Endoscopic stereo matching method and apparatus using direct attenuation model
An endoscopic stereo matching method and apparatus using a direct attenuation model is provided. A method for generating a depth image according to an embodiment includes: generating a stereo image; and estimating a depth image from the stereo image based on an attenuation trend of light of an illumination used to generate the stereo image. Accordingly, a dense depth image can be obtained by using images obtained from a capsule endoscope, and thus the geometric structure of the inside of the GI tract can be estimated.
1. A method for generating a depth image, the method comprising:
generating a stereo image in an environment where there is no external atmospheric light; and
estimating a depth image from the stereo image based on scene radiance information of the stereo image and an attenuation trend of a light of an illumination used to generate the stereo image,
wherein the light is different from the external atmospheric light, and
wherein the scene radiance information of the stereo image indicates information approximated as a mean of all pixel values of the stereo image.
2. The method of claim 1 , wherein the external atmospheric light is a natural light of sunlight.
3. The method of claim 1 , wherein the estimating the depth image comprises estimating the depth image by using the following equation:
d
(
p
)
=
In
(
J
(
p
)
)
-
In
(
I
(
p
)
)
β
(
p
)
where d(p) is a depth of a pixel p, J(p) is a scene radiance of the pixel p, I(p) is an observed intensity of the pixel p, and β(p) is a light attenuation coefficient.
4. The method of claim 3 , wherein the β(p) is a sum of an absorption coefficient and a scattering coefficient of light.
5. The method of claim 1 , wherein the generating comprises generating an internal stereo image filled with a homogeneous matter.
6. The method of claim 3 , wherein the β(p) is approximated as a constant value.
7. The method of claim 6 , wherein the J(p), corresponding to the scene radiance of the pixel p, is approximated as the mean of all pixel values.
8. The method of claim 1 , wherein the estimating the depth image comprises estimating the depth image by using the following equation:
d β ( p )=β d ( p )= In ( Ī )− In ( I ( p ))
where d β (p) is a depth image up to a scale factor β regarding the pixel p, Ī is a mean of all pixel values, and I(p) is an observed intensity of the pixel p.
9. The method of claim 8 , wherein the scale factor β is determined according to the following equations:
β
=
d
β
(
p
)
/
d
s
(
p
)
d
s
(
p
)
=
fB
p
x
L
-
p
x
R
where p x L and p x R are positions of matched points from left and right images along an x-axis, f is a focal distance of a left camera, and B is a distance between cameras.
10. An image system comprising:
a stereo camera configured to generate a stereo image in an environment where there is no external atmospheric light; and
a computing system configured to estimate a depth image based on scene radiance information of the stereo image and an attenuation trend of light of an illumination used to generate the stereo image,
wherein the scene radiance information of the stereo image indicates information associated with a mean of all pixel values of the stereo image.
11. A method for generating a depth image, the method comprising:
estimating a depth image from a stereo image based on scene radiance information of the stereo image and an attenuation trend of light of an illumination used to generate the stereo image,
wherein the stereo image is generated in an environment where there is no external atmospheric light, and the scene radiance information of the stereo image indicates information associated with a mean of all pixel values of the stereo image; and
outputting the estimated depth image.
12. The image system of claim 10 , wherein the external atmospheric light is a natural light of sunlight.
13. The image system of claim 10 , wherein, for the estimating the depth image, the computing system is configured to estimate the depth image by using the following equation:
d
(
p
)
=
In
(
J
(
p
)
)
-
In
(
I
(
p
)
)
β
(
p
)
where d(p) is a depth of a pixel p, J(p) is a scene radiance of the pixel p, I(p) is an observed intensity of the pixel p, and β(p) is a light attenuation coefficient.
14. The image system of claim 13 , wherein the β(p) is a sum of an absorption coefficient and a scattering coefficient of light.
15. The image system of claim 10 , wherein, for the generating, the computing system is configured to generate an internal stereo image filled with a homogeneous matter.
16. The image system of claim 13 , wherein the β(p) is approximated as a constant value.
17. The image system of claim 16 , wherein the J(p), corresponding to the scene radiance of the pixel p, is approximated as the mean of all pixel values.
18. The image system of claim 10 , wherein, for the estimating the depth image, the computing system is configured to estimate the depth image by using the following equation:
d β ( p )=β d ( p )= In ( Ī )− In ( I ( p ))
where d β (p) is a depth image up to a scale factor β regarding the pixel p, Ī is a mean of all pixel values, and I(p) is an observed intensity of the pixel p.
19. The image system of claim 18 , wherein the scale factor β is determined according to the following equations:
β
=
d
β
(
p
)
/
d
s
(
p
)
d
s
(
p
)
=
fB
p
x
L
-
p
x
R
where p x L and p x R are positions of matched points from left and right images along an x-axis, f is a focal distance of a left camera, and B is a distance between cameras.