Information processing device, information processing method, and recording medium

1. An information processing device comprising:

a Central Processing unit (CPU); and

a memory connected to the CPU and storing a program, wherein

the CPU reads the program from the memory and configures:

a variation-amount calculating unit that, for an input image as a target to be processed, calculates a variation amount between a value of a predetermined pixel of the input image and values of peripheral pixels of the predetermined pixel;

an attribute reliability calculating unit that, based on an attribute that is a property of pixels within a specified area in the input image and the variation amount, calculates attribute reliability that is reliability of the attribute of the pixel of the input image;

a regularization strength estimating unit that, based on image quality information that is information about image quality for the attribute and the attribute reliability, estimates a regularization strength of the pixel in the input image; and

an image reconstructing unit generates a reconstructed image that is an image acquired by reconstructing the input image by using the regularization strength;

wherein

for each pixel, the regularization strength estimating unit estimates the regularization strength such that the higher the attribute reliability is, the more the pixel conforms to the image quality information for the attribute, and estimates the regularization strength such that the lower the attribute reliability is, the more the pixel is apart from the image quality information for the attribute.

2. The information processing device according to claim 1 ,

wherein

the image reconstructing unit generates the reconstructed image based on a regularization term representing a strength of regularization for each pixel based on the regularization strength and an error term representing a difference between pixels of a degraded image that is an image acquired by deteriorating the reconstructed image and the input image.

3. The information processing device according to claim 1 ,

wherein

the attribute reliability calculating unit calculates reliability of the attribute, based on area information that is information of an area, for which the attribute is specified.

4. The information processing device according to claim 1 , the CPU further configures:

an image receiving unit that receives the input image;

an image outputting unit that transmits the reconstructed image; and

a training information receiving unit that receives training information including information of the attribute and the image quality information,

wherein

the training information receiving unit receives the regularization strength of a pixel in a predetermined area of the reconstructed image that is output by the image outputting unit, as the image quality information.

5. The information processing device according to claim 4 ,

wherein

the image reconstructing unit generates a plurality of reconstructed images,

the image outputting unit transmits the plurality of reconstructed images, and

the training information receiving unit receives the regularization strength of pixels in a plurality of areas in the plurality of reconstructed images that are transmitted by the image outputting unit as the image quality information.

6. The information processing device according to claim 4 , wherein

the training information receiving unit further including:

an area specifying unit that specifies an area in the reconstructed image, for which the regularization strength of the pixels is specified; and

a training information acquiring unit that acquires the attribute and the regularization strength of the pixels of the area as training information of the specified area.

7. The information processing device according to claim 1 , the CPU further configures:

a learning image receiving unit that receives a learning image that is an image different from the input image; and

a learning image training information receiving unit that receives second training information relating to the learning image,

wherein

the variation amount calculating unit calculates the variation amount based on the learning image; and

the attribute reliability calculating unit calculates the attribute reliability by using the second training information for the learning image.

8. An information processing method comprising:

for an input image as a target to be processed, calculating a variation amount between a value of a predetermined pixel of the input image and values of peripheral pixels of the predetermined pixel;

based on an attribute that is a property of pixels within a specified area in the input image and the variation amount, calculating attribute reliability that is reliability of the attribute of the pixel of the input image;

based on image quality information that is information about image quality for the attribute and the attribute reliability, estimating a regularization strength of the pixel in the input image:

generating a reconstructed image that is an image acquired by reconstructing the input image by using the regularization strength; and

for each pixel, estimating the regularization strength such that the higher the attribute reliability is, the more the pixel conforms to the image quality information for the attribute, and estimates the regularization strength such that the lower the attribute reliability is, the more the pixel is apart from the image quality information for the attribute.

9. The information processing method according to claim 8 , further comprising:

generating the reconstructed image based on a regularization term representing a strength of regularization for each pixel based on the regularization strength and an error term representing a difference between pixels of a degraded image that is an image acquired by deteriorating the reconstructed image and the input image.

10. The information processing method according to claim 8 , further comprising:

calculating reliability of the attribute, based on area information that is information of an area, for which the attribute is specified.

11. The information processing method according to claim 8 , further comprising:

receiving the input image;

transmitting the reconstructed image; and

receiving training information including information of the attribute and the image quality information, and

receiving the regularization strength of a pixel in a predetermined area of the reconstructed image that is output, as the image quality information.

12. A non-transitory computer-readable recording medium embodying a program, the program causing a computer to perform a method, the method comprising:

for an input image as a target to be processed, calculating a variation amount between a value of a predetermined pixel of the input image and values of peripheral pixels of the predetermined pixel;

based on an attribute that is a property of pixels within a specified area in the input image and the variation amount, calculating attribute reliability that is reliability of the attribute of the pixel of the input image;

based on image quality information that is information about image quality for the attribute and the attribute reliability, estimating a regularization strength of the pixel in the input image;

generating a reconstructed image that is an image acquired by reconstructing the input image by using the regularization strength; and

for each pixel, estimating the regularization strength such that the higher the attribute reliability is, the more the pixel conforms to the image quality information for the attribute, and estimates the regularization strength such that the lower the attribute reliability is, the more the pixel is apart from the image quality information for the attribute.

13. The medium according to claim 12 , further embodying the program to perform the method, the method further comprising:

generating the reconstructed image based on a regularization term representing a strength of regularization for each pixel based on the regularization strength and an error term representing a difference between pixels of a degraded image that is an image acquired by deteriorating the reconstructed image and the input image.

14. The medium according to claim 12 , further embodying the program to perform the method, the method further comprising:

calculating reliability of the attribute, based on area information that is information of an area, for which the attribute is specified.

15. The medium according to claim 12 , further embodying the program to perform the method, the method further comprising:

receiving the input image;

transmitting the reconstructed image; and

receiving training information including information of the attribute and the image quality information, and

receiving the regularization strength of a pixel in a predetermined area of the reconstructed image that is output, as the image quality information.