation unit 121, the estimation unit 131 determines the base number used for the estimation process. More specifically, the estimation unit 131 determines the number of bases by referring to a data table that defines the lowest average luminance value that has been created in advance stored in the storage unit 125 and the number of bases used for the estimation process. (S204).

In S204, after determining the base number, the estimation unit 131 acquires the pixel value of the pixel located at the corresponding coordinate in the m filter images (S205). Then, the estimation unit 131 receives the obtained pixel value as an input, obtains the weighting coefficient of the basis function by the above equation 7, and then estimates the spectral reflectance by the approximation of the basis function by the above equation 2 (S206). . After estimating the spectral reflectance, the control unit 141 determines whether or not the spectral reflectance estimation processing has been performed on all pixel values in the image (S207). When the control unit 141 determines that the spectral reflectance estimation process is not performed on all pixel values (“No” in S207), the process is performed on the next pixel value. The process returns to S204.
On the other hand, when the control unit 141 determines that the spectral reflectance estimation process has been performed on all pixel values (“Yes” in S207), the spectral reflectance estimation process ends.

  Further, the estimation unit 131 substitutes the estimated spectral reflectance into the above formula 1, and estimates the color signal spectrum of the subject for the pixel located at the corresponding coordinate in the m filter images. And the estimation part 131 estimates the filter image of the wavelength division subdivided rather than the acquired some filter image by implementing such a process about all the pixels which comprise an image.

As described above, in the estimation apparatus 11 according to the present embodiment, an error between the spectrum of the known spectral reflectance and the spectrum of the spectral reflectance estimated in the filter image that has obtained the lowest average luminance value is a threshold value. A data table is used in which the maximum base number among the base numbers that are within is associated with the lowest average brightness value among the average brightness values of each filter image. Then, for the lowest average luminance value among the average luminance values of the plurality of filter images with respect to the subject whose spectral reflectance spectrum is unknown, the base number used for estimation of the filter image is determined with reference to the data table. To do.
Therefore, according to the estimation apparatus 11 according to the present embodiment, the average luminance value of a plurality of filter images obtained by imaging a predetermined subject using the relationship between the S / N value and the estimated quality of the filter image. By comparing the lowest average luminance value with the data table, the optimum base number can be automatically determined by a simple method.

Moreover, in the estimation apparatus 11 according to the present embodiment, the average luminance value is used as an index of the S / N value of the filter image. Furthermore, in the estimation apparatus 11 according to the present embodiment, among the average luminance values of each filter image, the lowest average luminance value, the spectrum of the known spectral reflectance in the filter image that obtained the lowest average luminance value, and a plurality The optimum base number is determined by a combination with the maximum base number among the base numbers whose error from the spectral reflectance spectrum for each base number is within a predetermined threshold.
That is, in the estimation apparatus 11 according to the present embodiment, the basis is calculated from the average luminance value on the data table corresponding to the lowest average luminance value among the average luminance values of the plurality of filter images obtained by imaging a predetermined subject. Determine the number and estimate the filter image.
Thereby, it is possible to prevent a situation in which the estimation result shown in FIG.
Further, in the estimation device 11, the basis number on the data table is set to the maximum of the basis numbers whose error between the spectrum of the known spectral reflectance and the spectrum of the spectral reflectance for each of the plurality of basis numbers is within a predetermined threshold. The base number of. Thereby, the estimation error of the filter image can be minimized.
Therefore, according to the estimation apparatus 11 according to the present embodiment, it is possible to estimate a filter image that minimizes the estimation error regardless of the value of the average luminance value, and consequently improve the estimation quality of the filter image. be able to.

<Example 2>
Hereinafter, Example 2 will be described with reference to FIGS. 7 and 8.
In this embodiment, the spectral reflectance is individually estimated for all the pixels constituting the filter image, and the filter image is estimated. In addition, when estimating the spectral reflectance, spectral reflectance is estimated using all the base numbers, and the base number satisfying a predetermined condition for the spectrum of the spectral reflectance as an estimation result is set to an optimum value for each pixel. Let it be a base number. And the filter image of the wavelength division subdivided rather than the some filter image obtained by imaging is estimated.
Hereinafter, a configuration for performing the above-described estimation process and detailed processing contents will be described.

<Electrical configuration of estimation apparatus>
The electrical configuration of the estimation device 12 according to the present embodiment will be described with reference to FIG.
As illustrated in FIG. 7, the estimation device 12 according to the present embodiment includes an imaging unit 110, an imaging feature amount acquisition unit 112, a storage unit 126, an estimation unit 132, and a control unit 142. Yes.

  The imaging feature amount acquisition unit 112 acquires imaging feature amounts for each of the plurality of filter images from the plurality of filter images obtained by the imaging unit 110. In the present embodiment, acquisition of pixel values of all pixels constituting a plurality of filter images can be exemplified as the imaging feature amount.

The control part 142 controls operation | movement of the estimation apparatus 12 which concerns on a present Example based on the control program stored inside.
Further, as an example, the control unit 142 reads out a threshold value from the storage unit 126 and outputs the threshold value to the estimation unit 132 so that the estimation unit 132 described later determines the base number. In addition, the control unit 142 performs initial setting of the base number and update processing of the base number in the processing of the estimation device 12 described later. Furthermore, the control unit 142 inputs the spectral reflectance of the pixel located at the corresponding coordinates in the plurality of estimated filter images satisfying the predetermined condition from the estimation unit 132, and stores this together with the pixel information, the storage unit 126 described later. Remember me.

  The storage unit 126 stores a threshold used by the estimation unit 132 to be described later to determine the basis number. In addition, the storage unit 126 stores the spectral reflectance of the pixel located at the corresponding coordinate in the plurality of estimated filter images satisfying the predetermined condition together with the pixel information.

  The estimation unit 132 sequentially acquires pixel values of pixels located at corresponding coordinates in the plurality of filter images for all the pixels constituting the plurality of filter images acquired by the imaging feature amount acquisition unit 112. Then, the estimation unit 132 receives the acquired pixel value as input, and estimates the spectral reflectance for all the base numbers less than or equal to the filter number m. In addition, the estimation unit 132 performs a noise removal process on the spectrum of the spectral reflectance of the pixel located at the corresponding coordinate in the plurality of estimated filter images, and then uses a basis number that satisfies a predetermined condition. Based on the estimated spectral reflectance for each pixel, a filter image of a wavelength section subdivided from the plurality of filter images is estimated from the plurality of filter images.

In this embodiment, as an example, the estimation unit 132 uses the acquired pixel value as an input, and obtains the weighting coefficient of the basis function using the above equation 7 for all the basis numbers less than or equal to the filter number m. Next, the estimation unit 132 estimates the spectral reflectance by the approximation of the basis function according to the above formula 2. Then, the estimation unit 132 associates the spectral reflectance of each pixel that satisfies the predetermined condition of the spectral reflectance spectrum after the noise removal processing with respect to the estimated spectral reflectance spectrum. Output to. Further, the estimation unit 132 estimates, based on the spectral reflectance for each pixel stored in the storage unit 126, a filter image of a wavelength segment that is subdivided from the plurality of filter images. Can be illustrated.
Here, the predetermined condition is, for example, a relationship between a threshold value determined in advance by the spectral reflectance spectrum of a pixel located at a corresponding coordinate in a plurality of filter images and a differential value obtained by differential operation. The number of times that the differential value exceeds the first threshold is equal to or less than the predetermined number, or the sum of the differential values does not exceed a predetermined second threshold. Then, the basis number used for estimating the spectrum of the spectral reflectance that satisfies the predetermined condition is set as the optimum basis number.
In addition, when there are a plurality of base numbers that satisfy a predetermined condition, it is preferable to use an estimation result based on a larger base number.

<Processing of estimation device>
The process of the estimation apparatus 12 according to the present embodiment will be described with reference to FIG.
In the following description, in order to make the description more specific, the imaging unit 110 is exemplified as a multispectral camera, and an estimation process based on a filter image captured by the multispectral camera will be described.

As shown in FIG. 8, first, the control unit 142 determines the number of bases used for estimation based on the number of filters of the multispectral camera (S301). Here, if the number of filters of the multispectral camera is m, a value m _s less than or equal to m is set as the base number.

  Next, m multispectral images (a plurality of filter images) obtained by imaging a predetermined subject are acquired (S302). The estimation unit 132 acquires the pixel value of the pixel located at the corresponding coordinate in the m filter images acquired by the imaging feature amount acquisition unit 112 (S303). Then, the estimation unit 132 uses the pixel value acquired in S303 as an input, and uses the basis number determined in S301 to estimate the spectral reflectance of the pixel located at the corresponding coordinate in the acquired m filter images. Perform (S304).

Then, the estimation unit 132 performs secondary differentiation on the spectrum of the spectral reflectance of the pixel located at the corresponding coordinate in the filter image where the spectral reflectance is estimated, and calculates the sum of the secondary differential values. (S305). Next, the control unit 142 determines whether the sum exceeds a threshold value stored in advance in the storage unit 126 (S306). If the sum exceeds the threshold value ("Yes" in S306), Judge that there are many errors in the estimation results. At this time, the control unit 142 updates the base number used for estimation to m _s −1 (S307).

  On the other hand, when the sum of the secondary differential values does not exceed the threshold value (“No” in S306), the control unit 142 determines that the estimation result has little error, and the pixel located at the corresponding coordinate in the filter image As the final estimation result, the estimated spectral reflectance is stored in the storage unit 126 in association with the pixel information. Then, the control unit 142 determines whether or not the spectral reflectance estimation processing has been performed on all the pixels in the image (S308). If the control unit 142 determines that spectral reflectance estimation processing has not been performed for all pixels (“No” in S308), the control unit 142 returns the processing to S303 and performs processing for the next pixel. .

On the other hand, if the control unit 142 determines that spectral reflectance estimation processing has been performed for all pixels (“Yes” in S308), the control unit 142 estimates each pixel stored in the storage unit 126. The color signal spectrum of the subject for the pixel is estimated from the spectral reflectance obtained. Then, by executing this process for all the pixels constituting the image, the filter image of the wavelength segment subdivided from the acquired filter image is estimated, and the process ends.
In the above description, the second-order differentiation is used for the noise removal processing. However, the present invention is not limited to this as long as the method can remove only unnecessary noise without affecting the original spectrum.

  Further, in this embodiment, an optimal basis number is obtained on the condition that the sum of the differential values of spectral reflectance spectra of pixels located at corresponding coordinates in a plurality of filter images does not exceed the second threshold value. The method was illustrated. However, instead of this, the predetermined condition is that the number of times that the differential value of the spectrum of the spectral reflectance of the pixel located at the corresponding coordinate in the plurality of filter images exceeds the first threshold is equal to or less than the predetermined number of times. An optimum basis number may be obtained.

As described above, in the estimation apparatus 12 according to the present embodiment, all pixels constituting a plurality of filter images obtained by imaging a predetermined subject are located at corresponding coordinates in the plurality of filter images. The pixel values of the pixels are acquired sequentially. Then, using the acquired pixel value as an input, spectral reflectance is estimated for all the base numbers less than or equal to the filter number m. Further, with respect to the estimated spectral reflectance spectrum, the spectrum reflectance spectrum after the noise removal process satisfies a predetermined condition, and determines the number of bases for spectral reflectance spectrum estimation for each pixel. From the filter image, a filter image of a wavelength segment subdivided from the plurality of filter images is estimated. That is, the optimal base number is determined for each pixel constituting the filter image, and the filter image is estimated.
Therefore, according to the estimation apparatus 12 according to the present embodiment, the spectrum of the spectral reflectance estimated for each pixel is evaluated based on the relationship between the S / N value and the estimated quality of the filter image. Thereby, it is possible to directly evaluate the estimated quality of the filter image and determine the optimum base number for each pixel. Therefore, it is possible to improve the estimation quality related to the estimation of the filter image of the wavelength segment subdivided than the filter image.

  Further, according to the estimation device 12 according to the present embodiment, the control unit 142 performs spectral processing after noise removal processing on the spectral reflectance spectrum of the pixels located at the corresponding coordinates in the plurality of estimated filter images. Depending on whether or not the reflectance spectrum satisfies a predetermined condition, the magnitude of the error in the estimation result of the filter image is determined, and the optimum basis number is determined. That is, the estimation error of the filter image is used as an index of the S / N value of a plurality of captured filter images. For this reason, regarding the estimation of the filter image, the estimated quality of the filter image can be evaluated with an appropriate index.

<Example 3>
Hereinafter, Example 3 is demonstrated using FIGS. 9-11.
In this embodiment, as illustrated in FIG. 9, a subject with a known spectral reflectance (reference body B in the figure) and a predetermined subject (subject in the figure) are imaged simultaneously, The imaging feature amount for each filter image is acquired. Further, the spectral reflectance of a predetermined subject is estimated using the acquired pixel value of the known subject. The basis number is determined from the error between the estimated spectral reflectance spectrum and the known spectral reflectance spectrum, and estimation processing is performed on the entire image using this basis value.
Hereinafter, a configuration for performing the above-described estimation process and detailed processing contents will be described.

<Electrical configuration of estimation apparatus>
The electrical configuration of the estimation device 13 according to the present embodiment will be described with reference to FIG.
As shown in FIG. 10, the estimation apparatus 13 according to the present embodiment includes an imaging unit 110A, an imaging feature amount acquisition unit 113, a spectral reflectance estimation unit 127, a storage unit 128, an error calculation unit 129, and an estimation. The unit 133 and the control unit 143 are included.

The imaging unit 110A obtains a plurality of filter images by simultaneously imaging a subject with a known spectral reflectance and a predetermined subject.
The imaging feature amount acquisition unit 113 acquires the imaging feature amount for each of the plurality of filter images captured by the imaging unit 110A. Here, as the imaging feature amount, a pixel value of a pixel obtained by imaging a subject with a known spectral reflectance can be exemplified.

The spectral reflectance estimation unit 127 estimates the spectral reflectance of a predetermined subject based on the imaging feature amount for each of the plurality of filter images acquired by the imaging feature amount acquisition unit 113.
In the present embodiment, as an example, the spectral reflectance estimation unit 127 acquires a pixel value from a filter image of a subject (reference body B) having a known spectral reflectance, and sets the basis number to a value _ms that is equal to or less than the filter number m. Setting and estimating the spectral reflectance for all the set base numbers can be exemplified (see FIG. 9). Note that the spectral reflectance estimation method is the same as that in the first or second embodiment.

Storage unit 128, the spectral reflectance are storing m _s number the number of the base to be used in the estimation process in the spectral information or spectral reflectance estimation unit 127 of the spectral reflectance of a known object. Further, an error calculated by an error calculation unit 129 described later is stored.

The error calculation unit 129 calculates an error from the difference between the spectrum of the spectral reflectance estimated by the spectral reflectance estimation unit 127 and the spectrum of the known spectral reflectance.
In the present embodiment, as an example, spectral information of known spectral reflectance is acquired from the storage unit 128 via the control unit 143, and the acquired spectral information of spectral reflectance and the spectral reflectance estimation unit 127 are estimated. The difference between the measured spectral reflectance spectrum and the spectrum is calculated as an error (see FIG. 9).

  The estimation unit 133 subdivides the plurality of filter images from the plurality of filter images by estimating the spectral reflectance of the entire image of the predetermined subject based on the basis number determined by the control unit 143 described later. Estimate the filter image of the wavelength section. Note that the filter image estimation method is the same as in the first or second embodiment.

The control unit 143 controls the operation of the estimation device 13 according to the present embodiment based on a control program stored therein.
Further, in this embodiment, as an example, it reads from the m _s-number of base number storage unit 128 to use the spectral reflectance estimating section 127 in descending order of their numbers, by outputting the spectral reflectance estimation unit 127, spectral The reflectance estimation unit 127 sets the number of bases used for estimation. In addition, the control unit 143 reads out spectral information of known spectral reflectance from the storage unit 128 and outputs it to the error calculation unit 129. Further, the control unit 143 determines the basis number that minimizes the error from the errors calculated by the error calculation unit 129 and outputs the basis number to the estimation unit 133.

<Processing of estimation device>
Processing of the estimation device 13 according to the present embodiment will be described with reference to FIG.
In the following description, in order to make the description more specific, the imaging unit 110A is exemplified as a multispectral camera, and an estimation process based on a filter image captured by the multispectral camera will be described.

First, the control unit 143 determines the number of bases used by the spectral reflectance estimation unit 127 for estimation based on the number of filters of the multispectral camera (S401). Here, if the number of filters of the multispectral camera is m, a value m _s less than or equal to m is set as the base number.

  Next, the multispectral camera captures a subject with a known spectral reflectance and a predetermined subject to obtain m multispectral images (a plurality of filter images) (S402). The imaging feature amount acquisition unit 113 acquires the imaging feature amount for each of the plurality of filter images acquired by the multispectral camera. (S403).

  The spectral reflectance estimation unit 127 acquires a pixel value from the filter image of the subject whose spectral reflectance is known, and estimates the spectral reflectance with respect to the basis number set by the control unit 143 (S404).

  The error calculation unit 129 acquires the spectral information of the known spectral reflectance from the storage unit 128 via the control unit 143. Then, the error calculation unit 129 calculates a difference between the acquired spectral information of the spectral reflectance and the spectral reflectance spectrum estimated by the spectral reflectance estimation unit 127 as an error, and outputs the difference to the control unit 143. (S405).

  The control unit 143 determines whether or not the error input from the error calculation unit 129 is smaller than the errors input from the error calculation unit 129 so far (S406). If it is determined that the value is small (“Yes” in S406), the error stored in the storage unit 128 is rewritten and the spectral reflectance is estimated for all the base numbers set by the control unit 143. It is determined whether or not is completed (S408).

  On the other hand, when the control unit 143 determines that the error input from the error calculation unit 129 is larger than the error input from the error calculation unit 129 so far (“No” in S406), the control unit 143 updates the basis number. (S407), and the process returns to S404.

In S408, if the control unit 143 determines that the spectral reflectance estimation has not been completed for all the base numbers set by the control unit 143 ("No" in S408), the base is determined. The number is updated from m to m-1 (S407), and the process returns to S404.
On the other hand, when it is determined that the estimation of the spectral reflectance has been completed for all the base numbers set by the control unit 143 (“Yes” in S408), the control unit 143 stores the storage unit 128. The base number is determined from the error stored in, and the determined base number is output to the estimation unit 133 and the process is terminated.

  Based on the basis number input from the control unit 143, the estimation unit 133 uses the same method as in the first embodiment to filter the wavelength division filters subdivided from the plurality of filter images. Estimate the image.

As described above, in the estimation apparatus 13 according to the present embodiment, the imaging feature amount acquisition unit 113 performs imaging for each of a plurality of filter images obtained by simultaneously imaging a subject with a known spectral reflectance and a predetermined subject. The imaging feature amount of is acquired. Then, the spectral reflectance estimation unit 127 estimates the spectral reflectance of a predetermined subject for all the base numbers set by the control unit 143 using the acquired pixel values of the known subject. Further, the error calculation unit 129 calculates an error from the difference between the estimated spectral reflectance spectrum and the known spectral reflectance spectrum, and the control unit 143 determines the base number that minimizes the error. Further, the estimation unit 133 estimates the spectral reflectance of a predetermined subject based on the determined basis number, and from the plurality of filter images, filter images of wavelength sections that are subdivided from the plurality of filter images. presume.
Therefore, according to the estimation apparatus 13 according to the present embodiment, the spectrum of the estimated spectral reflectance is evaluated based on the relationship between the S / N value and the estimated quality of the filter image. Thereby, the estimated quality of the filter image can be directly evaluated to determine the optimum basis number.

In addition, according to the estimation device 13 according to the present embodiment, the estimation result that is a difference between the spectrum of the estimated spectral reflectance and the spectrum of the known spectral reflectance is obtained using the acquired pixel value of the known subject. The optimum basis number is determined by determining the magnitude of the error.
That is, the estimation error is used as an index of the S / N value of a plurality of filter images taken. For this reason, regarding the estimation of the filter image, the estimation quality can be evaluated with an appropriate index.

<Example 4>
In the above-described first to third embodiments, the illumination light spectrum is assumed to be known, and the spectral reflectance is estimated, so that the filter images of the wavelength segments subdivided from the plurality of filter images are subdivided from the plurality of filter images. The estimation method was described.
In this embodiment, the spectral reflectance is known and the illumination light spectrum is estimated, so that a filter image of a wavelength segment subdivided from the plurality of filter images is estimated from the plurality of filter images.

  Specifically, Equations 2 and 4 are changed to Equations 8 and 9 below to obtain arbitrary pixel values, and the weighting coefficient of the basis function of the illumination light spectrum is obtained. Then, the illumination light spectrum is estimated by substituting the weighting coefficient of the basis function of the illumination light spectrum into Equation 8 and approximating the illumination light spectrum with the basis function. Furthermore, by substituting the estimated illumination light spectrum into Equation 1 to estimate the color signal spectrum at an arbitrary pixel and performing these processes for all the pixels, a plurality of filter images can be obtained from the plurality of filter images. The filter image of the wavelength division subdivided more is estimated.

  As described above, according to the present embodiment, the spectral reflectance is assumed to be known, and the illumination light spectrum is estimated. It is possible to estimate the filtered image of the wavelength division with high estimation quality.

  In the above description from the embodiments to the examples, the plurality of filter images have been described as images having different wavelength bands obtained by spectrally separating light from near ultraviolet to near infrared incident from the subject to the imaging unit. However, the plurality of filter images may be images composed of different frequency bands obtained by spectrally separating near-ultraviolet to near-infrared light incident on the imaging unit from the subject.

  Note that the estimation apparatus of the present invention can be realized by recording the processing of the estimation apparatus on a recording medium readable by a computer system, causing the estimation apparatus to read and execute a program recorded on the recording medium. The computer system here includes an OS and hardware such as peripheral devices.

  Further, the “computer system” includes a homepage providing environment (or display environment) if a WWW (World Wide Web) system is used. The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line.

  The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, what is called a difference file (difference program) may be sufficient.

  Although the embodiments and examples of the present invention have been described in detail with reference to the drawings, the specific configuration is not limited to the embodiments or examples, and the scope of the present invention is not deviated. Design etc. are also included.

10 to 13; Estimation devices 110 and 110A; Imaging units 111 to 113; Imaging feature amount acquisition unit 121; First calculation unit 122; Second calculation unit 123; Spectrum estimation unit 124; Error calculation units 125, 126, and 128 Storage unit 127; spectral reflectance estimation unit 129; error calculation unit 130 to 133; estimation unit 140 to 143; control unit

Claims (11)

An imaging feature amount acquisition unit that acquires an imaging feature amount for each of the plurality of filter images from a plurality of filter images obtained by imaging a predetermined subject;
Based on the acquired imaging feature amount, an estimation unit that estimates a filter image of a wavelength segment subdivided from the plurality of filter images from the plurality of filter images;
An estimation device comprising: When the illumination light spectrum or spectral reflectance is known,
The imaging feature amount acquisition unit acquires an S / N value as the imaging feature amount,
The estimation unit includes
The S / N value acquired by the imaging feature amount acquisition unit is
When larger than a predetermined S / N value, the basis number of the basis function is brought close to the number of filters for acquiring the plurality of filter images,
If it is smaller than the predetermined S / N value, the basis number of the basis function is made smaller than the number of filters for obtaining the plurality of filter images,
Determining the basis number, and determining the illumination light spectrum or the spectral reflectance of the predetermined subject for each filter image based on the determined basis number;
The estimation apparatus according to claim 1, wherein a filter image of a wavelength segment that is subdivided from the plurality of filter images is estimated from the plurality of filter images. A storage unit that stores a data table in which desired imaging feature amounts and basis numbers of basis functions are associated;
A first calculation unit that calculates a desired imaging feature amount of the predetermined subject from the imaging feature amount acquired by the imaging feature amount acquisition unit;
With
When the illumination light spectrum is known,
The estimation unit is
Determining the basis number corresponding to the desired imaging feature amount calculated in the first calculation unit from the data table;
Based on the determined basis number, the spectral reflectance of the predetermined subject is obtained,
The estimation apparatus according to claim 1, wherein a filter image of a wavelength segment that is subdivided from the plurality of filter images is estimated from the plurality of filter images. The lowest average luminance value among the average luminance values for each of the plurality of filter images obtained by changing the brightness of the illumination light and imaging the plurality of subjects having different spectral reflectances that are already known. A second calculation unit for calculating a value as the desired imaging feature amount;
A spectrum estimation unit that estimates a spectrum of the spectral reflectance for each of the plurality of base numbers for the filter image for which the lowest average luminance value is calculated;
The spectrum of the spectral reflectance that is known and the spectrum of the spectral reflectance for each of the estimated basis numbers are contrasted, and the estimated spectrum with respect to the spectrum of the spectral reflectance is known. An error calculating unit that calculates an error in the spectrum of the spectral reflectance for each of a plurality of base numbers;
With
The first calculation unit calculates an average luminance value of each filter image from the luminance value acquired as the imaging feature amount, and calculates the lowest average luminance value among the calculated average luminance values of each filter image And
The storage unit includes the lowest average luminance value calculated by the second calculation unit, and the largest basis number among the basis numbers whose errors calculated by the error calculation unit are within a predetermined threshold value. The estimation apparatus according to claim 3, wherein the data table is stored in association with each other. The illumination light spectrum is known,
The imaging feature amount is a pixel value of a pixel constituting the filter image,
The estimation unit sequentially acquires pixel values of pixels located at corresponding coordinates in the plurality of filter images for all the pixels constituting the plurality of filter images acquired by the imaging feature amount acquisition unit,
Using the acquired pixel value as an input, spectral reflectance is estimated for all the basis numbers of the basis functions equal to or less than the number of filters,
When the estimated spectral reflectance spectrum after the noise removal process satisfies a predetermined condition, the pixels constituting the plurality of filter images are determined based on the basis number used for estimating the spectral reflectance spectrum. 2. The estimation apparatus according to claim 1, wherein the spectral reflectance is estimated, and a filter image of a wavelength segment subdivided from the plurality of filter images is estimated from the plurality of filter images.   The estimation device according to claim 5, wherein the estimation unit performs the noise removal processing by a differential operation.   The basis number that satisfies the predetermined condition is that the spectrum of the estimated spectral reflectance is in a relationship between a predetermined first threshold value or a second threshold value and a differential value obtained by the differential operation. This is used when the number of times the differential value exceeds the first threshold is less than the predetermined number, or when the spectrum of the estimated spectral reflectance that the sum of the differential values does not exceed the predetermined second threshold is estimated. The estimation apparatus according to claim 6, wherein the base number is the same. When the illumination light spectrum is known,
An imaging feature amount acquisition unit that acquires an imaging feature amount for each of a plurality of filter images obtained by imaging a subject with a known spectral reflectance and a predetermined subject;
A spectral reflectance estimator for estimating a spectral reflectance spectrum of the predetermined subject based on the acquired imaging feature amount for each of the plurality of filter images;
An error calculator that calculates an error from the difference between the estimated spectral reflectance spectrum and the known spectral reflectance spectrum;
Based on the calculated error, a basis number of a basis function is determined, a spectral reflectance for each of the plurality of filter images is estimated based on the determined basis number, and the plurality of filter images are An estimation unit for estimating a filter image of a wavelength segment subdivided from the filter image of
An estimation device comprising: The estimation unit is
The estimation apparatus according to claim 8, wherein the error is obtained for all the basis numbers, and the basis number that minimizes the error is determined. An estimation method in an estimation device comprising an imaging feature amount acquisition unit and an estimation unit,
A first step in which the imaging feature quantity acquisition unit acquires imaging feature quantities for each of the plurality of filter images from a plurality of filter images obtained by imaging a predetermined subject;
A second step in which the estimation unit estimates a filter image of a wavelength segment subdivided from the plurality of filter images from the plurality of filter images based on the acquired imaging feature amount;
An estimation method comprising: A program for causing a computer to execute an estimation method in an estimation apparatus including an imaging feature amount acquisition unit and an estimation unit,
A first step in which the imaging feature quantity acquisition unit acquires imaging feature quantities for each of the plurality of filter images from a plurality of filter images obtained by imaging a predetermined subject;
A second step in which the estimation unit estimates a filter image of a wavelength segment subdivided from the plurality of filter images from the plurality of filter images based on the acquired imaging feature amount;
A program that causes a computer to execute.