Image processing apparatus, image processing method and program
It aims to accurately identify a class concerning classification of an object for each area of an image. A plurality of subsets are created from a plurality of learning images, and an area discriminator for performing area discrimination of the learning images is learned for each subset created. Then, area discrimination of a plurality of learning evaluation images is performed by using the plurality of learned area discriminators. Further, an integrated identifier is learned based on an area discrimination result performed to the plurality of learning evaluation images and scene information previously associated with each of the plurality of learning evaluation images.
1. An image processing apparatus comprising:
a processor;
and a memory storing instructions which, when executed by the processor, cause the image processing apparatus to function as:
a creating unit configured to create a plurality of subsets, wherein each of the plurality of subsets is created by selecting, from a plurality of learning images, one or more learning images;
a learning unit configured to learn an area discriminator for discriminating a class of each area obtained by dividing the learning images, for each subset created by the creating unit;
a learning discriminating unit configured to discriminate a class of each area obtained by dividing a plurality of learning evaluation images, by using the plurality of area discriminators learned by the learning unit; and
an identification learning unit configured to learn an integrated identifier based on a discrimination result for each area of the plurality of learning evaluation images discriminated by the learning discriminating unit and scene information associated with each of the plurality of learning evaluation images.
2. The image processing apparatus according to claim 1 , wherein
scene information is associated with each of the plurality of learning images, and
the creating unit creates the plurality of subsets based on the scene information of the plurality of learning images.
3. The image processing apparatus according to claim 2 , wherein
the creating unit creates, based on the scene information of the learning image randomly selected from the plurality of learning images, the subset by selecting the learning image with which scene information similar to the scene information of the learning image selected from the plurality of learning images is associated.
4. The image processing apparatus according to claim 2 , wherein
the creating unit creates, based on the scene information of the learning image selected from the plurality of learning images by an instruction of a user, the subset by selecting the learning image with which scene information similar to the scene information of the learning image selected from the plurality of learning images is associated.
5. The image processing apparatus according to claim 2 , wherein
the creating unit performs clustering of the plurality of learning images based on similarity of the scene information, and creates a same subset by the learning images included in a cluster by the clustering.
6. The image processing apparatus according to claim 1 , wherein
the plurality of learning evaluation images include a part or all of the plurality of learning images.
7. The image processing apparatus according to claim 2 , wherein
the identification learning unit learns the integrated identifier based on similarity between the scene information of the plurality of learning images and the scene information of the plurality of learning evaluation images.
8. The image processing apparatus according to claim 1 , wherein the instructions, when executed by the processor, further cause the image processing apparatus to function as:
a re-learning unit configured to re-learn the plurality of area discriminators according to accuracy of the discrimination result by the learning discriminating unit.
9. The image processing apparatus according to claim 1 , wherein the instructions, when executed by the processor, further cause the image processing apparatus to function as:
an additional learning unit configured to additionally learn the area discriminator according to the discrimination result by the learning discriminating unit, wherein
the identification learning unit learns the integrated identifier based on a discrimination result of the area discriminator additionally learned by the additional learning unit.
10. The image processing apparatus according to claim 1 , wherein
an importance is associated with each of the plurality of learning images, and
the creating unit creates the plurality of subsets based on the respective importances of the plurality of learning images.
11. The image processing apparatus according to claim 1 , wherein the instructions, when executed by the processor, further cause the image processing apparatus to function as:
a holding unit configured to hold the plurality of area discriminators learned by the learning unit.
12. The image processing apparatus according to claim 1 , wherein
the scene information includes a histogram of color information in a plurality of blocks obtained by dividing an image.
13. The image processing apparatus according to claim 1 , wherein
the learning unit learns the area discriminator for discriminating the class of each area based on a feature amount extracted from each area of the learning image and accuracy data given to the each area.
14. The image processing apparatus according to claim 1 , wherein
the identification learning unit learns the integrated identifier based on a feature amount constituted by the discrimination result and the scene information, and accuracy data given to each area of the plurality of learning evaluation images.
15. The image processing apparatus according to claim 1 , wherein the instructions, when executed by the processor, further cause the image processing apparatus to function as:
an obtaining unit configured to an image to be identified;
a scene information obtaining unit configured to obtain scene information of the image to be identified;
a discriminating unit configured to discriminate a class of each area obtained by dividing the image to be identified, by using the plurality of area discriminators learned by the learning unit; and
an integrated identifying unit configured to identify, by using the learned integrated identifier, the class of each area of the image to be identified, based on a discrimination result for each area of the plurality of images to be identified discriminated by the discriminating unit and the obtained scene information.
16. An image processing method comprising:
creating a plurality of subsets, wherein each of the plurality of subsets is created by selecting, from a plurality of learning images, one or more learning images;
learning an area discriminator for discriminating a class of each area obtained by dividing the learning images, for each of the created subsets;
discriminating a class of each area obtained by dividing a plurality of learning evaluation images, by using the plurality of learned area discriminators; and
learning an integrated identifier based on a discrimination result for each area of the plurality of discriminated learning evaluation images and scene information associated with each of the plurality of learning evaluation images.
17. A non-transitory computer-readable recording medium that stores a program for causing a computer to function as the units of an image processing apparatus comprising:
a creating unit configured to create a plurality of subsets, wherein each of the plurality of subsets is created by selecting, from a plurality of learning images, one or more learning images;
a learning unit configured to learn an area discriminator for discriminating a class of each area obtained by dividing the learning images, for each subset created by the creating unit;
a learning discriminating unit configured to discriminate a class of each area obtained by dividing a plurality of learning evaluation images, by using the plurality of area discriminators learned by the learning unit; and
an identification learning unit configured to learn an integrated identifier based on a discrimination result for each area of the plurality of learning evaluation images discriminated by the learning discriminating unit and scene information associated with each of the plurality of learning evaluation images.