Image processing device and image processing method
View Patent ↗In order to provide an image processing device and the like making it possible to generate a target image in which edges of a structure are upheld and from which streaking artifacts are removed, a computation device determines a shape of a non-linear function on the basis of feature amounts of an original image and a smoothed image (S 101 ). Next, the computation device calculates a condition coefficient of the original image and the smoothed image by using the non-linear function for which the shape was determined in S 101 (S 102 ). Next, the computation device uses the condition coefficients calculated in S 102 to calculate a weighting coefficient for each of the pixels of the original image and the smoothed image (S 103 ). Next, the computation device adds weighting to the original image and the smoothed image to generate the target image (S 104 ).
1. An image processing device for adding respective mutually corresponding pixels of an original image, including streak artifact, generated from original information and of a smoothed image generated from the same original information and reduced at least in the streak artifact with weighting using a weighting coefficient and generating a target image that an edge of a structure is maintained and the streak artifact is removed, comprising:
a function shape determination unit that determines a shape of a nonlinear function on the basis of feature amounts of the original image and the smoothed image; and
a weighting coefficient calculation unit that calculates the weighing coefficient on the basis of the nonlinear function,
wherein the function shape determination unit includes
a division unit that divides regions of the original image and the smoothed image into plural mutually corresponding small regions,
a decreasing rate calculation unit that calculates variation values for the small regions pertaining to both of the original image and the smooth image and calculates a decreasing rate of the variation value of the smoothed image based on the variation value of the original image, for each of the small regions,
a region extraction unit that extracts a feature amount calculation region from a set of the small regions on the basis of the decreasing rate, and
a feature amount calculation unit that calculates the feature amount from a pixel value of the feature amount calculation region.
2. The image processing device according to claim 1 , wherein
the region extraction unit extracts the small region the decreasing rate of which is maximum within a predetermined range as the feature amount calculation region.
3. The image processing device according to claim 1 , wherein
the region extraction unit extracts the small region the variation value of which is maximum in the small regions counted from the small region the decreasing rate of which is maximum down to the small region ranked to a fixed order within a predetermined range as the feature amount calculation region.
4. The image processing device according to claim 1 , wherein
the nonlinear function is a smooth and continuous function that a difference in pixel value between a pixel of interest of the original image or the smoothed image and a pixel adjacent to the pixel of interest is set as a variable.
5. The image processing device according to claim 4 , wherein
the nonlinear function is a function which approaches 1 as the difference in pixel value is decreased and approaches 0 as the difference in pixel value is increased.
6. The image processing device according to claim 1 , wherein
the weighting coefficient calculation unit calculates condition coefficients indicating respective conditions of the original image and the smoothed image by using the nonlinear function and calculates the weighting coefficient on the basis of the condition coefficients.
7. An image processing method of adding respective mutually corresponding pixels of an original image generated from original information and of a smoothed image generated from the same original information and reduced at least in streak artifact with weighting using a weighting coefficient and generating a target image that an edge of a structure is maintained and the streak artifact is removed, comprising:
the function shape determination step of determining a shape of a nonlinear function on the basis of feature amounts of the original image and the smoothed image; and
the weighting coefficient calculation step of calculating the weighting coefficient on the basis of the nonlinear function,
wherein the function shape determination step includes
a division step of dividing regions of the original image and the smoothed image into plural mutually corresponding small regions,
a decreasing rate calculation step of calculating variation values for the small regions pertaining to both of the original image and the smooth image and calculating a decreasing rate of the variation value of the smoothed image based on the variation value of the original image, for each of the small regions,
a region extraction step of extracting a feature amount calculation region from a set of the small regions on the basis of the decreasing rate, and
a feature amount calculation step of calculating the feature amount from a pixel value of the feature amount calculation region.
8. The image processing method according to claim 7 , wherein
the region extraction step is adapted to extract the small region the decreasing rate of which is maximum within a predetermined range as the feature amount calculation region.
9. The image processing method according to claim 7 , wherein
the region extraction step is adapted to extract the small region the variation value of which is maximum within the small regions counted from the small region the decreasing rate of which is maximum down to the small region ranked to a fixed order within a predetermined range as the feature amount calculation region.
10. The image processing method according to claim 7 , wherein
the nonlinear function is a smooth and continuous function that a difference in pixel value between a pixel of interest of the original image or the smoothed image and a pixel adjacent to the pixel of interest is set as a variable.
11. The image processing method according to claim 10 , wherein the nonlinear function is a function which approaches 1 as the difference in pixel value is decreased and approaches 0 as the difference in pixel value is increased.
12. The image processing method according to claim 7 , wherein
the weighting coefficient calculation step is adapted to calculate condition coefficients indicating respective conditions of the original image and the smoothed image by using the nonlinear function and calculate the weighting coefficient on the basis of the condition coefficients.
13. The image processing method according to claim 7 , wherein the nonlinear function is defined by a condition coefficient g s (x) of a pixel of interest s, calculated based on the following formula (1), in which x ={xl, . . . , xj} is an image vector, N s is a set of adjacent pixels r relative to the pixel of interest s, p is an arbitrary parameter for adjusting the gradient of the generalized Gaussian function, ν is an arbitrary parameter for adjusting a bending position of the generalized Gaussian function, W sr is a weighting coefficient according to a distance between the pixel of interest s and the adjacent pixel r, and wherein the shape of the nonlinear function is determined by the value of ν
g
s
(
x
,
v
)
≡
exp
{
-
∑
r
∈
N
s
w
sr
x
s
-
x
r
p
2
v
p
}
.
(
1
)
14. The image processing device according to claim 1 , wherein the nonlinear function is defined by a condition coefficient g s (x) of a pixel of interest s, calculated based on the following formula (1), in which x ={xl, . . . , xj} is an image vector, N s is a set of adjacent pixels r relative to the pixel of interest s, p is an arbitrary parameter for adjusting the gradient of the generalized Gaussian function, ν is an arbitrary parameter for adjusting a bending position of the generalized Gaussian function, W sr is a weighting coefficient according to a distance between the pixel of interest s and the adjacent pixel r, and wherein the shape of the nonlinear function is determined by the value of ν
g
s
(
x
,
v
)
≡
exp
{
-
∑
r
∈
N
s
w
sr
x
s
-
x
r
p
2
v
p
}
.
(
1
)