Hi,
I tried to orthorectify aerial photographs. That worked fine by now, until i got scans wich appear to be darker as before. At least this is the first observation I made. The effect occured with that scans is, that after rectification some of the dark areas become white (or rather null, querying that cells result in -0.722622310236638).
Has anyone had similar experiences when rectifying aerial photographs. And hopefully a hint how to avoid this damages. I suspect that the orthorectified images become double precision whereas the imported scans are integer...
OK, I found a workaround.
with r.mapcalc I identify all values lt 1 and gt 254 and set them to 1 resp. 254. Actually not a nice way but the result is good enough for the moment.
Anyway, I still would be thankfull for an explanation of the phenomenon, moreover a better solution to avoid that behaviour.
best regards
Stefan
Am 21.02.2014 00:26, schrieb Stefan Kiefer:
Hi,
I tried to orthorectify aerial photographs. That worked fine by now, until i got scans wich appear to be darker as before. At least this is the first observation I made. The effect occured with that scans is, that after rectification some of the dark areas become white (or rather null, querying that cells result in -0.722622310236638).
Has anyone had similar experiences when rectifying aerial photographs. And hopefully a hint how to avoid this damages. I suspect that the orthorectified images become double precision whereas the imported scans are integer...
On Fri, Feb 21, 2014 at 12:26 AM, Stefan Kiefer <st_kiefer@web.de> wrote:
Hi,
I tried to orthorectify aerial photographs. That worked fine by now, until
i got scans wich appear to be darker as before. At least this is the first
observation I made. The effect occured with that scans is, that after
rectification some of the dark areas become white (or rather null, querying
that cells result in -0.722622310236638).
This should only happen with method=cubic or method=cubic_f. The other
resampling methods nearest,bilinear,bilinear_f should not produce
these resampling overshoots.
Markus M
Has anyone had similar experiences when rectifying aerial photographs. And
hopefully a hint how to avoid this damages. I suspect that the
orthorectified images become double precision whereas the imported scans are
integer...
Hi Markus,
thanks for that hint, that might be the solution. As I must confess I did not only try any of the other methods, believing cubic would yield the best results because the scans are from an area with high energy of relief.
I will try this when I’m back in office and give a feedback of my experiences then.
On Fri, Feb 21, 2014 at 12:26 AM, Stefan Kiefer st_kiefer@web.de wrote:
Hi,
I tried to orthorectify aerial photographs. That worked fine by now, until
i got scans wich appear to be darker as before. At least this is the first
observation I made. The effect occured with that scans is, that after
rectification some of the dark areas become white (or rather null, querying
that cells result in -0.722622310236638).
This should only happen with method=cubic or method=cubic_f. The other
resampling methods nearest,bilinear,bilinear_f should not produce
these resampling overshoots.
Markus M
Has anyone had similar experiences when rectifying aerial photographs. And
hopefully a hint how to avoid this damages. I suspect that the
orthorectified images become double precision whereas the imported scans are
integer…
On Fri, Feb 21, 2014 at 9:27 AM, . . <st_kiefer@web.de> wrote:
Hi Markus,
thanks for that hint, that might be the solution. As I must confess I did
not only try any of the other methods, believing cubic would yield the best
results because the scans are from an area with high energy of relief.
I would leave it as it is. Bicubic gives nice results, but these
overshoots must be taken care of. Your method of forcing values back
to the range of 1,254 seems appropriate to me for handling resampling
overshoots.
Markus M
I will try this when I'm back in office and give a feedback of my
experiences then.
best regards
Stefan
Markus Metz <markus.metz.giswork@gmail.com> hat am 21. Februar 2014 um
08:05 geschrieben:
On Fri, Feb 21, 2014 at 12:26 AM, Stefan Kiefer <st_kiefer@web.de> wrote:
> Hi,
> I tried to orthorectify aerial photographs. That worked fine by now,
> until
> i got scans wich appear to be darker as before. At least this is the
> first
> observation I made. The effect occured with that scans is, that after
> rectification some of the dark areas become white (or rather null,
> querying
> that cells result in -0.722622310236638).
This should only happen with method=cubic or method=cubic_f. The other
resampling methods nearest,bilinear,bilinear_f should not produce
these resampling overshoots.
Markus M
> Has anyone had similar experiences when rectifying aerial photographs.
> And
> hopefully a hint how to avoid this damages. I suspect that the
> orthorectified images become double precision whereas the imported scans
> are
> integer...
>
> best regards
>
> Stefan
> _______________________________________________
> grass-user mailing list
> grass-user@lists.osgeo.org
> http://lists.osgeo.org/mailman/listinfo/grass-user
Hi,
as Markus wrote there are no overshoots with bilinear method. Nevertheless I agree with keeping cubic and forcing the range. Particular because I only rely on greyscale aerial photographs, so it doesent matter if its more or less black or white. Especially when humans can't distinguish more than 130 shades of grey. An apropriate rectification is more desirable.
Thanks again for the support and the discussion.
best regards
Stefan
Am 21.02.2014 13:57, schrieb Markus Metz:
On Fri, Feb 21, 2014 at 9:27 AM, . . <st_kiefer@web.de> wrote:
Hi Markus,
thanks for that hint, that might be the solution. As I must confess I did
not only try any of the other methods, believing cubic would yield the best
results because the scans are from an area with high energy of relief.
I would leave it as it is. Bicubic gives nice results, but these
overshoots must be taken care of. Your method of forcing values back
to the range of 1,254 seems appropriate to me for handling resampling
overshoots.
Markus M
I will try this when I'm back in office and give a feedback of my
experiences then.
best regards
Stefan
Markus Metz <markus.metz.giswork@gmail.com> hat am 21. Februar 2014 um
08:05 geschrieben:
On Fri, Feb 21, 2014 at 12:26 AM, Stefan Kiefer <st_kiefer@web.de> wrote:
Hi,
I tried to orthorectify aerial photographs. That worked fine by now,
until
i got scans wich appear to be darker as before. At least this is the
first
observation I made. The effect occured with that scans is, that after
rectification some of the dark areas become white (or rather null,
querying
that cells result in -0.722622310236638).
This should only happen with method=cubic or method=cubic_f. The other
resampling methods nearest,bilinear,bilinear_f should not produce
these resampling overshoots.
Markus M
Has anyone had similar experiences when rectifying aerial photographs.
And
hopefully a hint how to avoid this damages. I suspect that the
orthorectified images become double precision whereas the imported scans
are
integer...