Image analysis method, learning image or analysis image generation method, learned model generation method, image analysis apparatus, and image analysis program
An example apparatus comprising an image acquisition unit configured to acquire a plurality of images having temporal or spatial continuity; a channel assignment unit configured to assign a channel different from each other to at least a part of gradation information on a color and/or gradation information on brightness that can be acquired from each image of the plurality of images based on a predetermined rule; a composite image generation unit configured to generate one composite image in which gradation information on at least a part of each image can be identified by the channel by extracting and combining gradation information to which the channel is assigned from each of the plurality of images; and an inference unit configured to analyze the composite image and to infer the plurality of images.
1 . An image analysis method comprising:
(a) acquiring a plurality of images having temporal or spatial continuity;
(b) assigning a channel different from each other to gradation information corresponding to a color having a hue different from each other, for each image of the plurality of images;
(c) generating one color composite image in which gradation information on at least a part of each image can be identified by combining pieces of gradation information corresponding to colors having different hues by extracting the gradation information to which the channel is assigned from each of the plurality of images and combining the extracted gradation information; and
(d) analyzing the color composite image and inferring the plurality of images.
2 . The image analysis method according to claim 1 , wherein (d) analyzing the color composite image and inferring the plurality of images includes:
inputting the generated color composite image to a learned model on which machine learning is performed in advance based on a plurality of color composite images generated from a plurality of sample images used for learning, and
obtaining an output of the learned model as an inference result.
3 . The image analysis method according to claim 1 ,
wherein the plurality of images are a plurality of images extracted at regular time intervals from a moving image including a plurality of images having temporal continuity obtained by shooting a moving image of a predetermined moving object, and
wherein (d) analyzing the color composite image and inferring the plurality of images includes executing an inference related to a motion pattern of the predetermined moving object from the color composite image generated based on the moving image.
4 . The image analysis method according to claim 3 , further comprising
(e) acquiring positional information on the predetermined moving object at a time of acquiring the plurality of images, and
wherein when a plurality of objects is to be analyzed,
(a) acquiring a plurality of images having temporal or spatial continuity includes acquiring the plurality of images for each of the plurality of objects,
(e) acquiring positional information on the object at a time of acquiring a plurality of images includes acquiring the positional information on each of the plurality of objects,
(b) assigning a channel different from each other to gradation information on a color having a hue different from each other includes assigning a channel to each of a plurality of acquired images for each object,
(c) generating one color composite image in which gradation information on at least a part of each image can be identified includes generating the color composite image for each of the objects, and
(d) analyzing the color composite image and inferring the plurality of images includes executing an inference related to motion patterns of the plurality of objects with a plurality of the composite images generated for each of the objects and the positional information on each of the plurality of objects as inputs.
5 . The image analysis method according to claim 1 ,
wherein the plurality of images are a plurality of tomographic images having continuity in a specific direction when a three-dimensional region is represented by stacking a plurality of acquired tomographic images in the specific direction, or a plurality of tomographic images extracted from a three-dimensional model so as to have continuity in a specific direction when a three-dimensional region is represented by a three-dimensional model from which a tomographic image can be optionally extracted, and
wherein (d) analyzing the color composite image and inferring the plurality of images includes executing an inference related to the three-dimensional region from the color composite image.
6 . An image generation method comprising:
(a) acquiring a plurality of images having temporal or spatial continuity;
(b) assigning a channel different from each other to gradation information corresponding to a color having a hue different from each other, for each image of the plurality of images; and
(c) generating one color composite image in which gradation information on at least a part of each image can be identified by combining pieces of gradation information corresponding to colors having different hues by extracting the gradation information to which the channel is assigned from each image of the plurality of images and combining the extracted gradation information.
7 . The image generation method according to claim 6 , further comprising:
(d) acquiring correct answer data when the inference on the color composite image is executed;
(e) inputting the color composite image to a model including a neural network, causing the inference to be executed, and causing an inference result to be output; and
(f) updating a parameter of the model using the inference result and the correct answer data.
8 . An image analysis apparatus comprising a hardware processor that:
acquires a plurality of images having temporal or spatial continuity;
assigns a channel different from each other to gradation information corresponding to a color having a hue different from each other, for from each image of the plurality of images;
generates one color composite image in which gradation information on at least a part of each image can be identified by combining pieces of gradation information corresponding to colors having different hues by extracting the gradation information to which the channel is assigned from each of the plurality of images and combining the extracted gradation information; and
analyzes the color composite image and infers the plurality of images.
9 . A non-transitory recording medium storing a computer readable program, the program causing a computer to implement:
(a) acquiring a plurality of images having temporal or spatial continuity;
(b) assigning a channel different from each other to gradation information corresponding to a color having a hue different from each other, for each image of the plurality of images;
(c) generating one color composite image in which gradation information on at least a part of each image can be identified by combining pieces of gradation information corresponding to colors having different hues by extracting the gradation information to which the channel is assigned from each of the plurality of images and combining the extracted gradation information; and
(d) analyzing the color composite image and inferring the plurality of images.