[darktable-devel] HDR pink colors

[bva]

When I try to create HDR from 3 CR2 images the resulting DNG is in
pink colors (see attachment). The same result with tone mapping on or
off. All other modules are off. I'm doing something wrong or this is a
bug?

darktable --version
this is darktable 0.9.1+85~ge4912c6

[bva]

And I just noticed that there is no this bug in rtopencl branch.

Post by bva
When I try to create HDR from 3 CR2 images the resulting DNG is in
pink colors (see attachment). The same result with tone mapping on or
off. All other modules are off. I'm doing something wrong or this is a
bug?
darktable --version
this is darktable 0.9.1+85~ge4912c6

[johannes hanika]

i tried and could reproduce, but then i shot 3 fresh brackets and
merged them, all was fine. might it be a broken cache issue again?
this looks like you're getting the wrong CFA layout from somewhere.
does it do heavy mazing, when you zoom in?

can you try to delete ~/.cache/darktable/mipmaps* and try again if
it's still pink?

-jo

Post by bva
And I just noticed that there is no this bug in rtopencl branch.

Post by bva
When I try to create HDR from 3 CR2 images the resulting DNG is in
pink colors (see attachment). The same result with tone mapping on or
off. All other modules are off. I'm doing something wrong or this is a
bug?
darktable --version
this is darktable 0.9.1+85~ge4912c6

------------------------------------------------------------------------------
Got Input?   Slashdot Needs You.
Take our quick survey online.  Come on, we don't ask for help often.
Plus, you'll get a chance to win $100 to spend on ThinkGeek.
http://p.sf.net/sfu/slashdot-survey
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[bva]

Post by johannes hanika
this looks like you're getting the wrong CFA layout from somewhere.

CFA? Sorry, could you please write a long version of this
abbreviation? I have no idea what CFA means. My english is very poor.

Post by johannes hanika
does it do heavy mazing, when you zoom in?

yes.

Post by johannes hanika
can you try to delete ~/.cache/darktable/mipmaps* and try again if
it's still pink?

I deleted cache. Then I tryed to make new HDR from 3 CR2s with the
same result :(

I want to say again: in rtopencl branch I can't reproduce this with
one version of "mipmaps" file for both rtopencl & master. May be this
information can help.

[johannes hanika]

Post by bva

Post by johannes hanika
this looks like you're getting the wrong CFA layout from somewhere.

CFA? Sorry, could you please write a long version of this
abbreviation? I have no idea what CFA means. My english is very poor.

Post by johannes hanika
does it do heavy mazing, when you zoom in?

yes.

Post by johannes hanika
can you try to delete ~/.cache/darktable/mipmaps* and try again if
it's still pink?

I deleted cache. Then I tryed to make new HDR from 3 CR2s with the
same result :(
I want to say again: in rtopencl branch I can't reproduce this with
one version of "mipmaps" file for both rtopencl & master. May be this
information can help.

can you post the output of both

git show master | head -1

and

git show rtopencl | head -1

please? i think your rtopencl branch might be out of sync with
origin/rtopencl and thus the difference. there should be no notable
difference between rtopencl and master about demosaicing things. maybe
i can use that as a base for git bisect and can find out what the hell
is going on here..

-jo

[David Vincent-Jones]

After a lot of (hard) work with wide dynamic range images this is my
take on the Pink HDR problem:

Pink artifact result solely from totally blown highlights. dt simply
does not handle blown highlights well (at this time).

It is too easy to simply take 3 bracketed shots and assume that the
subject range requirements will be met. If any of the images include
blown highlights then the resulting DNG file will not prove useful. The
approach that I take is to ensure that the brightest image is fully
within the sensor's capability and then add darker component images.
This works well in dt using its tone-mapping with prodigious
capabilities in the low tonal range.

David

Post by johannes hanika

Post by bva

Post by johannes hanika
this looks like you're getting the wrong CFA layout from somewhere.

CFA? Sorry, could you please write a long version of this
abbreviation? I have no idea what CFA means. My english is very poor.

Post by johannes hanika
does it do heavy mazing, when you zoom in?

yes.

Post by johannes hanika
can you try to delete ~/.cache/darktable/mipmaps* and try again if
it's still pink?

I deleted cache. Then I tryed to make new HDR from 3 CR2s with the
same result :(
I want to say again: in rtopencl branch I can't reproduce this with
one version of "mipmaps" file for both rtopencl & master. May be this
information can help.

can you post the output of both
git show master | head -1
and
git show rtopencl | head -1
please? i think your rtopencl branch might be out of sync with
origin/rtopencl and thus the difference. there should be no notable
difference between rtopencl and master about demosaicing things. maybe
i can use that as a base for git bisect and can find out what the hell
is going on here..
-jo
------------------------------------------------------------------------------
BlackBerry® DevCon Americas, Oct. 18-20, San Francisco, CA
The must-attend event for mobile developers. Connect with experts.
Get tools for creating Super Apps. See the latest technologies.
Sessions, hands-on labs, demos & much more. Register early & save!
http://p.sf.net/sfu/rim-blackberry-1
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[johannes hanika]

hey.

On Wed, Aug 3, 2011 at 12:21 PM, David Vincent-Jones

Post by David Vincent-Jones
After a lot of (hard) work with wide dynamic range images this is my
Pink artifact result solely from totally blown highlights. dt simply
does not handle blown highlights well (at this time).

that's not true. the pipe changed so highlight clipping can now be
done before tone mapping. you have to leave highlights switched on, as
for regular raws, too. are you on the latest git master? 0.9.1 should
also include that hdr change.

Post by David Vincent-Jones
It is too easy to simply take 3 bracketed shots and assume that the
subject range requirements will be met. If any of the images include
blown highlights then the resulting DNG file will not prove useful. The
approach that I take is to ensure that the brightest image is fully
within the sensor's capability and then add darker component images.
This works well in dt using its tone-mapping with prodigious
capabilities in the low tonal range.

not true anymore, see above. the pink bva was talking about is because
the CFA (color filter array) layout was interpreted the wrong way
(e.g. rggb instead of gbrg or similar)

-jo

Post by David Vincent-Jones

Post by johannes hanika

Post by bva

Post by johannes hanika
this looks like you're getting the wrong CFA layout from somewhere.

CFA? Sorry, could you please write a long version of this
abbreviation? I have no idea what CFA means. My english is very poor.

Post by johannes hanika
does it do heavy mazing, when you zoom in?

yes.

Post by johannes hanika
can you try to delete ~/.cache/darktable/mipmaps* and try again if
it's still pink?

I deleted cache. Then I tryed to make new HDR from 3 CR2s with the
same result :(
I want to say again: in rtopencl branch I can't reproduce this with
one version of "mipmaps" file for both rtopencl & master. May be this
information can help.

can you post the output of both
git show master | head -1
and
git show rtopencl | head -1
please? i think your rtopencl branch might be out of sync with
origin/rtopencl and thus the difference. there should be no notable
difference between rtopencl and master about demosaicing things. maybe
i can use that as a base for git bisect and can find out what the hell
is going on here..
-jo
------------------------------------------------------------------------------
BlackBerry® DevCon Americas, Oct. 18-20, San Francisco, CA
The must-attend event for mobile developers. Connect with experts.
Get tools for creating Super Apps. See the latest technologies.
Sessions, hands-on labs, demos & much more. Register early & save!
http://p.sf.net/sfu/rim-blackberry-1
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[David Vincent-Jones]

I am using the latest ppa on Naty . ... With Highlight-reconstruction
turned on as the first item on my stack, there is absolutely no way,
that I can find, to eliminate the Pinks from an image with blown
highlights.Have there been even later changes?

David

Post by johannes hanika
hey.
On Wed, Aug 3, 2011 at 12:21 PM, David Vincent-Jones

Post by David Vincent-Jones
After a lot of (hard) work with wide dynamic range images this is my
Pink artifact result solely from totally blown highlights. dt simply
does not handle blown highlights well (at this time).

that's not true. the pipe changed so highlight clipping can now be
done before tone mapping. you have to leave highlights switched on, as
for regular raws, too. are you on the latest git master? 0.9.1 should
also include that hdr change.

Post by David Vincent-Jones
It is too easy to simply take 3 bracketed shots and assume that the
subject range requirements will be met. If any of the images include
blown highlights then the resulting DNG file will not prove useful. The
approach that I take is to ensure that the brightest image is fully
within the sensor's capability and then add darker component images.
This works well in dt using its tone-mapping with prodigious
capabilities in the low tonal range.

not true anymore, see above. the pink bva was talking about is because
the CFA (color filter array) layout was interpreted the wrong way
(e.g. rggb instead of gbrg or similar)
-jo

Post by David Vincent-Jones

Post by johannes hanika

Post by bva

Post by johannes hanika
this looks like you're getting the wrong CFA layout from somewhere.

CFA? Sorry, could you please write a long version of this
abbreviation? I have no idea what CFA means. My english is very poor.

Post by johannes hanika
does it do heavy mazing, when you zoom in?

yes.

Post by johannes hanika
can you try to delete ~/.cache/darktable/mipmaps* and try again if
it's still pink?

I deleted cache. Then I tryed to make new HDR from 3 CR2s with the
same result :(
I want to say again: in rtopencl branch I can't reproduce this with
one version of "mipmaps" file for both rtopencl & master. May be this
information can help.

can you post the output of both
git show master | head -1
and
git show rtopencl | head -1
please? i think your rtopencl branch might be out of sync with
origin/rtopencl and thus the difference. there should be no notable
difference between rtopencl and master about demosaicing things. maybe
i can use that as a base for git bisect and can find out what the hell
is going on here..
-jo
------------------------------------------------------------------------------
BlackBerry® DevCon Americas, Oct. 18-20, San Francisco, CA
The must-attend event for mobile developers. Connect with experts.
Get tools for creating Super Apps. See the latest technologies.
Sessions, hands-on labs, demos & much more. Register early & save!
http://p.sf.net/sfu/rim-blackberry-1
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[johannes hanika]

could you post darktable --version or the exact hash in the header?

On Wed, Aug 3, 2011 at 12:47 PM, David Vincent-Jones

Post by David Vincent-Jones
I am using the latest ppa on Naty . ... With Highlight-reconstruction
turned on as the first item on my stack, there is absolutely no way,
that I can find, to eliminate the Pinks from an image with blown
highlights.Have there been even later changes?
David

Post by johannes hanika
hey.
On Wed, Aug 3, 2011 at 12:21 PM, David Vincent-Jones

Post by David Vincent-Jones
After a lot of (hard) work with wide dynamic range images this is my
Pink artifact result solely from totally blown highlights. dt simply
does not handle blown highlights well (at this time).

that's not true. the pipe changed so highlight clipping can now be
done before tone mapping. you have to leave highlights switched on, as
for regular raws, too. are you on the latest git master? 0.9.1 should
also include that hdr change.

Post by David Vincent-Jones
It is too easy to simply take 3 bracketed shots and assume that the
subject range requirements will be met. If any of the images include
blown highlights then the resulting DNG file will not prove useful. The
approach that I take is to ensure that the brightest image is fully
within the sensor's capability and then add darker component images.
This works well in dt using its tone-mapping with prodigious
capabilities in the low tonal range.

not true anymore, see above. the pink bva was talking about is because
the CFA (color filter array) layout was interpreted the wrong way
(e.g. rggb instead of gbrg or similar)
-jo

Post by David Vincent-Jones

Post by johannes hanika

Post by bva

Post by johannes hanika
this looks like you're getting the wrong CFA layout from somewhere.

CFA? Sorry, could you please write a long version of this
abbreviation? I have no idea what CFA means. My english is very poor.

Post by johannes hanika
does it do heavy mazing, when you zoom in?

yes.

Post by johannes hanika
can you try to delete ~/.cache/darktable/mipmaps* and try again if
it's still pink?

I deleted cache. Then I tryed to make new HDR from 3 CR2s with the
same result :(
I want to say again: in rtopencl branch I can't reproduce this with
one version of "mipmaps" file for both rtopencl & master. May be this
information can help.

can you post the output of both
git show master | head -1
and
git show rtopencl | head -1
please? i think your rtopencl branch might be out of sync with
origin/rtopencl and thus the difference. there should be no notable
difference between rtopencl and master about demosaicing things. maybe
i can use that as a base for git bisect and can find out what the hell
is going on here..
-jo
------------------------------------------------------------------------------
BlackBerry® DevCon Americas, Oct. 18-20, San Francisco, CA
The must-attend event for mobile developers. Connect with experts.
Get tools for creating Super Apps. See the latest technologies.
Sessions, hands-on labs, demos & much more. Register early & save!
http://p.sf.net/sfu/rim-blackberry-1
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[David Vincent-Jones]

Version 0.9.1

Post by johannes hanika
could you post darktable --version or the exact hash in the header?
On Wed, Aug 3, 2011 at 12:47 PM, David Vincent-Jones

Post by David Vincent-Jones
I am using the latest ppa on Naty . ... With Highlight-reconstruction
turned on as the first item on my stack, there is absolutely no way,
that I can find, to eliminate the Pinks from an image with blown
highlights.Have there been even later changes?
David

Post by johannes hanika
hey.
On Wed, Aug 3, 2011 at 12:21 PM, David Vincent-Jones

Post by David Vincent-Jones
After a lot of (hard) work with wide dynamic range images this is my
Pink artifact result solely from totally blown highlights. dt simply
does not handle blown highlights well (at this time).

that's not true. the pipe changed so highlight clipping can now be
done before tone mapping. you have to leave highlights switched on, as
for regular raws, too. are you on the latest git master? 0.9.1 should
also include that hdr change.

Post by David Vincent-Jones
It is too easy to simply take 3 bracketed shots and assume that the
subject range requirements will be met. If any of the images include
blown highlights then the resulting DNG file will not prove useful. The
approach that I take is to ensure that the brightest image is fully
within the sensor's capability and then add darker component images.
This works well in dt using its tone-mapping with prodigious
capabilities in the low tonal range.

not true anymore, see above. the pink bva was talking about is because
the CFA (color filter array) layout was interpreted the wrong way
(e.g. rggb instead of gbrg or similar)
-jo

Post by David Vincent-Jones

Post by johannes hanika

Post by bva

Post by johannes hanika
this looks like you're getting the wrong CFA layout from somewhere.

CFA? Sorry, could you please write a long version of this
abbreviation? I have no idea what CFA means. My english is very poor.

Post by johannes hanika
does it do heavy mazing, when you zoom in?

yes.

Post by johannes hanika
can you try to delete ~/.cache/darktable/mipmaps* and try again if
it's still pink?

I deleted cache. Then I tryed to make new HDR from 3 CR2s with the
same result :(
I want to say again: in rtopencl branch I can't reproduce this with
one version of "mipmaps" file for both rtopencl & master. May be this
information can help.

can you post the output of both
git show master | head -1
and
git show rtopencl | head -1
please? i think your rtopencl branch might be out of sync with
origin/rtopencl and thus the difference. there should be no notable
difference between rtopencl and master about demosaicing things. maybe
i can use that as a base for git bisect and can find out what the hell
is going on here..
-jo
------------------------------------------------------------------------------
BlackBerry® DevCon Americas, Oct. 18-20, San Francisco, CA
The must-attend event for mobile developers. Connect with experts.
Get tools for creating Super Apps. See the latest technologies.
Sessions, hands-on labs, demos & much more. Register early & save!
http://p.sf.net/sfu/rim-blackberry-1
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[johannes hanika]

did you re-create the hdr dng since you upgraded?

On Wed, Aug 3, 2011 at 12:53 PM, David Vincent-Jones

Post by David Vincent-Jones
Version 0.9.1

Post by johannes hanika
could you post darktable --version or the exact hash in the header?
On Wed, Aug 3, 2011 at 12:47 PM, David Vincent-Jones

Post by David Vincent-Jones
I am using the latest ppa on Naty . ... With Highlight-reconstruction
turned on as the first item on my stack, there is absolutely no way,
that I can find, to eliminate the Pinks from an image with blown
highlights.Have there been even later changes?
David

Post by johannes hanika
hey.
On Wed, Aug 3, 2011 at 12:21 PM, David Vincent-Jones

Post by David Vincent-Jones
After a lot of (hard) work with wide dynamic range images this is my
Pink artifact result solely from totally blown highlights. dt simply
does not handle blown highlights well (at this time).

that's not true. the pipe changed so highlight clipping can now be
done before tone mapping. you have to leave highlights switched on, as
for regular raws, too. are you on the latest git master? 0.9.1 should
also include that hdr change.

Post by David Vincent-Jones
It is too easy to simply take 3 bracketed shots and assume that the
subject range requirements will be met. If any of the images include
blown highlights then the resulting DNG file will not prove useful. The
approach that I take is to ensure that the brightest image is fully
within the sensor's capability and then add darker component images.
This works well in dt using its tone-mapping with prodigious
capabilities in the low tonal range.

not true anymore, see above. the pink bva was talking about is because
the CFA (color filter array) layout was interpreted the wrong way
(e.g. rggb instead of gbrg or similar)
-jo

Post by David Vincent-Jones

Post by johannes hanika

Post by bva

Post by johannes hanika
this looks like you're getting the wrong CFA layout from somewhere.

CFA? Sorry, could you please write a long version of this
abbreviation? I have no idea what CFA means. My english is very poor.

Post by johannes hanika
does it do heavy mazing, when you zoom in?

yes.

Post by johannes hanika
can you try to delete ~/.cache/darktable/mipmaps* and try again if
it's still pink?

I deleted cache. Then I tryed to make new HDR from 3 CR2s with the
same result :(
I want to say again: in rtopencl branch I can't reproduce this with
one version of "mipmaps" file for both rtopencl & master. May be this
information can help.

can you post the output of both
git show master | head -1
and
git show rtopencl | head -1
please? i think your rtopencl branch might be out of sync with
origin/rtopencl and thus the difference. there should be no notable
difference between rtopencl and master about demosaicing things. maybe
i can use that as a base for git bisect and can find out what the hell
is going on here..
-jo
------------------------------------------------------------------------------
BlackBerry® DevCon Americas, Oct. 18-20, San Francisco, CA
The must-attend event for mobile developers. Connect with experts.
Get tools for creating Super Apps. See the latest technologies.
Sessions, hands-on labs, demos & much more. Register early & save!
http://p.sf.net/sfu/rim-blackberry-1
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[David Vincent-Jones]

Here are the images. HDR created today!
All only processed with highlight reco on plus exposure on auto.
The light image recovers OK simply during exposure.
The HRD picks all of the problems from the raw light image and is a
disaster.

Post by johannes hanika
did you re-create the hdr dng since you upgraded?
On Wed, Aug 3, 2011 at 12:53 PM, David Vincent-Jones

Post by David Vincent-Jones
Version 0.9.1

Post by johannes hanika
could you post darktable --version or the exact hash in the header?
On Wed, Aug 3, 2011 at 12:47 PM, David Vincent-Jones

Post by David Vincent-Jones
I am using the latest ppa on Naty . ... With Highlight-reconstruction
turned on as the first item on my stack, there is absolutely no way,
that I can find, to eliminate the Pinks from an image with blown
highlights.Have there been even later changes?
David

Post by johannes hanika
hey.
On Wed, Aug 3, 2011 at 12:21 PM, David Vincent-Jones

Post by David Vincent-Jones
After a lot of (hard) work with wide dynamic range images this is my
Pink artifact result solely from totally blown highlights. dt simply
does not handle blown highlights well (at this time).

that's not true. the pipe changed so highlight clipping can now be
done before tone mapping. you have to leave highlights switched on, as
for regular raws, too. are you on the latest git master? 0.9.1 should
also include that hdr change.

Post by David Vincent-Jones
It is too easy to simply take 3 bracketed shots and assume that the
subject range requirements will be met. If any of the images include
blown highlights then the resulting DNG file will not prove useful. The
approach that I take is to ensure that the brightest image is fully
within the sensor's capability and then add darker component images.
This works well in dt using its tone-mapping with prodigious
capabilities in the low tonal range.

not true anymore, see above. the pink bva was talking about is because
the CFA (color filter array) layout was interpreted the wrong way
(e.g. rggb instead of gbrg or similar)
-jo

Post by David Vincent-Jones

Post by johannes hanika

Post by bva

Post by johannes hanika
this looks like you're getting the wrong CFA layout from somewhere.

CFA? Sorry, could you please write a long version of this
abbreviation? I have no idea what CFA means. My english is very poor.

Post by johannes hanika
does it do heavy mazing, when you zoom in?

yes.

Post by johannes hanika
can you try to delete ~/.cache/darktable/mipmaps* and try again if
it's still pink?

I deleted cache. Then I tryed to make new HDR from 3 CR2s with the
same result :(
I want to say again: in rtopencl branch I can't reproduce this with
one version of "mipmaps" file for both rtopencl & master. May be this
information can help.

can you post the output of both
git show master | head -1
and
git show rtopencl | head -1
please? i think your rtopencl branch might be out of sync with
origin/rtopencl and thus the difference. there should be no notable
difference between rtopencl and master about demosaicing things. maybe
i can use that as a base for git bisect and can find out what the hell
is going on here..
-jo
------------------------------------------------------------------------------
BlackBerry® DevCon Americas, Oct. 18-20, San Francisco, CA
The must-attend event for mobile developers. Connect with experts.
Get tools for creating Super Apps. See the latest technologies.
Sessions, hands-on labs, demos & much more. Register early & save!
http://p.sf.net/sfu/rim-blackberry-1
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[johannes hanika]

that highlight doesn't even look overexposed? something is really very
wrong here. i have the suspicion that we don't always detect dng as
what they should be? and treat them as ldr instead or whatever.. need
to look around some more.

-jo

On Wed, Aug 3, 2011 at 1:18 PM, David Vincent-Jones

Post by David Vincent-Jones
Here are the images. HDR created today!
All only processed with highlight reco on plus exposure on auto.
The light image recovers OK simply during exposure.
The HRD picks all of the problems from the raw light image and is a
disaster.

Post by johannes hanika
did you re-create the hdr dng since you upgraded?
On Wed, Aug 3, 2011 at 12:53 PM, David Vincent-Jones

Post by David Vincent-Jones
Version 0.9.1

Post by johannes hanika
could you post darktable --version or the exact hash in the header?
On Wed, Aug 3, 2011 at 12:47 PM, David Vincent-Jones

Post by David Vincent-Jones
I am using the latest ppa on Naty . ... With Highlight-reconstruction
turned on as the first item on my stack, there is absolutely no way,
that I can find, to eliminate the Pinks from an image with blown
highlights.Have there been even later changes?
David

Post by johannes hanika
hey.
On Wed, Aug 3, 2011 at 12:21 PM, David Vincent-Jones

Post by David Vincent-Jones
After a lot of (hard) work with wide dynamic range images this is my
Pink artifact result solely from totally blown highlights. dt simply
does not handle blown highlights well (at this time).

that's not true. the pipe changed so highlight clipping can now be
done before tone mapping. you have to leave highlights switched on, as
for regular raws, too. are you on the latest git master? 0.9.1 should
also include that hdr change.

Post by David Vincent-Jones
It is too easy to simply take 3 bracketed shots and assume that the
subject range requirements will be met. If any of the images include
blown highlights then the resulting DNG file will not prove useful. The
approach that I take is to ensure that the brightest image is fully
within the sensor's capability and then add darker component images.
This works well in dt using its tone-mapping with prodigious
capabilities in the low tonal range.

not true anymore, see above. the pink bva was talking about is because
the CFA (color filter array) layout was interpreted the wrong way
(e.g. rggb instead of gbrg or similar)
-jo

Post by David Vincent-Jones

Post by johannes hanika

Post by bva

Post by johannes hanika
this looks like you're getting the wrong CFA layout from somewhere.

CFA? Sorry, could you please write a long version of this
abbreviation? I have no idea what CFA means. My english is very poor.

Post by johannes hanika
does it do heavy mazing, when you zoom in?

yes.

Post by johannes hanika
can you try to delete ~/.cache/darktable/mipmaps* and try again if
it's still pink?

I deleted cache. Then I tryed to make new HDR from 3 CR2s with the
same result :(
I want to say again: in rtopencl branch I can't reproduce this with
one version of "mipmaps" file for both rtopencl & master. May be this
information can help.

can you post the output of both
git show master | head -1
and
git show rtopencl | head -1
please? i think your rtopencl branch might be out of sync with
origin/rtopencl and thus the difference. there should be no notable
difference between rtopencl and master about demosaicing things. maybe
i can use that as a base for git bisect and can find out what the hell
is going on here..
-jo
------------------------------------------------------------------------------
BlackBerry® DevCon Americas, Oct. 18-20, San Francisco, CA
The must-attend event for mobile developers. Connect with experts.
Get tools for creating Super Apps. See the latest technologies.
Sessions, hands-on labs, demos & much more. Register early & save!
http://p.sf.net/sfu/rim-blackberry-1
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[bva]

Post by johannes hanika
git show master | head -1

commit e4912c6fa7d69228d809a819cabecd64ec6910a9

Post by johannes hanika
git show rtopencl | head -1

commit 1abec226a4fda21ddaf60b6e1e3d88107fa8e4ad

Post by johannes hanika
your image shows completely different artifacts, not sure if it's

I agre with you. I had tried to make HDR from only two pictures
(under- and normal- exposed) and got exactly the same result as in my
first message. So, David, my problem and your's are different.

[johannes hanika]

bva,

i think i have fixed your problem in latest git. can you confirm?

jo

Post by bva

Post by johannes hanika
git show master | head -1

commit e4912c6fa7d69228d809a819cabecd64ec6910a9

Post by johannes hanika
git show rtopencl | head -1

commit 1abec226a4fda21ddaf60b6e1e3d88107fa8e4ad

Post by johannes hanika
your image shows completely different artifacts, not sure if it's

I agre with you. I had tried to make HDR from only two pictures
(under- and normal- exposed) and got exactly the same result as in my
first message. So, David, my problem and your's are different.

[bva]

This problem is gone! Huge thank you! :)

Now, no pink image, no maze. I'm not shure (cause my test pictures was
captured without tripod, I need to re-shot this), but it seems like
now I have the same problem as David describes: pinky white areas.
This is my first HDR try and I'm not shure: may be this is ok and
should be removed on tone mapping phase?

Another one thing about HDR but it is a bit offtopic. After pressing
"create hdr" button DT informs me that DNG file just created. But it
not always appears in the filmroll. Dependency is not clear for me
right now. Some times it appears, sometimes not.

Post by johannes hanika
bva,
i think i have fixed your problem in latest git. can you confirm?
jo

Post by bva

Post by johannes hanika
git show master | head -1

commit e4912c6fa7d69228d809a819cabecd64ec6910a9

Post by johannes hanika
git show rtopencl | head -1

commit 1abec226a4fda21ddaf60b6e1e3d88107fa8e4ad

Post by johannes hanika
your image shows completely different artifacts, not sure if it's

I agre with you. I had tried to make HDR from only two pictures
(under- and normal- exposed) and got exactly the same result as in my
first message. So, David, my problem and your's are different.

[johannes hanika]

Post by bva
This problem is gone! Huge thank you! :)
Now, no pink image, no maze. I'm not shure (cause my test pictures was
captured without tripod, I need to re-shot this), but it seems like
now I have the same problem as David describes: pinky white areas.
This is my first HDR try and I'm not shure: may be this is ok and
should be removed on tone mapping phase?

i can absolutely not reproduce. you sure you switched on the
highlights plugin? i.e. left it switched on?

Post by bva
Another one thing about HDR but it is a bit offtopic. After pressing
"create hdr" button DT informs me that DNG file just created. But it
not always appears in the filmroll. Dependency is not clear for me
right now. Some times it appears, sometimes not.

filter settings? sort order?

-jo

Post by bva

Post by johannes hanika
bva,
i think i have fixed your problem in latest git. can you confirm?
jo

Post by bva

Post by johannes hanika
git show master | head -1

commit e4912c6fa7d69228d809a819cabecd64ec6910a9

Post by johannes hanika
git show rtopencl | head -1

commit 1abec226a4fda21ddaf60b6e1e3d88107fa8e4ad

Post by johannes hanika
your image shows completely different artifacts, not sure if it's

I agre with you. I had tried to make HDR from only two pictures
(under- and normal- exposed) and got exactly the same result as in my
first message. So, David, my problem and your's are different.

[bva]

Post by johannes hanika

Post by bva
This problem is gone! Huge thank you! :)
Now, no pink image, no maze. I'm not shure (cause my test pictures was
captured without tripod, I need to re-shot this), but it seems like
now I have the same problem as David describes: pinky white areas.
This is my first HDR try and I'm not shure: may be this is ok and
should be removed on tone mapping phase?

i can absolutely not reproduce. you sure you switched on the
highlights plugin? i.e. left it switched on?

On is only base curve, tone mapping, sharpen.

Post by johannes hanika

Post by bva
Another one thing about HDR but it is a bit offtopic. After pressing
"create hdr" button DT informs me that DNG file just created. But it
not always appears in the filmroll. Dependency is not clear for me
right now. Some times it appears, sometimes not.

filter settings? sort order?

Sort order. I am sorry :)

[johannes hanika]

Post by bva

Post by johannes hanika

Post by bva
This problem is gone! Huge thank you! :)
Now, no pink image, no maze. I'm not shure (cause my test pictures was
captured without tripod, I need to re-shot this), but it seems like
now I have the same problem as David describes: pinky white areas.
This is my first HDR try and I'm not shure: may be this is ok and
should be removed on tone mapping phase?

i can absolutely not reproduce. you sure you switched on the
highlights plugin? i.e. left it switched on?

On is only base curve, tone mapping, sharpen.

so switch highlights back on!

:)

Post by bva

Post by johannes hanika

Post by bva
Another one thing about HDR but it is a bit offtopic. After pressing
"create hdr" button DT informs me that DNG file just created. But it
not always appears in the filmroll. Dependency is not clear for me
right now. Some times it appears, sometimes not.

filter settings? sort order?

Sort order. I am sorry :)

[bva]

I have some free time to walk with tripod and try to do some HDR
shots. As you know, the problem with pink maze was fixed.

But now I have the same (I think so) problem as David. Whole image is
like it's covered with a pink film.

As David say, this is problem appears only when one of the images in
the set has overexposed areas.

Post by johannes hanika

Post by bva

Post by johannes hanika

Post by bva
This problem is gone! Huge thank you! :)
Now, no pink image, no maze. I'm not shure (cause my test pictures was
captured without tripod, I need to re-shot this), but it seems like
now I have the same problem as David describes: pinky white areas.
This is my first HDR try and I'm not shure: may be this is ok and
should be removed on tone mapping phase?

i can absolutely not reproduce. you sure you switched on the
highlights plugin? i.e. left it switched on?

On is only base curve, tone mapping, sharpen.

so switch highlights back on!
:)

Post by bva

Post by johannes hanika

Post by bva
Another one thing about HDR but it is a bit offtopic. After pressing
"create hdr" button DT informs me that DNG file just created. But it
not always appears in the filmroll. Dependency is not clear for me
right now. Some times it appears, sometimes not.

filter settings? sort order?

Sort order. I am sorry :)

[bva]

Hmm... Take a look to the 2nd screenshot. The preview is ok! And when
I moving image (zoomed in for a bit) in main view, it becomes normal
for a short time (while "working.." shown) and then, when "working.."
message is dissapears, it become pink again.

Post by bva
I have some free time to walk with tripod and try to do some HDR
shots. As you know, the problem with pink maze was fixed.
But now I have the same (I think so) problem as David. Whole image is
like it's covered with a pink film.
As David say, this is problem appears only when one of the images in
the set has overexposed areas.

Post by johannes hanika

Post by bva

Post by johannes hanika

Post by bva
This problem is gone! Huge thank you! :)
Now, no pink image, no maze. I'm not shure (cause my test pictures was
captured without tripod, I need to re-shot this), but it seems like
now I have the same problem as David describes: pinky white areas.
This is my first HDR try and I'm not shure: may be this is ok and
should be removed on tone mapping phase?

i can absolutely not reproduce. you sure you switched on the
highlights plugin? i.e. left it switched on?

On is only base curve, tone mapping, sharpen.

so switch highlights back on!
:)

Post by bva

Post by johannes hanika

Post by bva
Another one thing about HDR but it is a bit offtopic. After pressing
"create hdr" button DT informs me that DNG file just created. But it
not always appears in the filmroll. Dependency is not clear for me
right now. Some times it appears, sometimes not.

filter settings? sort order?

Sort order. I am sorry :)

[Tobias Ellinghaus]

Post by bva
Hmm... Take a look to the 2nd screenshot. The preview is ok! And when
I moving image (zoomed in for a bit) in main view, it becomes normal
for a short time (while "working.." shown) and then, when "working.."
message is dissapears, it become pink again.

Just for the record: This sounds exactly like the behaviour I encountered when
I tried to implement the invert plugin for raw images.

Tobias

Post by bva

Post by bva
I have some free time to walk with tripod and try to do some HDR
shots. As you know, the problem with pink maze was fixed.
But now I have the same (I think so) problem as David. Whole image is
like it's covered with a pink film.
As David say, this is problem appears only when one of the images in
the set has overexposed areas.

Post by johannes hanika

Post by bva

Post by johannes hanika

Post by bva
This problem is gone! Huge thank you! :)
Now, no pink image, no maze. I'm not shure (cause my test pictures
was captured without tripod, I need to re-shot this), but it seems
like now I have the same problem as David describes: pinky white
areas. This is my first HDR try and I'm not shure: may be this is ok
and should be removed on tone mapping phase?

i can absolutely not reproduce. you sure you switched on the
highlights plugin? i.e. left it switched on?

On is only base curve, tone mapping, sharpen.

so switch highlights back on!
:)

Post by bva

Post by johannes hanika

Post by bva
Another one thing about HDR but it is a bit offtopic. After pressing
"create hdr" button DT informs me that DNG file just created. But it
not always appears in the filmroll. Dependency is not clear for me
right now. Some times it appears, sometimes not.

filter settings? sort order?

Sort order. I am sorry :)

[johannes hanika]

do you have a few sample shots for that? i think your white point
clipping value might be too high. that doesn't seem to be the case for
my cam, i can't reproduce such behaviour.

-jo

Post by bva
I have some free time to walk with tripod and try to do some HDR
shots. As you know, the problem with pink maze was fixed.
But now I have the same (I think so) problem as David. Whole image is
like it's covered with a pink film.
As David say, this is problem appears only when one of the images in
the set has overexposed areas.

Post by johannes hanika

Post by bva

Post by johannes hanika

Post by bva
This problem is gone! Huge thank you! :)
Now, no pink image, no maze. I'm not shure (cause my test pictures was
captured without tripod, I need to re-shot this), but it seems like
now I have the same problem as David describes: pinky white areas.
This is my first HDR try and I'm not shure: may be this is ok and
should be removed on tone mapping phase?

i can absolutely not reproduce. you sure you switched on the
highlights plugin? i.e. left it switched on?

On is only base curve, tone mapping, sharpen.

so switch highlights back on!
:)

Post by bva

Post by johannes hanika

Post by bva
Another one thing about HDR but it is a bit offtopic. After pressing
"create hdr" button DT informs me that DNG file just created. But it
not always appears in the filmroll. Dependency is not clear for me
right now. Some times it appears, sometimes not.

filter settings? sort order?

Sort order. I am sorry :)

[bva]

Post by johannes hanika
do you have a few sample shots for that? i think your white point
clipping value might be too high. that doesn't seem to be the case for
my cam, i can't reproduce such behaviour.

I will check clipping and send you test images tomorrow cause today I
am away from my home.

[johannes hanika]

Post by bva

Post by johannes hanika
do you have a few sample shots for that? i think your white point
clipping value might be too high. that doesn't seem to be the case for
my cam, i can't reproduce such behaviour.

I will check clipping and send you test images tomorrow cause today I
am away from my home.

clipping would be the white point in cameras.xml. you can try to lower
the value for your camera there and see how it goes. tone mapping
would expose the pink artifact way more than the regular pipeline, so
you might not have noticed otherwise.

-jo

[Ulrich Pegelow]

Post by johannes hanika

Post by bva

Post by johannes hanika
do you have a few sample shots for that? i think your white point
clipping value might be too high. that doesn't seem to be the case for
my cam, i can't reproduce such behaviour.

I will check clipping and send you test images tomorrow cause today I
am away from my home.

clipping would be the white point in cameras.xml. you can try to lower
the value for your camera there and see how it goes. tone mapping
would expose the pink artifact way more than the regular pipeline, so
you might not have noticed otherwise.

Hi,

if I got all points right in this thread, there was an observation that
tonemap worked well for some period of time in branch rtopencl. I assume that
this might have to to with handling of processed_maximum[3]. Tiling (which was
implemented in rtopencl branch) did not handle this parameters correctly in
the beginning. Consequence was that a recalculation was done repetitively for
each tile, typically in the sense of processed_maximum[k] *= some_factor[k].

I fixed that around August 4th, with a further fix yesterday.

Maybe this gives a hint what is going wrong here.

BTW I also see issues with tonemap. Full image is just plain white or gray.
Preview image is fine. And I also have some overexposed parts in some of the
images.

Ulrich

[johannes hanika]

On Sat, Aug 20, 2011 at 8:33 PM, Ulrich Pegelow

Post by Ulrich Pegelow

Post by johannes hanika

Post by bva

Post by johannes hanika
do you have a few sample shots for that? i think your white point
clipping value might be too high. that doesn't seem to be the case for
my cam, i can't reproduce such behaviour.

I will check clipping and send you test images tomorrow cause today I
am away from my home.

clipping would be the white point in cameras.xml. you can try to lower
the value for your camera there and see how it goes. tone mapping
would expose the pink artifact way more than the regular pipeline, so
you might not have noticed otherwise.

Hi,
if I got all points right in this thread, there was an observation that
tonemap worked well for some period of time in branch rtopencl. I assume that
this might have to to with handling of processed_maximum[3]. Tiling (which was
implemented in rtopencl branch) did not handle this parameters correctly in
the beginning. Consequence was that a recalculation was done repetitively for
each tile, typically in the sense of processed_maximum[k] *= some_factor[k].
I fixed that around August 4th, with a further fix yesterday.
Maybe this gives a hint what is going wrong here.

the processed_maximum[] has only to be carried through the pipe until
highlight clipping has taken place. that's before tonemapping. could
still be the cause if the threshold was lowered significantly by some
other plugin on the way.

Post by Ulrich Pegelow
BTW I also see issues with tonemap. Full image is just plain white or gray.
Preview image is fine. And I also have some overexposed parts in some of the
images.

[David Vincent-Jones]

My current test comments:

1.) On the 3 test images that I have used as a primary HDR test, none
of the raw base images *now* displays any pink tint whatsoever. The pink
only appears as artifact on the created dng.

2.) With a water image (single shot) with specular highlight problems
the current 'unstable' ppa totally eliminates the pink problem (ie.
problem now appears solved for single images).

My conclusion is that the pink problem is now solely being created
during the dng creation.

David

Post by johannes hanika
do you have a few sample shots for that? i think your white point
clipping value might be too high. that doesn't seem to be the case for
my cam, i can't reproduce such behaviour.
-jo

Post by bva
I have some free time to walk with tripod and try to do some HDR
shots. As you know, the problem with pink maze was fixed.
But now I have the same (I think so) problem as David. Whole image is
like it's covered with a pink film.
As David say, this is problem appears only when one of the images in
the set has overexposed areas.

Post by johannes hanika

Post by bva

Post by johannes hanika

Post by bva
This problem is gone! Huge thank you! :)
Now, no pink image, no maze. I'm not shure (cause my test pictures was
captured without tripod, I need to re-shot this), but it seems like
now I have the same problem as David describes: pinky white areas.
This is my first HDR try and I'm not shure: may be this is ok and
should be removed on tone mapping phase?

i can absolutely not reproduce. you sure you switched on the
highlights plugin? i.e. left it switched on?

On is only base curve, tone mapping, sharpen.

so switch highlights back on!
:)

Post by bva

Post by johannes hanika

Post by bva
Another one thing about HDR but it is a bit offtopic. After pressing
"create hdr" button DT informs me that DNG file just created. But it
not always appears in the filmroll. Dependency is not clear for me
right now. Some times it appears, sometimes not.

filter settings? sort order?

Sort order. I am sorry :)

------------------------------------------------------------------------------
Get a FREE DOWNLOAD! and learn more about uberSVN rich system,
user administration capabilities and model configuration. Take
the hassle out of deploying and managing Subversion and the
tools developers use with it. http://p.sf.net/sfu/wandisco-d2d-2
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[bva]

There is my test RAW set: http://ge.tt/8rNW807?c

So, the reason is a wrong "white" parameter value in cameras.xml.
Original value is 16383. Image is looking good with values about
13000, but overexposed areas are still pink. Example with value 13010
is attached.
How this values in cameras.xml are calculated or measured? Hit and
miss tactics is too long and stupid.

Is this problem have a chance to be fixed before 0.9.2?

1.)   On the 3 test images that I have used as a primary HDR test, none
of the raw base images *now* displays any pink tint whatsoever. The pink
only appears as artifact on the created dng.
2.)   With a water image (single shot) with specular highlight problems
the current 'unstable' ppa totally eliminates the pink problem (ie.
problem now appears solved for single images).
My conclusion is that the pink problem is now solely  being created
during the dng creation.
David

Post by johannes hanika
do you have a few sample shots for that? i think your white point
clipping value might be too high. that doesn't seem to be the case for
my cam, i can't reproduce such behaviour.
-jo

Post by bva
I have some free time to walk with tripod and try to do some HDR
shots. As you know, the problem with pink maze was fixed.
But now I have the same (I think so) problem as David. Whole image is
like it's covered with a pink film.
As David say, this is problem appears only when one of the images in
the set has overexposed areas.

Post by johannes hanika

Post by bva

Post by johannes hanika

Post by bva
This problem is gone! Huge thank you! :)
Now, no pink image, no maze. I'm not shure (cause my test pictures was
captured without tripod, I need to re-shot this), but it seems like
now I have the same problem as David describes: pinky white areas.
This is my first HDR try and I'm not shure: may be this is ok and
should be removed on tone mapping phase?

i can absolutely not reproduce. you sure you switched on the
highlights plugin? i.e. left it switched on?

On is only base curve, tone mapping, sharpen.

so switch highlights back on!
:)

Post by bva

Post by johannes hanika

Post by bva
Another one thing about HDR but it is a bit offtopic. After pressing
"create hdr" button DT informs me that DNG file just created. But it
not always appears in the filmroll. Dependency is not clear for me
right now. Some times it appears, sometimes not.

filter settings? sort order?

Sort order. I am sorry :)

------------------------------------------------------------------------------
Get a FREE DOWNLOAD! and learn more about uberSVN rich system,
user administration capabilities and model configuration. Take
the hassle out of deploying and managing Subversion and the
tools developers use with it. http://p.sf.net/sfu/wandisco-d2d-2
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[bva]

And a quick quesstion (but a bit offtopic).

Is these aoutlines are a sort of artifacts or just a simple tripod fail?

[johannes hanika]

i would guess the oversharp edges (gradient reversals) are due to the
bilateral filter used for tonemapping, which sometimes overzealously
detects edges.

the red/green would be chromatic aberration, see the lens plugin or
the plugin by exactly that name for a fix.

-jo

Post by bva
And a quick quesstion (but a bit offtopic).
Is these aoutlines are a sort of artifacts or just a simple tripod fail?
------------------------------------------------------------------------------
Get a FREE DOWNLOAD! and learn more about uberSVN rich system,
user administration capabilities and model configuration. Take
the hassle out of deploying and managing Subversion and the
tools developers use with it. http://p.sf.net/sfu/wandisco-d2d-2
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[Jan]

Post by bva
And a quick quesstion (but a bit offtopic).
Is these aoutlines are a sort of artifacts or just a simple tripod fail?

Compared to 0.9.1, which produced completely pink casted images for me, the HDR
functionality seems to have improved a lot in current version, thanks! :)

But: I still see strong pink colors at hard edges. It's similar to the
image by bva, but much stronger. Please see the following exported image:

Loading Image...

What could be the reason for this?
I can provide the raw files of the series, if someone wants to check them.

[johannes hanika]

Post by Jan

Post by bva
And a quick quesstion (but a bit offtopic).
Is these aoutlines are a sort of artifacts or just a simple tripod fail?

Compared to 0.9.1, which produced completely pink casted images for me, the HDR
functionality seems to have improved a lot in current version, thanks! :)
But: I still see strong pink colors at hard edges. It's similar to the
http://www.abload.de/image.php?img=img_0001m7og.jpg
What could be the reason for this?
I can provide the raw files of the series, if someone wants to check them.

this does look like a tripod fail (see the wooden planks on top of that wall?).

unfortunately our hdr merger is very, very silly. it does not align
anything .. henrik's version might be smarter (or at least more robust
to that due to some smoothed blending)

j.

[Jan]

Post by johannes hanika

Post by Jan
What could be the reason for this?
I can provide the raw files of the series, if someone wants to check them.

this does look like a tripod fail (see the wooden planks on top of that wall?).

You are right. Will try with an aligned series later.

But even without 100%-perfect alignment, the HDR might be usable if it
hadn't those pink shades :)

[johannes hanika]

Post by bva
There is my test RAW set: http://ge.tt/8rNW807?c
So, the reason is a wrong "white" parameter value in cameras.xml.
Original value is 16383. Image is looking good with values about
13000, but overexposed areas are still pink. Example with value 13010
is attached.

16383 sounds dangerously close to 2^14 == 16384 .. so i think that
might be wrong/just a first guess, and just never caused very visible
problems so far. did you never get visible purple fringes around
overexposed areas? like a tree against the sun, or a window border?

i didn't quite follow you: you're saying 13000 is all good, or 13000
is still broken for very bright highlights? in the latter case we'll
have to double check the dng creation in
src/control/jobs/control_jobs.c, if the weighting of overexposed
pixels is correct.

-jo

Post by bva
How this values in cameras.xml are calculated or measured? Hit and
miss tactics is too long and stupid.
Is this problem have a chance to be fixed before 0.9.2?

1.)   On the 3 test images that I have used as a primary HDR test, none
of the raw base images *now* displays any pink tint whatsoever. The pink
only appears as artifact on the created dng.
2.)   With a water image (single shot) with specular highlight problems
the current 'unstable' ppa totally eliminates the pink problem (ie.
problem now appears solved for single images).
My conclusion is that the pink problem is now solely  being created
during the dng creation.
David

Post by johannes hanika
do you have a few sample shots for that? i think your white point
clipping value might be too high. that doesn't seem to be the case for
my cam, i can't reproduce such behaviour.
-jo

Post by bva
I have some free time to walk with tripod and try to do some HDR
shots. As you know, the problem with pink maze was fixed.
But now I have the same (I think so) problem as David. Whole image is
like it's covered with a pink film.
As David say, this is problem appears only when one of the images in
the set has overexposed areas.

Post by johannes hanika

Post by bva

Post by johannes hanika

Post by bva
This problem is gone! Huge thank you! :)
Now, no pink image, no maze. I'm not shure (cause my test pictures was
captured without tripod, I need to re-shot this), but it seems like
now I have the same problem as David describes: pinky white areas.
This is my first HDR try and I'm not shure: may be this is ok and
should be removed on tone mapping phase?

i can absolutely not reproduce. you sure you switched on the
highlights plugin? i.e. left it switched on?

On is only base curve, tone mapping, sharpen.

so switch highlights back on!
:)

Post by bva

Post by johannes hanika

Post by bva
Another one thing about HDR but it is a bit offtopic. After pressing
"create hdr" button DT informs me that DNG file just created. But it
not always appears in the filmroll. Dependency is not clear for me
right now. Some times it appears, sometimes not.

filter settings? sort order?

Sort order. I am sorry :)

------------------------------------------------------------------------------
Get a FREE DOWNLOAD! and learn more about uberSVN rich system,
user administration capabilities and model configuration. Take
the hassle out of deploying and managing Subversion and the
tools developers use with it. http://p.sf.net/sfu/wandisco-d2d-2
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[bva]

Post by johannes hanika

Post by bva
There is my test RAW set: http://ge.tt/8rNW807?c
So, the reason is a wrong "white" parameter value in cameras.xml.
Original value is 16383. Image is looking good with values about
13000, but overexposed areas are still pink. Example with value 13010
is attached.

16383 sounds dangerously close to 2^14 == 16384 .. so i think that
might be wrong/just a first guess, and just never caused very visible
problems so far. did you never get visible purple fringes around
overexposed areas? like a tree against the sun, or a window border?

Sometimes I have something like a purple thread around the edges of
contrast areas (like a tree against the sun). But I think these are a
chromatic aberrations. This effect is easily can be removed by CA
removing plugin.

Post by johannes hanika
i didn't quite follow you: you're saying 13000 is all good, or 13000
is still broken for very bright highlights? in the latter case we'll
have to double check the dng creation in
src/control/jobs/control_jobs.c, if the weighting of overexposed
pixels is correct.

When I say "13000 is looking good" I mean that the pink tint is
disappeared. But pink overexposed areas are still here (like a roof of
a cars on the screenshot).

Post by johannes hanika
-jo

Post by bva
How this values in cameras.xml are calculated or measured? Hit and
miss tactics is too long and stupid.
Is this problem have a chance to be fixed before 0.9.2?

1.)   On the 3 test images that I have used as a primary HDR test, none
of the raw base images *now* displays any pink tint whatsoever. The pink
only appears as artifact on the created dng.
2.)   With a water image (single shot) with specular highlight problems
the current 'unstable' ppa totally eliminates the pink problem (ie.
problem now appears solved for single images).
My conclusion is that the pink problem is now solely  being created
during the dng creation.
David

Post by johannes hanika
do you have a few sample shots for that? i think your white point
clipping value might be too high. that doesn't seem to be the case for
my cam, i can't reproduce such behaviour.
-jo

Post by bva
I have some free time to walk with tripod and try to do some HDR
shots. As you know, the problem with pink maze was fixed.
But now I have the same (I think so) problem as David. Whole image is
like it's covered with a pink film.
As David say, this is problem appears only when one of the images in
the set has overexposed areas.

Post by johannes hanika

Post by bva

Post by johannes hanika

Post by bva
This problem is gone! Huge thank you! :)
Now, no pink image, no maze. I'm not shure (cause my test pictures was
captured without tripod, I need to re-shot this), but it seems like
now I have the same problem as David describes: pinky white areas.
This is my first HDR try and I'm not shure: may be this is ok and
should be removed on tone mapping phase?

i can absolutely not reproduce. you sure you switched on the
highlights plugin? i.e. left it switched on?

On is only base curve, tone mapping, sharpen.

so switch highlights back on!
:)

Post by bva

Post by johannes hanika

Post by bva
Another one thing about HDR but it is a bit offtopic. After pressing
"create hdr" button DT informs me that DNG file just created. But it
not always appears in the filmroll. Dependency is not clear for me
right now. Some times it appears, sometimes not.

filter settings? sort order?

Sort order. I am sorry :)

------------------------------------------------------------------------------
Get a FREE DOWNLOAD! and learn more about uberSVN rich system,
user administration capabilities and model configuration. Take
the hassle out of deploying and managing Subversion and the
tools developers use with it. http://p.sf.net/sfu/wandisco-d2d-2
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[johannes hanika]

..could it be you only get that for portrait orientation?

Post by bva
When I try to create HDR from 3 CR2 images the resulting DNG is in
pink colors (see attachment). The same result with tone mapping on or
off. All other modules are off. I'm doing something wrong or this is a
bug?
darktable --version
this is darktable 0.9.1+85~ge4912c6
------------------------------------------------------------------------------
Got Input?   Slashdot Needs You.
Take our quick survey online.  Come on, we don't ask for help often.
Plus, you'll get a chance to win $100 to spend on ThinkGeek.
http://p.sf.net/sfu/slashdot-survey
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[David Vincent-Jones]

The set I sent to you were in landscape mode.

When I made a dng from just 2 images, (the normal and the dark) the
results were just fine but as soon as I introduced the third image with
the blown highlight it appeared to 'infect' the entire set. This looks
suspiciously to me as if the DNG processing could be the problem.

Keep in mind that the bright image, processed by itself, can be easily
'tamed' (obviously with loss of highlight detail) but tamed without
showing the pink cast.

David

Post by johannes hanika
..could it be you only get that for portrait orientation?

[johannes hanika]

On Wed, Aug 3, 2011 at 4:16 PM, David Vincent-Jones

Post by David Vincent-Jones
The set I sent to you were in landscape mode.

your image shows completely different artifacts, not sure if it's
related to the OP's problem.

Post by David Vincent-Jones
When I made a dng from just 2 images, (the normal and the dark) the
results were just fine but as soon as I introduced the third image with
the blown highlight it appeared to 'infect' the entire set. This looks
suspiciously to me as if the DNG processing could be the problem.
Keep in mind that the bright image, processed by itself, can be easily
'tamed' (obviously with loss of highlight detail) but tamed without
showing the pink cast.
David

Post by johannes hanika
..could it be you only get that for portrait orientation?

[David Vincent-Jones]

Post by johannes hanika
..could it be you only get that for portrait orientation?

Extra information:
I notice now that the jpg that I sent to you from the 'light-image'
contains some artifact on part of the frame. Not sure what that is ...
never seen anything like that on my normal png output. I will follow up
on that tomorrow.

David

[bva]

johannes, I will test all what you asked before when I will get home
in ~30 hours.

Post by David Vincent-Jones

Post by johannes hanika
..could it be you only get that for portrait orientation?

I notice now that the jpg that I sent to you from the 'light-image'
contains some artifact on part of the frame. Not sure what that is ...
never seen anything like that on my normal png output. I will follow up
on that tomorrow.
David

[johannes hanika]

On Wed, Aug 3, 2011 at 6:21 AM, David Vincent-Jones

Post by David Vincent-Jones

Post by johannes hanika
..could it be you only get that for portrait orientation?

I notice now that the jpg that I sent to you from the 'light-image'
contains some artifact on part of the frame. Not sure what that is ...
never seen anything like that on my normal png output. I will follow up
on that tomorrow.

like a mazing artifact? when zoomed in 1:1? i'm not quite sure what i
see in your images, the screen captures don't tell me anything. could
you try again with the fix in current git and see if it goes away,
too? or else tell me how to reproduce/send me a test image.

thanks
jo

[David Vincent-Jones]

Unfortunately I am running the ppa version ... tried to follow the git
install instructions and there appears to be something missing. Would be
nice if that worked as written!

I will however make some further tests ... this appears to only be a
problem that I have seen on the single jpg ... my normal output is png
and those images all appear to be clean.

David

Post by johannes hanika
On Wed, Aug 3, 2011 at 6:21 AM, David Vincent-Jones

Post by David Vincent-Jones

Post by johannes hanika
..could it be you only get that for portrait orientation?

I notice now that the jpg that I sent to you from the 'light-image'
contains some artifact on part of the frame. Not sure what that is ...
never seen anything like that on my normal png output. I will follow up
on that tomorrow.

like a mazing artifact? when zoomed in 1:1? i'm not quite sure what i
see in your images, the screen captures don't tell me anything. could
you try again with the fix in current git and see if it goes away,
too? or else tell me how to reproduce/send me a test image.
thanks
jo

[David Vincent-Jones]

Further testing:

Used the same image set: Each of the 3 images were initially tested to
look at the extent of the 'pink' problem. All could be easily adjusted
with a minimum (to zero) showing of the problem. ... The 3 images then
were made into a dng and exposure shifted so that no r,g,b value
exceeded 240. Result very bad 'pink' all over the image.
Finally made the attached jpg that is not showing the prior artifact so
at least that problem appears to have been something odd and non
repeatable.

If the HRD process has been revamped I hope that those changes will
show-up within the next ppa.

If there are more up-to-date instructions for the git that includes all
needed lib files I will have a crack at a git install again.

David

Post by johannes hanika
On Wed, Aug 3, 2011 at 6:21 AM, David Vincent-Jones

Post by David Vincent-Jones

Post by johannes hanika
..could it be you only get that for portrait orientation?

I notice now that the jpg that I sent to you from the 'light-image'
contains some artifact on part of the frame. Not sure what that is ...
never seen anything like that on my normal png output. I will follow up
on that tomorrow.

like a mazing artifact? when zoomed in 1:1? i'm not quite sure what i
see in your images, the screen captures don't tell me anything. could
you try again with the fix in current git and see if it goes away,
too? or else tell me how to reproduce/send me a test image.
thanks
jo

[Alexandre Prokoudine]

Post by bva
When I try to create HDR from 3 CR2 images the resulting DNG is in
pink colors (see attachment). The same result with tone mapping on or
off. All other modules are off. I'm doing something wrong or this is a
bug?

Just for the heck of it, here is my example:

Loading Image...

Thus far only enfuse provides a meaningful output, so I really hope
dinamic uploads his fusion plug-in soon :)

Alexandre Prokoudine
http://libregraphicsworld.org

[johannes hanika]

if i had time to fix it i would really need a raw sample/exposure
series. because with my raws it works just fine. so i'm guessing the
white level for your cameras are too high, to not lose detail.

purple fringes around overexposed areas are not C/A, no matter how
easy you can remove them..

-j.

On Sun, Aug 21, 2011 at 8:22 PM, Alexandre Prokoudine

Post by Alexandre Prokoudine

Post by bva
When I try to create HDR from 3 CR2 images the resulting DNG is in
pink colors (see attachment). The same result with tone mapping on or
off. All other modules are off. I'm doing something wrong or this is a
bug?

http://i.imgur.com/VeaaH.jpg
Thus far only enfuse provides a meaningful output, so I really hope
dinamic uploads his fusion plug-in soon :)
Alexandre Prokoudine
http://libregraphicsworld.org
------------------------------------------------------------------------------
Get a FREE DOWNLOAD! and learn more about uberSVN rich system,
user administration capabilities and model configuration. Take
the hassle out of deploying and managing Subversion and the
tools developers use with it. http://p.sf.net/sfu/wandisco-d2d-2
_______________________________________________
darktable-devel mailing list
https://lists.sourceforge.net/lists/listinfo/darktable-devel

[bva]

Post by johannes hanika
if i had time to fix it i would really need a raw sample/exposure
series. because with my raws it works just fine. so i'm guessing the
white level for your cameras are too high, to not lose detail.

I can provide you as many series as you want. Please, tell me if
something special is needed from the shot (e.g. overexposed areas/high
contrast shapes/something else).

Post by johannes hanika
purple fringes around overexposed areas are not C/A, no matter how
easy you can remove them..

I am glad that this is not fault of my lenses :)

[Alexandre Prokoudine]

Post by johannes hanika
if i had time to fix it i would really need a raw sample/exposure
series. because with my raws it works just fine.

I can upload it for you

Post by johannes hanika
so i'm guessing the white level for your cameras are too high, to not lose detail.

Dunno. UFRaw converted JPEGs worked for enfuse just fine back when I
shot it (around 2008). No purple madness :)

Alexandre Prokoudine
http://libregraphicsworld.org

[johannes hanika]

On Sun, Aug 21, 2011 at 8:56 PM, Alexandre Prokoudine

Post by Alexandre Prokoudine

Post by johannes hanika
if i had time to fix it i would really need a raw sample/exposure
series. because with my raws it works just fine.

I can upload it for you

Post by johannes hanika
so i'm guessing the white level for your cameras are too high, to not lose detail.

Dunno. UFRaw converted JPEGs worked for enfuse just fine back when I
shot it (around 2008). No purple madness :)

come on, that's almost like using gif to store your images :)

where are the raw files?

[Alexandre Prokoudine]

Post by johannes hanika
where are the raw files?

Uploaded them for dinamic last night :)

http://linuxgraphics.ru/darktable/bracketed.tar.bz2

Alexandre Prokoudine
http://libregraphicsworld.org

[bva]

Post by Alexandre Prokoudine

Post by johannes hanika
where are the raw files?

Uploaded them for dinamic last night :)
http://linuxgraphics.ru/darktable/bracketed.tar.bz2

And from me: http://ge.tt/9i1Fx17?c

[johannes hanika]

Post by bva

Post by Alexandre Prokoudine

Post by johannes hanika
where are the raw files?

Uploaded them for dinamic last night :)
http://linuxgraphics.ru/darktable/bracketed.tar.bz2

And from me: http://ge.tt/9i1Fx17?c

cool, these i can get :)

[johannes hanika]

On Mon, Aug 22, 2011 at 9:41 PM, Alexandre Prokoudine

Post by Alexandre Prokoudine

Post by johannes hanika
where are the raw files?

Uploaded them for dinamic last night :)
http://linuxgraphics.ru/darktable/bracketed.tar.bz2

404

[Alexandre Prokoudine]

Post by johannes hanika
On Mon, Aug 22, 2011 at 9:41 PM, Alexandre Prokoudine

Post by Alexandre Prokoudine

Post by johannes hanika
where are the raw files?

Uploaded them for dinamic last night :)
http://linuxgraphics.ru/darktable/bracketed.tar.bz2

404

http://linuxgraphics.ru/files/darktable/bracketed.tar.bz2

Alexandre Prokoudine
http://libregraphicsworld.org

[johannes hanika]

On Mon, Aug 22, 2011 at 10:17 PM, Alexandre Prokoudine

Post by Alexandre Prokoudine

Post by johannes hanika
On Mon, Aug 22, 2011 at 9:41 PM, Alexandre Prokoudine

Post by Alexandre Prokoudine

Post by johannes hanika
where are the raw files?

Uploaded them for dinamic last night :)
http://linuxgraphics.ru/darktable/bracketed.tar.bz2

404

http://linuxgraphics.ru/files/darktable/bracketed.tar.bz2

:)