Optical image processing method, machine learning method, trained model, machine learning pre-processing method, optical image processing module, optical image processing program, and optical image processing system
An optical image processing module includes an image acquisition unit that acquires an optical image obtained by capturing an image of light from a target object, a selection unit that selects a trained model from a plurality of trained models each of which is built in advance through machine learning using condition information including photodetector information indicating a type of photodetector, and a processing unit that inputs the optical image to the selected trained model and executes image processing of removing noise from the optical image.
1 . An optical image processing method comprising:
acquiring an optical image obtained by capturing an image of light from a target object;
selecting a trained model from a plurality of trained models each of which is built in advance through machine learning, by using imaging information relating to capturing an image of the target object;
inputting the optical image to the selected trained model and executing image processing of removing noise from the optical image; and
accepting an input of condition information indicating imaging conditions when an image of the target object is captured,
wherein the imaging information includes the condition information,
each of the plurality of trained models is a trained model built in advance through machine learning using a predetermined noise distribution model, and
selecting of the trained model includes selecting the trained model built in advance corresponding to the condition information.
2 . The optical image processing method according to claim 1 , wherein the condition information includes information indicating a type of photodetector used to capture an image of the target object,
each of the plurality of trained models is a trained model built in accordance with the type of photodetector which is capable of being input as the condition information, and
selecting of the trained model includes selecting the trained model built in advance using the predetermined noise distribution model corresponding to the type of photodetector indicated by the condition information.
3 . The optical image processing method according to claim 1 , wherein the condition information includes information indicating a gain setting value of a photodetector used to capture an image of the target object,
each of the plurality of trained models is a trained model built in accordance with a gain setting value of a photodetector which is capable of being input as the condition information, and
selecting of the trained model includes selecting the trained model built in advance corresponding to the gain setting value of the photodetector indicated by the condition information.
4 . The optical image processing method according to claim 1 , wherein the condition information includes information indicating a readout mode of a photodetector used to capture an image of the target object,
each of the plurality of trained models is a trained model built in accordance with a readout mode of a photodetector which is capable of being input as the condition information, and
selecting of the trained model includes selecting the trained model built in advance corresponding to the readout mode of the photodetector indicated by the condition information.
5 . The optical image processing method according to claim 1 , wherein the noise distribution model includes at least one of a normal distribution model and a Poisson distribution model.
6 . The optical image processing method according to claim 1 , wherein the noise distribution model includes a Bessel function distribution model.
7 . The optical image processing method according to claim 1 , further comprising:
acquiring an optical image of a structure having a predetermined structure by capturing an image of light from the structure; and
specifying image characteristics of the optical image of the structure,
wherein the imaging information includes the image characteristics,
each of the plurality of trained models is a trained model built in advance through machine learning using image data, and
selecting of the trained model includes selecting the trained model built in advance on the basis of the image characteristics.
8 . A machine learning pre-processing method in a machine learning method, the machine learning method comprising using a structure image to which noise is added based on a predetermined noise distribution model as a training image and using the training image and noise-removed image data which is data obtained by removing noise from the training image as training data to build a trained model that outputs the noise-removed image data on the basis of the training image through machine learning, the machine learning pre-processing method comprising generating a structure image to which noise is added based on the noise distribution model as the training image.
9 . A trained model built using the machine learning pre-processing method according to claim 8 , the trained model causing a processor to execute image processing of removing noise from an optical image of a target object.
10 . The machine learning pre-processing method according to claim 8 , further comprising accepting an input of condition information including photodetector information indicating a type of photodetector used to capture an image of a target object,
wherein generating of the structure image includes determining the noise distribution model to be used from the photodetector information.
11 . The machine learning pre-processing method according to claim 10 , wherein the noise distribution model includes at least one of a normal distribution model and a Poisson distribution model.
12 . The machine learning pre-processing method according to claim 10 , wherein the noise distribution model includes a Bessel function distribution model.
13 . An optical image processing module comprising a processor configured to:
acquire an optical image obtained by capturing an image of light from a target object;
select a trained model from a plurality of trained models each of which is built in advance through machine learning by using imaging information relating to capturing an image of the target object; and
input the optical image to the selected trained model and execute image processing of removing noise from the optical image,
wherein the processor accepts an input of condition information indicating imaging conditions when an image of the target object is captured,
the imaging information includes the condition information,
each of the plurality of trained models is a trained model built in advance through machine learning using a predetermined noise distribution model, and
the processor selects the trained model built in advance corresponding to the condition information.
14 . The optical image processing module according to claim 13 , wherein the condition information includes information indicating a type of photodetector used to capture an image of the target object,
each of the plurality of trained models is a trained model built in accordance with the type of photodetector which is capable of being input as the condition information, and
the processor selects the trained model built in advance using the predetermined noise distribution model corresponding to the type of photodetector indicated by the condition information.
15 . The optical image processing module according to claim 13 , wherein the condition information includes information indicating a gain setting value of a photodetector used to capture an image of the target object,
each of the plurality of trained models is a trained model built in accordance with a gain setting value of a photodetector which is capable of being input as the condition information, and
the processor selects the trained model built in advance corresponding to the gain setting value of the photodetector indicated by the condition information.
16 . The optical image processing module according to claim 13 , wherein the condition information includes information indicating a readout mode of a photodetector used to capture an image of the target object,
each of the plurality of trained models is a trained model built in accordance with a readout mode of a photodetector which is capable of being input as the condition information, and
the processor selects the trained model built in advance corresponding to the readout mode of the photodetector indicated by the condition information.
17 . The optical image processing module according to claim 13 , wherein the noise distribution model includes at least one of a normal distribution model and a Poisson distribution model.
18 . The optical image processing module according to claim 13 , wherein the noise distribution model includes a Bessel function distribution model.
19 . The optical image processing module according to claim 13 , wherein
the processor acquires an optical image of a structure having a predetermined structure obtained by capturing an image of light from the structure; and
specifies image characteristics of the optical image of the structure,
wherein the imaging information includes the image characteristics,
each of the plurality of trained models is a trained model built in advance through machine learning using image data, and
the processor selects the trained model built in advance on the basis of the image characteristics.
20 . The optical image processing module according to claim 13 , wherein the processor uses a structure image to which noise is added based on a predetermined noise distribution model as a training image and use the training image and noise-removed image data which is data obtained by removing noise from the training image as training data to build a trained model that outputs the noise-removed image data on the basis of the training image through machine learning.
21 . A non-transitory computer-readable medium causing a processor to function as:
acquiring an optical image obtained by capturing an image of light from a target object;
using imaging information relating to capturing an image of the target object to select a trained model from a plurality of trained models each of which is built in advance through machine learning;
inputting the optical image to the selected trained model and executing image processing of removing noise from the optical image;
accepting an input of condition information indicating imaging conditions when an image of the target object is captured, the imaging information including the condition information, and each of the plurality of trained models is a trained model built in advance through machine learning using a predetermined noise distribution model; and
selecting the trained model built in advance corresponding to the condition information.
22 . An optical image processing system comprising:
the optical image processing module according to claim 13 ; and
an imaging device configured to acquire the optical image by capturing an image of light from the target object.