I think that if you uses the subsampling option you can perform consecutive 
registrations of increasing accuracy with a smaller metric windows. in this 
way, you can maybe obtain similar results reducing the time.


Best regards
Carlos 

El miércoles, 26 de noviembre de 2014 17:44:04 UTC+1, alobo escribió:
>
> After having used otbcli_FineRegistration quite a lot, 
> I have come to the conclusion that the Metric window defined by Metric 
> Radius X and 
> Metric Radius Y must be quite large, otherwise the calculated offset 
> values are prone 
> to be too local (it is easy to find sporious high correlations at 
> positions that are not geometrically correct if the window is over a 
> homogeneous area) and unrealistically variable. 
>
> The problem is that the program is very slow for large Metric windows. 
> Neverthless, there is no reason why the metric window should be 
> positioned at every pixel. 
> FineRegistration could have "jumpX", jumpY" arguments so that the 
> metric is calculated 
> every jumpX and jumpY pixels, which would speed up computation a lot. 
> For example 
>
> -erx 5 -ery 5 -mrx 121 -mry 121 -jumpx 31 -jumpy 31 
> would be enough 
> The final 
> X and Y offset "images" would then be just interpolated. It is more 
> important a large mrx and mry 
> than having the values calculated at every pixel. 
>
> This should not be hard to implement for someone knowing well the 
> code. I would be willing to intensively test the new implementation. 
>
> Agus 
>

-- 
-- 
Check the OTB FAQ at
http://www.orfeo-toolbox.org/FAQ.html

You received this message because you are subscribed to the Google
Groups "otb-users" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/otb-users?hl=en
--- 
You received this message because you are subscribed to the Google Groups 
"otb-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to