Stereo matching system and stereo matching method using the same
A stereo matching system and a stereo matching method using the same. Here, a Sum of Edge Differences (SED) method as a disparity estimation method utilizing edge information is added to a disparity estimation method utilizing a local method to perform stereo matching. As such, it is possible to correct false matching in a non-texture region generated when stereo matching is performed using only a local method, thereby enabling good stereo matching.
1. A stereo matching system comprising:
a stereo image obtaining part adapted to obtain a left image and a right image;
a first disparity estimation part adapted to estimate disparity of the left image and the right image utilizing a local method;
an edge information obtaining part adapted to obtain edge information from the left image and the right image;
a second disparity estimation part adapted to estimate disparity through Sum of Edge Differences (SED) utilizing the edge information; and
a disparity synthesizing part adapted to synthesize disparities of the first disparity estimation part and the second disparity estimation part to produce a disparity map.
2. The stereo matching system according to claim 1 , wherein the first disparity estimation part utilizes Sum of Absolute Differences (SAD) to estimate the disparity of the left image and the right image.
3. The stereo matching system according to claim 1 , wherein the disparity synthesizing part synthesizes Sum of Absolute Differences (SAD) and the SED to produce the disparity map.
4. A stereo matching method comprising:
obtaining a left image and a right image;
estimating disparity of the left image and the right image utilizing a local method;
obtaining edge information from the left image and the right image;
estimating disparity through Sum of Edge Differences (SED) utilizing the edge information; and
synthesizing disparities of the estimating of the disparity of the left image and the right image utilizing the local method and the estimating of the disparity through the SED utilizing the edge information to produce a disparity map.
5. The stereo matching method according to claim 4 , wherein the estimating of the disparity of the left image and the right image utilizing the local method comprises utilizing Sum of Absolute Differences (SAD) to estimate the disparity of the right image and the left image.
6. The stereo matching method according to claim 4 , wherein the synthesizing of the disparities comprises synthesizing Sum of Absolute Differences (SAD) and the SED to produce the disparity map.
7. The stereo matching method according to claim 4 , wherein the SED satisfies the following:
SED
(
x
,
y
,
z
)
=
∑
V
DL
(
x
,
y
,
V
)
-
DR
(
x
+
d
,
y
,
V
)
,
wherein, in DL(x,y,V), x and y represent coordinate information of a reference pixel of a left image, V represents a first direction from the reference pixel to an edge of the left image,
wherein DL(x,y,V) represents a distance from the reference pixel of the left image to the edge of the left image along the first direction,
wherein, in DR(x+d,y,V), x+d and y represent coordinate information of a reference pixel of the right image moved by a disparity d in an X-axis direction corresponding to the reference pixel of the left image, and V represents the first direction from the reference pixel of the right image to an edge of the right image corresponding to the edge of the left image, and
wherein DR(x+d,y,V) represents a distance from the reference pixel of the right image to the edge of the right image along the first direction.
8. The stereo matching method according to claim 7 , wherein the first direction is a single direction or a plurality of directions.
9. The stereo matching method according to claim 4 , wherein the synthesizing of the disparities satisfies the following:
C ( d )= C local ( d )+λ C sed ( d ),
wherein C local (d) is a parameter representing the disparity estimation method utilizing the local method, C sed (d) is a parameter representing the SED as the disparity estimation method utilizing the edge information, and λ is a constant representing importance of the SED.
10. The stereo matching method according to claim 9 , further comprising reducing the C(d) to a minimum value.
11. The stereo matching method according to claim 4 , wherein the estimating of the disparity of the left image and the right image utilizing the local method is performed concurrently with the obtaining of the edge information from the left image and the right image and the estimating of the disparity through the SED utilizing the edge information.
12. The stereo matching method according to claim 4 , wherein the estimating of the disparity of the left image and the right image utilizing the local method is performed after the obtaining of the edge information from the left image and the right image and the estimating of the disparity through the SED utilizing the edge information.
13. A stereo matching system comprising:
first means for obtaining a left image and a right image;
second means for estimating disparity of the left image and the right image utilizing a local method;
third means for obtaining edge information from the left image and the right image;
fourth means for estimating disparity through Sum of Edge Differences (SED) utilizing the edge information; and
fifth means for synthesizing disparities of the estimating of the disparity of the left image and the right image utilizing the local method and the estimating of the disparity through the SED utilizing the edge information to produce a disparity map.
14. The stereo matching system according to claim 13 , wherein the second means utilizes Sum of Absolute Differences (SAD) to estimate the disparity of the left image and the right image.
15. The stereo matching system according to claim 13 , wherein the fifth means synthesizes Sum of Absolute Differences (SAD) and the SED to produce the disparity map.
16. The stereo matching system according to claim 13 , wherein the SED satisfies the following:
SED
(
x
,
y
,
z
)
=
∑
V
DL
(
x
,
y
,
V
)
-
DR
(
x
+
d
,
y
,
V
)
,
wherein, in DL(x,y,V), x and y represent coordinate information of a reference pixel of a left image, V represents a first direction from the reference pixel to an edge of the left image,
wherein DL(x,y,V) represents a distance from the reference pixel of the left image to the edge of the left image along the first direction,
wherein, in DR(x+d,y,V), x+d and y represent coordinate information of a reference pixel of the right image moved by a disparity d in an X-axis direction corresponding to the reference pixel of the left image, and V represents the first direction from the reference pixel of the right image to an edge of the right image corresponding to the edge of the left image, and
wherein DR(x+d,y,V) represents a distance from the reference pixel of the right image to the edge of the right image along the first direction.
17. The stereo matching system according to claim 16 , wherein the first direction is a single direction or a plurality of directions.
18. The stereo matching system according to claim 13 , wherein the synthesizing of the disparities:
C ( d )= C local ( d )+λ C sed ( d ),
wherein C local (d) is a parameter representing the second means utilizing the local method, C sed (d) is a parameter representing the SED as the fourth means utilizing the edge information, and λ is a constant representing importance of the SED.
19. The stereo matching system according to claim 18 , further comprising means for reducing the C(d) to a minimum value.
20. The stereo matching system according to claim 13 , further comprising means for performing the second means concurrently with or after the third means and the fourth means.