Image restoration apparatus, image restoration method, image restoration program, restorer generation apparatus, restorer generation method, restorer generation program, determiner generation apparatus, determiner generation method, determiner generation program, article determination apparatus, article determination method, and article determination program
An image restoration apparatus includes: an image acquisition unit that acquires a mask image generated by masking, by a mask region, a predetermined region of an article image in which an article is imaged; and an image output unit that outputs a restored image corresponding to the mask image acquired by the image acquisition unit by using a restorer pre-trained by machine learning so as to output the restored image, which is restored from the mask image to reproduce the article image in a pseudo manner, according to an input of the mask image.
1. An image restoration apparatus comprising:
a processor and/or hardware circuitry configured to implement:
an image acquisition unit that acquires a mask image generated by masking, by a mask region, a predetermined region of an article image in which an article is imaged; and
an image output unit that outputs a restored image corresponding to the mask image acquired by the image acquisition unit by using a restorer pre-trained by machine learning so as to output the restored image, which is restored from the mask image to reproduce the article image in a pseudo manner, according to an input of the mask image.
2. The image restoration apparatus according to claim 1 , wherein
the image acquisition unit acquires, as the mask image, a non-defective article mask image generated by masking, as the predetermined region, a defect candidate region, which is predetermined so as to correspond to a portion of the article in which a defect frequently occurs during manufacturing, in a non-defective article image that is the article image in which the article including no defect is imaged, and
the image output unit outputs a pseudo-defective article image corresponding to the non-defective article mask image acquired by the image acquisition unit by using the restorer, which is pre-trained by the machine learning based on a defective article image, which is the article image in which the article including the defect is imaged, and a defective article mask image as the mask image generated by masking a defective region, as the predetermined region, corresponding to the defect in the defective article image so as to output, as the restored image, the pseudo-defective article image, which is restored to reproduce the defective article image in a pseudo manner.
3. The image restoration apparatus according to claim 1 , wherein
the image acquisition unit increases the number of the mask images by executing a first image process including at least expansion or contraction on the mask region.
4. The image restoration apparatus according to claim 1 , wherein
the image acquisition unit increases the number of the mask images by executing a second image process including one or more of noise addition, gain adjustment, contrast adjustment, and averaging on a region other than the mask region in the mask image.
5. The image restoration apparatus according to claim 1 , wherein
the image acquisition unit acquires the mask image by synthesizing the article image and a template image including the mask region at a predetermined position.
6. An image restoration method comprising:
an image acquisition step of acquiring a mask image generated by masking, by a mask region, a predetermined region of an article image in which an article is imaged; and
an image output step of outputting a restored image corresponding to the mask image acquired by the image acquisition step by using a restorer pre-trained by machine learning so as to output the restored image, which is restored from the mask image, according to an input of the mask image to reproduce the article image in a pseudo manner.
7. A non-transitory computer readable storage medium storing an image restoration program for causing a computer to execute:
an image acquisition step of acquiring a mask image generated by masking, by a mask region, a predetermined region of an article image in which an article is imaged; and
an image output step of outputting a restored image corresponding to the mask image acquired by the image acquisition step by using a restorer pre-trained by machine learning so as to output the restored image, which is restored from the mask image, according to an input of the mask image to reproduce the article image in a pseudo manner.