US 9,177,366 B2 · Edge-preserving noise filtering

Abstract

The invention relates to an image processing apparatus (1) comprising an image providing unit (2) for providing a first image of an object and a second image of the same object and a filtering unit (3) which filters the first image depending on a first degree of similarity between image values of the first image and on a second degree of similarity between image values of the second image. This allows filtering the first image depending on the likeliness that image values belong to the same part of the object, for example, to tissue or to bone material of the object, if the object is a human being or an animal, even if due to noise the image values of one of the first image and the second image are disturbed, thereby improving, for example, an edge preserving property of the filtering procedure.

An image processing apparatus for processing an image, the image processing apparatus comprising:

an image providing unit, executed by a microprocessor, that provides a first image of an object and a second image of the same object,
a filtering unit, executed by the microprocessor, that filters the first image, wherein the filtering unit filters the first image depending on a first degree of similarity between image values of the first image and on a second degree of similarity between image values of the second image,
wherein the filtering unit weightedly averages image values of image elements of the first image by determining an average image value for a first image element of the first image through:
u
^
(
1
)
⁡
(
x
→
i
)
=
∑
x
→
j
∈
Δ
i
⁢
L
1
(
u
→
(
1
)
⁡
(
x
→
i
)
-
u
→
(
1
)
⁢
(
x
→
j
)
⁢
)
L
2
⁡
(
u
→
(
2
)
⁡
(
x
→
i
)
-
u
→
(
2
)
⁡
(
x
→
j
)
)
⁢
K
⁡
(
x
→
i
-
x
→
j
)
⁢
u
(
1
)
⁡
(
x
→
j
)
∑
x
→
j
∈
Δ
i
⁢
L
1
⁡
(
u
→
(
1
)
⁡
(
x
→
i
)
-
u
→
(
1
)
⁢
(
x
→
j
)
)
⁢
L
2
⁡
(
u
→
(
2
)
⁡
(
x
→
i
)
-
u
→
(
2
)
⁡
(
x
→
j
)
)
⁢
K
⁡
(
x
→
i
-
x
→
j
)
wherein û(1)({right arrow over (x)}i) denotes the average image value for the first image element at a position {right arrow over (x)}i of the first image, Δi denotes a first region which comprises the first image element at the position {right arrow over (x)}i, and {right arrow over (x)}j denotes positions of first region image elements within the first region Δi,
wherein L1({right arrow over (u)}(1)({right arrow over (x)}i)−{right arrow over (u)}(1)({right arrow over (x)}j)) denotes the first degree of similarity between a) the image value of the first region image element {right arrow over (x)}j and image values of image elements surrounding the respective first region element at the position {right arrow over (x)}j and b) the image value of the first image element at the position {right arrow over (x)}i and image values of image elements surrounding the first image element of the first image;
wherein L2({right arrow over (u)}(2)({right arrow over (x)}i)−{right arrow over (u)}(2)({right arrow over (x)}j)) denotes a second degree of similarity between a) the image value of the second region image element at the position {right arrow over (x)}j within a second region, which corresponds to the first region, of the second image and optionally image values of image elements surrounding the second region element at the position {right arrow over (x)}j, and b) the image value of a second image element at the position {right arrow over (x)}i, which corresponds to the first image element at the same position in the first image, of the second image and optionally image values of image elements surrounding the second image element;
wherein K({right arrow over (x)}i−{right arrow over (x)}j) denotes a distance function which depends on distances between the
image elements at the positions {right arrow over (x)}i and {right arrow over (x)}j;
wherein {right arrow over (u)}(1)({right arrow over (x)}j) denotes an unfiltered image value at the position {right arrow over (x)}j in the first image;
and
wherein a k-th degree of similarity is based on:
Lk({right arrow over (u)}(k)({right arrow over (x)}i)−{right arrow over (u)}(k)({right arrow over (x)}j))=e−∥{right arrow over (u)}(k)({right arrow over (x)}i)−{right arrow over (u)}(k)({right arrow over (x)}j)∥2/2σL(k)({right arrow over (x)}i)σL(k)({right arrow over (x)}j), (4)
wherein σL(k)({right arrow over (x)}i) and σL(k)({right arrow over (x)}j) are local noise estimates at the position {right arrow over (x)}i and {right arrow over (x)}j, respectively.

186 more elements

Edge-preserving noise filtering

Life of this patent

Edge-preserving noise filtering

Life of this patent

Abstract

Claims19