Topic: Orthorectification of images with high off-nadir angle

Is it correct that if I have two images with the same off-nadir angle higher than 5 degrees that then I will need to use stereo pairs.  Or is it only if the difference between the off-nadir angles of the images is higher than 5 degrees?
Can I compare images with different resolution, one with 5 m and one with 0,5 m resolution?
I have been warned of using spot 5 images, is it some problem with them in cosi corr?
As from the posts in the forum is it true that I will need DEM for the whole image or can I use DEM only for the area of interest, if the images I compare don´t cover the excactly same area how do I then select the area I need DEM for?
Best regards
Haraldur Hallsteinsson


Re: Orthorectification of images with high off-nadir angle

Dear Haraldur,

the smaller the difference in incidence angle, the better. If your two images have the same off-nadir angle, it is not too bad. You still want a DEM to correct for the parallax effect of the images, but relatively between each other, the parallax will be minimum.

You can compare image with different resolution, but before that, you'll need to have the image resampled to the SAME resolution (to the coarsest resolution). In you example (5m/pix and 0.5m/pix), the 0.5m/pix image will be resampled to 5m/pix.

SPOT 5 image are usually very good quality images. It all depends on what you want to do. If you're after a 10 cm displacement, SPOT 5 imagery is too coarse for that. It is good if you're looking at displacement of 1m and up.

In theory you need a DEM only on the overlapping area of the images. However it is always good to have a DEM that covers both images extent. This could facilitate getting tie-points between the first image and the DEM.



Re: Orthorectification of images with high off-nadir angle

Thank you for the answers, I´m orthorectifying and image to image registration my images which are WV1 images or part of images in ENVI.  I´m having trouble getting them to fit when I´m registering them, how good does the registration need to be, I have added up to 300 tie points.  Is it better to orthorectify them before or after the registration?
The images have high but similar off nadir angle.


Re: Orthorectification of images with high off-nadir angle

Dear Haraldur,

the sequence should be, first you "register" them through tie-points selection (it does not mean that you are going to actually register the images, but you are taking tie-points such that DURING the orthorectification, the tie-points will help the coregistration), then second you orthorectify the images while accounting for the tie-points.

If I only consider COSI-Corr process here, your process should be:
- Select tie-points between your first WV1 image and your DEM or other orthoimage well registered to the DEM
- Optimize the tie-points if possible (it'll depends on your reference image if it's a DEM or an orthoimage. Optmization with a shaded DEM is hard)
- Orthorectify your first WV1 image

- Select tie-points between your second WV1 image and the ortho you just generated
- Optimize the tie-points
- Orthorectify the second image, with the optimized tie-points (we call them optimized GCPS)

At this stage, your two orthoimages should be well coregistered. If it is not the case then the reasons are either or a combination of:
- your DEM is way too coarse for your images which are taken on rough topography
- some tie-points (for the first and/or second image) are wrong
- the tie-points optimization failed

Please check the User Guide which will give you details on each of the steps involved.