Camera system and image registration method thereof
View Patent ↗Provided are a camera system and an image registration method for registering images taken by multiple cameras moving in the same manner. The image registration method includes: acquiring multiple sets of images of a reference subject, wherein each of the multiple sets of images comprises images taken by the multiple cameras facing a same direction and images of different sets are taken with the multiple cameras facing different directions; performing image registration using the multiple sets of images; and rendering a result image showing a result of the image registration.
1. An image registration method performed by a processor to register images taken by multiple cameras moving in a synchronized manner at same relative positions each other, the image registration method comprising:
acquiring multiple sets of images of a reference subject, wherein
an n-th camera takes an n-th set of images,
each set of the multiple sets of images has more than one image, and
in each set of the multiple sets of images, directions of the camera taking the images in the set are different each other;
generating reference images respectively corresponding to the multiple sets of images by synthesizing images of each of the multiple sets of images;
comparing a first field of view of a first camera with a second field of view of a second camera;
performing image registration according to a first registration algorithm when a difference between the first field of view and the second field of view is greater than a reference value;
performing image registration according to a second registration algorithm when the difference between the first field of view and the second field of view is less than or equal to the reference value; and
rendering a result image showing a result of the image registration,
wherein the first registration algorithm estimates a conversion model by matching corresponding feature points between first and second reference images, and
wherein the second registration algorithm estimates the conversion model by matching corresponding feature points between the multiple sets of images without using the first and second reference images.
2. The image registration method of claim 1 , wherein each of the multiple sets of images shows a different wavelength characteristic of the reference subject.
3. The image registration method of claim 1 , wherein the generating the reference images comprises:
generating feature point images by masking regions outside the reference subject in the multiple sets of images; and
synthesizing the feature point images for each of the multiple sets of images.
4. The image registration method of claim 1 , wherein a center coordinate of the reference subject in each reference image is used as a coordinate of a feature point of the reference image.
5. A camera system comprising:
multiple cameras configured to move in a synchronized manner at same relative positions each other; and
a controller configured to:
acquire multiple sets of images of a reference subject, wherein
an n-th camera takes an n-th set of images,
each set of the multiple sets of images has more than one image, and
in each set of the multiple sets of images, directions of the camera taking the images in the set are different each other,
generate reference images respectively corresponding to the multiple sets of images by synthesizing images of each of the multiple sets of images,
compare a first field of view of a first camera with a second field of view of a second camera,
perform image registration according to a first registration algorithm when a difference between the first field of view and the second field of view is greater than a reference value, and
perform image registration according to a second registration algorithm when the difference between the first field of view and the second field of view is less than or equal to the reference value, and
perform rendering of a result image showing a result of the image registration,
wherein the first registration algorithm estimates a conversion model by matching corresponding feature points between first and second reference images, and
wherein the second registration algorithm estimates the conversion model by matching corresponding feature points between the multiple sets of images without using the first and second reference images.
6. The camera system of claim 5 , wherein a center coordinate of the reference subject in each reference image is used as a coordinate of a feature point of the reference image.
7. The camera system of claim 5 , wherein the controller is further configured to generate feature point images by masking regions outside the reference subject in the multiple sets of images, and synthesize images for each of the multiple sets of images.
8. The camera system of claim 5 , wherein each of the multiple sets of images shows a different wavelength characteristic of the reference subject.