Image stitching method using viewpoint transformation and system therefor
View Patent ↗An image stitching method using viewpoint transformation and a system therefor are provided. The method includes obtaining images captured by a plurality of cameras included in the camera system, performing viewpoint transformation for each of the images using a depth map for the images, and stitching the images, the viewpoint transformation of which is performed.
1. An image stitching method using viewpoint transformation executed by a computer included in a camera system, the method comprising:
obtaining images captured by a plurality of cameras included in the camera system;
performing viewpoint transformation for each of the images using a depth map for the images; and
stitching the images, the viewpoint transformation of which is performed,
wherein the performing of the viewpoint transformation for each of the images comprises:
setting a viewpoint transformation formula including a backward warping equation;
determining a movement parameter in the viewpoint transformation formula, the movement parameter being a value associated with a distance to move when a location where each of the images is captured moves to a virtual specific location; and
transforming viewpoints of the images using the viewpoint transformation formula to be matched,
wherein the backward warping equation is:
I
(
x
,
y
)
=
I
′
(
x
′
(
fB
+
m
z
D
)
+
fm
x
D
fB
-
c
x
,
y
′
(
fB
+
m
y
D
)
+
fm
z
D
fB
-
c
y
)
and in the backward warping equation, I(x, y) denotes the images captured by a plurality of cameras included in the camera system, x, y denote the pixel locations in each the images captured by the plurality of cameras included in the camera system, I′(x′, y′) denotes transformed images resulting from the viewpoint transformation which is performed, x′, y′ refer to the pixel locations in each the transformed images resulting from the viewpoint transformation which is performed, f denotes to a focal length of each of the plurality of cameras, B denotes to a length of a baseline of the camera system, D denotes to a parallax between the plurality of cameras, m x , m y , m z denote the movement parameter, and c x , c y denote coordinates of an image center passing through an optical axes.
2. The method of claim 1 , wherein the determining of the movement parameter in the viewpoint transformation formula comprises:
determining the movement parameter based on a distance to move when an optical center of each of the plurality of cameras moves to a virtual reference location of the camera system.
3. The method of claim 2 , wherein the determining of the movement parameter in the viewpoint transformation formula comprises:
determining the movement parameter based on the length of the baseline of the camera system.
4. The method of claim 1 , wherein the transforming of the images to be matched comprises:
transforming the viewpoint of each of the images into a viewpoint of an image captured at a reference location of the camera system.
5. The method of claim 1 , wherein the performing of the viewpoint transformation for each of the images further comprises:
filling a hole of the depth map before performing the viewpoint transformation for each of the images.
6. The method of claim 1 , wherein the performing of the viewpoint transformation for each of the images further comprises:
filling a hole of each of the images, the viewpoint transformation of which is performed.
7. The method of claim 1 , wherein the stitching of the images, the viewpoint transformation of which is performed, further comprises:
transforming a viewpoint of the stitched image to fill a hole of the stitched image.