2. The image processing device according to claim 1, wherein the weight calculation means comprising: gradient calculation means for calculating a gradient of a feature value of a pixel of the input image; direction calculation means for calculating a degree of equality in directions of gradient of the feature values; and weight data calculation means for calculating the weight based on the gradient and a degree of equality in the directions.

3. The image processing device according to claim 1 or 2, wherein, the weight data calculation means, in an area where magnitude of the gradient is more than a predetermined value, and a degree of equality in directions of the gradient is less than a predetermined value, calculates a weight reduced by a predetermined value.

4. The image processing device according to any one of claims 1 to 3, comprising: a direction constraint calculation means for calculating a direction constraint with which an edge direction perpendicular to a gradient direction of a feature value of the input image is smooth in the restored image, wherein the image restoring means restores an image by using the direction constraint in addition to the reconstruction constraint and the regularization constraint. 48 •

5. The image processing device according to any one of claims 1 to 4, further comprising: · a reference pixel pruning means for imposing a limitation so that some of pixels are not used in calculation of the regularization constraint by the regularization term calculation means.

6. The image processing device according to claim 5, wherein the reference pixel pruning means deletes a part of weights calculated by the weight calculation means.

7. An image processing method comprising: determining an area where a feature value of an input image is saved, based on a gradient of a feature value of a pixel of the input image and a direction of the gradient, and calculating aweight for reducing a regularization constraint that is a constraint based on regularization of image processing in the area where the feature value is saved; calculating a regularization constraint of a high resolution image restored based on the input image by using the weight; calculating a reconstruction constraint that is a constraint based on reconstruction of the high resolution image; and restoring the high resolution image from the input image based on the regularization constraint and the reconstruction constraint.

8. A computer-readable recording medium recording a program for causing a computer to execute: processing of determining an area where a feature value of an input image is saved, based on a gradient of a feature value of a pixel of the input image and a direction of the gradient, and calculating a weight for reducing a regularization constraint that is a constraint based on regularization of image processing in the area where the feature value is saved; processing of calculating a regularization constraint of a high resolution image restored based on the input image by using the weight; processing of calculating a reconstruction constraint that is a constraint based on reconstruction of the high resolution image; and processing of restoring the high resolution image from the input image based on the regularization constraint and the reconstruction constraint. 49 •

9. An image processing device comprising: weight calculation means for calculating a weight for reducing a regularization constraint that is a constraint based on regularization of image processing in a texture area of an input image; regularization term calculation means for calculating the regularization constraint by using the weight in the texture area and not using the weight in an area other than the texture area; and image restoring means for restoring a high resolution image based on the regularization constraint and a reconstruction constraint that is a constraint based on reconstruction of the high resolution image.

10. An image processing method comprising: calculating a weight for reducing a regularization constraint that is a constraint based on regularization of image processing in a texture area of an input image; calculating the regularization constraint by using the weight in the texture area and not using the weight in an area other than the texture area; and restoring a high resolution image based on the regularization constraint and a reconstruction constraint that is a constraint based on reconstruction of the high resolution image.

11. A computer-readable recording medium recording a program for causing a computer to execute: processing of calculating a weight for reducing a: regularization constraint that is a constraint based on regularization of image processing in a texture area of an input image; processing of calculating the regularization constraint by using the weight in the texture area and not using the weight in an area other than the texture area; and processing of restoring a high resolution image based on the regularization constraint and a reconstruction constraint that is a constraint based on reconstruction of the high resolution image.

For example, the reference pixel pruning unit 70 may include a
mask corresponding to a pixel determined in advance, and apply the mask
to the weight. In this case, the regularization term calculation unit 31
does not calculate a regularization constraint for a pixel corresponding to
the masked weight.
[0188]
The reference pixel pruning unit 70 may randomly select a
component of the weight, and set the value of the component to "0". For·
example, when the weight calculation unit 50 calculates the weight
indicated in the numerical expression 7, the reference pixel pruning unit 70
randomly selects a part of the components of the vector A or the weight
matrix (diag (A)), and set the value thereof to "0".
[0189]
Then, when the reference pixel pruning unit 70 sets the component
to "0" as descrihed above, the regularization term calculation unit 31 does
not calculate a difference of the component of which value is "0".
[0190]
As described above, the regularization term calculation unit 31
reduces the calculation processing of the difference based on the result of
the reference pixel pruning unit 70.
[0191]
The reference pixel pruning unit 70 may not operate the weight, and
may notify the regularization term calculation unit 3 I of information about
pruning (for example, a flag) as information different from the weight.
In this case, like the first exemplary embodiment, the regularization term
calculation unit 31 may receive the weight from the weight calculation unit
50, and may reduce pixels for which the regularization constraint is
calculated, based on the pruning information from the reference pixel

We Claim:
I.· An image processing device comprising:
weight calculation means for determining an area where a feature
value of an input image is saved, based on a gradient of a feature value of a
pixel of the input image and a direction of the gradient, and calculating a
weight for reducing a regularization constraint that is a constraint based on
regularization of image processing in the area where the feature value is
saved;
regularization term calculation means for calculating a
regularization constraint of a high resolution image restored based on the
input image by using the weight;
reconstruction constraint calculation means for calculating a
reconstruction constraint that is a constraint based on reconstruction of the
high resolution image; and
image restoring means for restoring the high resolution image from
the input image based on the regularization constraint and the
reconstruction constraint.
2. The image processing device according to claim 1, wherein
the weight calculation means comprising:
gradient calculation means for calculating a gradient of a feature
value of a pixel of the input image;
direction calculation means for calculating a degree of equality in
directions of gradient of the feature values; and
weight data calculation means for calculating the weight based on
the gradient and a degree of equality in the directions.
3. The image processing device according to claim 1 or 2,
wherein,
the weight data calculation means, in an area where magnitude of
the gradient is more than a predetermined value, and a degree of equality in
directions of the gradient is less than a predetermined value, calculates a
weight reduced by a predetermined value.
4. The image processing device according to any one of claims 1
to 3, comprising:
a direction constraint calculation means for calculating a direction
constraint with which an edge direction perpendicular to a gradient
direction of a feature value of the input image is smooth in the restored
image, wherein
the image restoring means restores an image by using the direction
constraint in addition to the reconstruction constraint and the regularization
constraint.
48
•
5. The image processing device according to any one of claims 1
to 4, further comprising:
· a reference pixel pruning means for imposing a limitation so that
some of pixels are not used in calculation of the regularization constraint by
the regularization term calculation means.
6. The image processing device according to claim 5, wherein
the reference pixel pruning means deletes a part of weights
calculated by the weight calculation means.
7. An image processing method comprising:
determining an area where a feature value of an input image is
saved, based on a gradient of a feature value of a pixel of the input image
and a direction of the gradient, and calculating aweight for reducing a
regularization constraint that is a constraint based on regularization of
image processing in the area where the feature value is saved;
calculating a regularization constraint of a high resolution image
restored based on the input image by using the weight;
calculating a reconstruction constraint that is a constraint based on
reconstruction of the high resolution image; and
restoring the high resolution image from the input image based on
the regularization constraint and the reconstruction constraint.
8. A computer-readable recording medium recording a program
for causing a computer to execute:
processing of determining an area where a feature value of an input
image is saved, based on a gradient of a feature value of a pixel of the
input image and a direction of the gradient, and calculating a weight for
reducing a regularization constraint that is a constraint based on
regularization of image processing in the area where the feature value is
saved;
processing of calculating a regularization constraint of a high
resolution image restored based on the input image by using the weight;
processing of calculating a reconstruction constraint that is a
constraint based on reconstruction of the high resolution image; and
processing of restoring the high resolution image from the input
image based on the regularization constraint and the reconstruction
constraint.
49
•
9. An image processing device comprising:
weight calculation means for calculating a weight for reducing a
regularization constraint that is a constraint based on regularization of
image processing in a texture area of an input image;
regularization term calculation means for calculating the
regularization constraint by using the weight in the texture area and not
using the weight in an area other than the texture area; and
image restoring means for restoring a high resolution image based
on the regularization constraint and a reconstruction constraint that is a
constraint based on reconstruction of the high resolution image.
10. An image processing method comprising:
calculating a weight for reducing a regularization constraint that is
a constraint based on regularization of image processing in a texture area
of an input image;
calculating the regularization constraint by using the weight in the
texture area and not using the weight in an area other than the texture area;
and
restoring a high resolution image based on the regularization
constraint and a reconstruction constraint that is a constraint based on
reconstruction of the high resolution image.
11. A computer-readable recording medium recording a program
for causing a computer to execute:
processing of calculating a weight for reducing a: regularization
constraint that is a constraint based on regularization of image processing
in a texture area of an input image;
processing of calculating the regularization constraint by using the
weight in the texture area and not using the weight in an area other than the
texture area; and
processing of restoring a high resolution image based on the
regularization constraint and a reconstruction constraint that is a constraint
based on reconstruction of the high resolution image.