Results on Visualization Experiments
Basic Image Combinations
Two different images of a common area are transformed (rigid only)
so that they are registered.
| Synthetic Aperture Radar magnitude
|
| Panchromatic images can be of very high resolution
|
| SAR and Panchromatic. Buildings and 2D ground structures
enhanced by panchromatic image
|
Image Warping and Combinations
The warping application works as follows: Given two images (SAR,
optical, Infra-red, street maps, images of 3D models...)
tries to find the best interpolation between the two images, so the two
images can be overlayed. That's based on some point correspodences
interactively chosen by the user.
Two transformations have been tested for the warping:
-
RBF
Given a set of point correspondences well-spaced in the image, between
the source and the target images, we compute a Radial Basis Function.
This is a non-linear transformation. The anchor points will suffer no
deformation at all under the RBF. The image points will be deformed
according to their Euclidean distance to the anchor points in the linear
RBF. We have also implemented a Gaussian RBF and the Hardy polynomes.
The linear works fine.
-
Rigid Transformation
Given at least two point correspondences between source and target
images,
it computes an rigid transformation between the two images:
- an isotropic scale,
- 2D rotation and
- 2D translation
This transformation is applied to a regular mesh well-spaced in the
image (instead of each point of the image to speed up) to generate the
final warped image.
Combination of SAR and a Panchromatic image
|
|
The initial images and the selected points used in the registration.
|
|
|
|
|
The above image are 4 different combinations. For the full range see:
Movie -- AVI
The AVI movie allows you to see the different levels of blending.
|
Combination of SAR and a computed DEM
the DEM is coded by elevation,
with green corresponding to the ground.
|
|
The Input images with the selected points used in the registration.
|
|
|
|
|
The above image are 4 different combinations. For the full range see:
Movie -- AVI
The AVI movie allows you to see the different levels of blending.
|
Hyper Spectral Classification and Pan
Hyperspectral based classification mixed with Panchromatic image.
4 frames and a pointer to the movie
|
|
|
|
The above image are 4 different combinations. For the full range see:
Movie -- AVI
The AVI movie allows you to see the different levels of blending.
|
Once the source image has been warped to "fit" the target image, both
are overlayed, and a slider allows you to change the opacity
(alpha-blending) from 0 to 1, so you can see all the posibilities of
blending. The .avi file shows the blending sequence for the warped
source image and the target image.
Relation to other Software
This is just the initial phase of our work. The typical method used
by Image Links,
Autodesk,
Autometric,
etc. is to transform the image to a common orthographic, geo-registered
frame of reference and combine these transformed images. Our approach
would allow the combining of oblique views.