Back projected dispartiy map

15 views
Skip to first unread message

Riccardo Pedrelli

unread,
Jun 26, 2026, 5:57:54 AM (13 days ago) Jun 26
to Ames Stereo Pipeline Support
Hi all,

If I understood correctly, when working with mapprojected images, after producing the high resolution disparity map, the triangulation is performed by back projected on the original camera position, and based on that and the camera model, the height is trinagulated. Did I got this right?

I'm working with radar data (so using RPC camera model), and I would be very intrested on the disparity map in the original image coordinates, instead of the map projected one. Is it possible to generate it? 

Thank you in advnace,
Riccardo

Oleg Alexandrov

unread,
Jun 26, 2026, 12:39:12 PM (12 days ago) Jun 26
to Riccardo Pedrelli, Ames Stereo Pipeline Support
If I understood correctly, when working with mapprojected images, after producing the high resolution disparity map, the triangulation is performed by back projected on the original camera position, and based on that and the camera model, the height is trinagulated. Did I got this right?

Yes, that is correct.
 
I'm working with radar data (so using RPC camera model), and I would be very intrested on the disparity map in the original image coordinates, instead of the map projected one. Is it possible to generate it? 

We have this disparity map, but it was not documented. I am fixing that now. The option is --unalign-disparity. This is obtained from the aligned disparity, via some interpolation process, so it should be good enough but likely not as precise as the one between aligned images. 

You can run only the stereo_tri step with this option to recover this disparity. It will be written in the output directory while having "unalign" as part of its name. 

It is suggested to run disparitydebug with the option --raw on it that will extract the two bands from it while handling correctly missing data. Ideally these two images should be 1-1 with your raw left and right cameras.

Note that our algorithms are having a had time with radar and SAR as these are not optical modalities.



--
You received this message because you are subscribed to the Google Groups "Ames Stereo Pipeline Support" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ames-stereo-pipeline...@googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/ames-stereo-pipeline-support/9b2ced2d-e168-4a27-bc28-b4a3673fca58n%40googlegroups.com.
Reply all
Reply to author
Forward
0 new messages