UFRaw

On Sun, Jul 31, 2011 at 10:21 PM, Vladimir Nadvornik wrote:
> Hi,
>
> Is the lensfun configuration in id files supposed to work in current CVS?

UFRaw doesn't write lensfun data into id files. It's not a bug, it
simply isn't implemented.

Alexandre Prokoudine
http://libregraphicsworld.org

Dne neděle 31 Červenec 2011 21:03:25 Alexandre Prokoudine napsal(a):
> On Sun, Jul 31, 2011 at 10:21 PM, Vladimir Nadvornik wrote:
> > Hi,
> > 
> > Is the lensfun configuration in id files supposed to work in current CVS?
> 
> UFRaw doesn't write lensfun data into id files. It's not a bug, it
> simply isn't implemented.
> 

Well, from the recent comment in 
https://sourceforge.net/tracker/?func=detail&aid=3370224&group_id=127649&atid=709086
I got the impression that this is work in progress.

Vladimir

There are 3 types of files:

1. UFRaw ID files. These are created with the 'Create ID file' option.
They can be loaded with the command 'ufraw FILENAME.ufraw'.

2. Configuration files. These are the same as ID file, but they are
loaded with the command 'ufraw --conf=FILENAME.ufraw FILENAME.raw'.

3. The resource file $HOME/.ufrawrc.

UFRaw can read and write lensfun data to ID files for a while now.

Reading lensfun data from configuration files was just added to CVS a
few days ago.

UFRaw does not read or write lensfun data to the resource file. I
think that most users would considered it a bug if UFRaw would
remember lensfun setting between activations. Lensfun has its own lens
database at /usr/share/lensun/. If your lens is not in the database,
you can add it.

Udi

Dne neděle 31 Červenec 2011 20:21:30 Vladimir Nadvornik napsal(a):
> Hi,
> 
> Is the lensfun configuration in id files supposed to work in current CVS?
> 
> I have the following problem:
> 
> 1. use ufraw from CVS, delete the ~/.ufrawrc
> 2. run
> 
> ./ufraw IMG_1102.CR2
> 
> (the file is from Canon S95)
> ufraw identifes the lens and the correction is correctly applied
> 
> 3. click on auto-crop
> 4. set Create ID -> Only
> 5. Save
> 6. Run
> ./ufraw IMG_1102.ufraw
> 
> the correction is not applied and the lens description is empty
> 
> 
> Is anybody working on this? I could try to debug it if needed.
> 

I have found the problem.

When an ID file is read, the method LensModel::Event is called too
early, when Lensfun.Camera is not set.

The Lensfun.LensDB()->FindLenses on ufraw_lensfun.cc line 145 then sometimes 
returns NULL and the LensModel is left empty.

I have tested this:

Sigma 10-20 on Canon camera => FAIL
Canon 100mm on Canon camera => works OK
Canon S95 compact camera    => FAIL

The lens name is apparently not sufficient in all cases and the camera name is 
required too.

Vladimir

Dne pondělí 01 Srpen 2011 23:21:04 Vladimir Nadvornik napsal(a):
> Dne neděle 31 Červenec 2011 20:21:30 Vladimir Nadvornik napsal(a):
> > Hi,
> > 
> > Is the lensfun configuration in id files supposed to work in current CVS?
> > 
> > I have the following problem:
> > 
> > 1. use ufraw from CVS, delete the ~/.ufrawrc
> > 2. run
> > 
> > ./ufraw IMG_1102.CR2
> > 
> > (the file is from Canon S95)
> > ufraw identifes the lens and the correction is correctly applied
> > 
> > 3. click on auto-crop
> > 4. set Create ID -> Only
> > 5. Save
> > 6. Run
> > ./ufraw IMG_1102.ufraw
> > 
> > the correction is not applied and the lens description is empty
> > 
> > 
> > Is anybody working on this? I could try to debug it if needed.
> 
> I have found the problem.
> 
> When an ID file is read, the method LensModel::Event is called too
> early, when Lensfun.Camera is not set.
> 
> The Lensfun.LensDB()->FindLenses on ufraw_lensfun.cc line 145 then
> sometimes returns NULL and the LensModel is left empty.
> 
> I have tested this:
> 
> Sigma 10-20 on Canon camera => FAIL
> Canon 100mm on Canon camera => works OK
> Canon S95 compact camera    => FAIL
> 
> The lens name is apparently not sufficient in all cases and the camera name
> is required too.
> 

I was trying to debug it further. I think that during loading of an ID file, 
UFRaw should not search the lens in the database at all, but just use the 
values from the ID file. But then there is a problem with missing lens crop 
factor, which is not stored in the ID file. I think that it should be added.

Currently it first takes all lens parameters from the database and then 
overwrites the values that are present in the ID file. This can lead to 
inconsistent results when the lensfun database is updated (a change of lens 
crop factor is quite likely, for example when a full frame lens is first 
measured with smaller sensor and later on full frame camera).

So I think that UFRaw should use either one set of parameters or another, but 
don't combine values from both. And the set in ID file is incomplete.

Vladimir

> I was trying to debug it further. I think that during loading of an ID file,
> UFRaw should not search the lens in the database at all, but just use the
> values from the ID file. But then there is a problem with missing lens crop
> factor, which is not stored in the ID file. I think that it should be added.

There is more than the crop factor in the LensDB, there are also the
valid focal lens, etc.

I found several bugs in this code path, hopefully I will have a patch
ready soon.

> Currently it first takes all lens parameters from the database and then
> overwrites the values that are present in the ID file. This can lead to
> inconsistent results when the lensfun database is updated (a change of lens
> crop factor is quite likely, for example when a full frame lens is first
> measured with smaller sensor and later on full frame camera).

Good point.

> So I think that UFRaw should use either one set of parameters or another, but
> don't combine values from both. And the set in ID file is incomplete.

When LensfunAuto is true, UFRaw could alway read the parameters from
LensDB. This is what I do for configuration files. For ID files I
usually aim at reproducing the original output, so it is not optimal.

If UFRaw does use the parameters from the ID file, it needs to treat
them as manually set, at least if they are different from the LensDB
params. This is tricky and a bit ugly.

Udi

I've fixed some related bugs in CVS. I did some basic testing, but any
more testing would be welcome.

As long as there are no crop factor changes in LensDB, I think that
everything should work correctly.

Udi

Dne čtvrtek 04 Srpen 2011 06:03:49 Udi Fuchs napsal(a):
> I've fixed some related bugs in CVS. I did some basic testing, but any
> more testing would be welcome.
> 
> As long as there are no crop factor changes in LensDB, I think that
> everything should work correctly.
> 

Thank you!

Now it works fine for all "normal" lens.

I have only found problems with manual fish-eye "BelOMO, MC Peleng 3.5/8"
I selected the lens manually and set focal length to 8.0 (the camera has
incorrect value in exif - Canon 7D has 50mm, Canon 350D has 0mm).
Then I changed lens geometry from "unknown" to "fish-eye" (this might be also 
a bug, "fish-eye" type is already in the DB) and target geometry to 
"panoramic". After saving and loading an ID file the image area was black. The 
lens values looked correct.
"fish-eye" to "rectilinear" does not work at all - probably because the lens 
on APS-C camera covers more than 180 degrees diagonally and the rectilinear 
projection maps it to infinity.

Vladimir