Multi-view image fusion by image space equalization and stereo-based rectification from two different cameras
Methods to solve the problem of performing fusion of images acquired with two cameras with different type sensors, for example a visible (VIS) digital camera and an short wave infrared (SWIR) camera, include performing image space equalization on images acquired with the different type sensors before performing rectification and registration of such images in a fusion process.
1. A method for fusing images from two different cameras, a first camera and a second camera, the method comprising:
a) performing a transformation that translates an image space of the first camera into an image space of the second camera to obtain a stereo rig in the form of a pair of space equalized images, the transformation comprising:
(i) calculating coordinates of each pixel in a sensor space of a pin-hole model of the second camera that has a second image sensor with a second resolution that is lower than a first resolution of a first image sensor of the first camera,
(ii) based on the calculated coordinates, transforming a sensor space of the first camera that has the first image sensor with the first resolution into the image space of the second camera that has the second image sensor with the second resolution that is lower than the first resolution by calculating new pixel coordinates u, v in the first camera corresponding to new pixel positions in the first camera x R y R using the transformation
u
(
x
R
)
=
p
R
+
(
x
R
-
p
R
)
·
F
L
·
H
R
F
R
·
H
L
v
(
y
R
)
=
q
R
+
(
y
R
-
q
R
)
·
F
L
·
W
R
F
R
·
W
L
wherein H L , W L are respectively a height and a width of the first image sensor, H R , W R are respectively a height and a width of the second image sensor, p R , q R are pixel positions of a principal point of the second camera in respectively horizontal and vertical directions, F L is a pixel relative focal length of the first camera and F R is a pixel relative focal length of the second camera, and
(iii) setting up positions of image pixels of the first camera in the image space of the second camera to obtain the pair of space equalized images;
b) performing rectification on the pair of space equalized images to obtain rectified first and second images;
c) performing registration on the rectified first and second images to obtain registered first and second images; and
d) fusing the registered first and second images into a fused image.
2. The method of claim 1 , wherein the first camera differs from the second camera in at least one parameter selected from focal length, distortion, gain and spectral range.
3. The method of claim 1 , wherein the transforming of the sensor space of the first camera comprises outputting a look-up table of pixel positions correlating between pixels positions of the first sensor and pixel positions of the second image.
4. The method of claim 1 , wherein the first camera and the second camera operate in differing spectral ranges.
5. The method of claim 1 , wherein the first camera is a visible (VIS) range camera sensitive in the VIS spectral range and wherein the second camera is a short-wave infrared (SWIR) camera sensitive in the SWIR range.
6. The method of claim 1 , wherein the performing of the rectification comprises rectifying the pair of space equalized images that originate from the first image sensor and the second image sensor which differ in at least their resolution and spectral range to obtain the rectified first and second images.
7. The method of claim 1 , wherein the performing of the registration comprises performing registration on the rectified first and second images that originate from the first image sensor and the second image sensor which differ in at least their resolution and spectral range to obtain the registered first and second images.