[Digikam-devel] New image editor noise reduction plugin...

[Gilles Caulier-2]

Hi all,

I have recently backported the new despeckle algorithm from gimp 2.2.x to
digiKam like noise reduction tool with support of 16 bits images.

After to have played with it, i have been unsatisfied... The result are a
little bit strange, especially with recursive option.

After to have serching on the web the best NR technics, my conclusion is that
this algorithm is a wrong way. despeckle use RGB color space to fix noise and
this is a wrong issue.

After to any investigation, i have found a great GPL plugin tool named
dcamnoise2. The code is really impressive and very well documented...

I have backported the code this week end, and the result look great :

http://digikam3rdparty.free.fr/Screenshots/new_noisereduction_plugin.png

The only one NR tool that i have already used witch provide a similar result
is Noise Nija plugin for photoshop... If you want to test it, there is a diff
file against current NR tool implementation from svn trunk here :

http://digikam3rdparty.free.fr/misc.tarballs/nr.diff

To test it, just go to digikamimageplugins/noisereduction directory, and apply
it like this:

# diff -p0 < nr.diff

make, and install the plugin. You can play with it in digikam. There are some
noisy images to use here :

http://digikam3rdparty.free.fr/Images_2_Test_DImg/NOISE/

To use it, especially with parameters, I recommend you to read the comments of
the algorithm author from the original gimp plugin source code here :

http://digikam3rdparty.free.fr/misc.tarballs/dcamnoise2-0.63.c

Nota :

- The most important settings are in General tab.
- The plugin do not support yet 16 bits images. It's just a question of time.
I will do it later, if the fedback is right from digiKam users.

Please, let's me hear your viewpoints. Thanks in advance...

Gilles Caulier
_______________________________________________
Digikam-devel mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-devel

[Gilles Caulier-2]

Le Mardi 14 Mars 2006 12:03 AM, Caulier Gilles a écrit :

> Hi all,
>
> I have recently backported the new despeckle algorithm from gimp 2.2.x to
> digiKam like noise reduction tool with support of 16 bits images.
>
> After to have played with it, i have been unsatisfied... The result are a
> little bit strange, especially with recursive option.
>
> After to have serching on the web the best NR technics, my conclusion is
> that this algorithm is a wrong way. despeckle use RGB color space to fix
> noise and this is a wrong issue.
>
> After to any investigation, i have found a great GPL plugin tool named
> dcamnoise2. The code is really impressive and very well documented...
>
> I have backported the code this week end, and the result look great :
>
> http://digikam3rdparty.free.fr/Screenshots/new_noisereduction_plugin.png
>
> The only one NR tool that i have already used witch provide a similar
> result is Noise Nija plugin for photoshop... If you want to test it, there
> is a diff file against current NR tool implementation from svn trunk here :
>
> http://digikam3rdparty.free.fr/misc.tarballs/nr.diff
>
> To test it, just go to digikamimageplugins/noisereduction directory, and
> apply it like this:
>
> # diff -p0 < nr.diff
>
> make, and install the plugin. You can play with it in digikam. There are
> some noisy images to use here :
>
> http://digikam3rdparty.free.fr/Images_2_Test_DImg/NOISE/
>
> To use it, especially with parameters, I recommend you to read the comments
> of the algorithm author from the original gimp plugin source code here :
>
> http://digikam3rdparty.free.fr/misc.tarballs/dcamnoise2-0.63.c
>
> Nota :
>
> - The most important settings are in General tab.
> - The plugin do not support yet 16 bits images. It's just a question of
> time. I will do it later, if the fedback is right from digiKam users.
>
> Please, let's me hear your viewpoints. Thanks in advance...

Forget to said that i need some RAW file samples taken with important noise to
test 16 bits image support. You can capture these pictures to boost ISO
settings of your camera for example.

Thanks in advance

Gilles Caulier
_______________________________________________
Digikam-devel mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-devel

Hi,

This looks great ! thanks Gilles

>>To test it, just go to digikamimageplugins/noisereduction directory, and
>>apply it like this:
>>
>># diff -p0 < nr.diff
>>

For those who are not used to diff/patch, in fact you need to apply it
like this:

patch -p0 < nr.diff

Julien
_______________________________________________
Digikam-devel mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-devel

[Gilles Caulier-2]

On Tuesday 14 March 2006 08:57 am, Julien Narboux wrote:

> Hi,
>
> This looks great ! thanks Gilles
>
> >>To test it, just go to digikamimageplugins/noisereduction directory, and
> >>apply it like this:
> >>
> >># diff -p0 < nr.diff
>
> For those who are not used to diff/patch, in fact you need to apply it
> like this:
>
> patch -p0 < nr.diff
>

Oups, sorry (:=))) I need a big coffee... I have done this mail yesterday late
in evening. I have lost some slepping hours (:=)))...

Gilles
_______________________________________________
Digikam-devel mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-devel

[Gilles Caulier-2]

A new version of the diff file is availalble at the same place:

http://digikam3rdparty.free.fr/misc.tarballs/nr.diff

 I have fixed some settings limit to prevent crash, especially with advanced
parameters, and fixed some uninitialized variable in memory (thanks Mr
valgrind)

Gilles


On Tuesday 14 March 2006 12:03 am, Caulier Gilles wrote:

> Hi all,
>
> I have recently backported the new despeckle algorithm from gimp 2.2.x to
> digiKam like noise reduction tool with support of 16 bits images.
>
> After to have played with it, i have been unsatisfied... The result are a
> little bit strange, especially with recursive option.
>
> After to have serching on the web the best NR technics, my conclusion is
> that this algorithm is a wrong way. despeckle use RGB color space to fix
> noise and this is a wrong issue.
>
> After to any investigation, i have found a great GPL plugin tool named
> dcamnoise2. The code is really impressive and very well documented...
>
> I have backported the code this week end, and the result look great :
>
> http://digikam3rdparty.free.fr/Screenshots/new_noisereduction_plugin.png
>
> The only one NR tool that i have already used witch provide a similar
> result is Noise Nija plugin for photoshop... If you want to test it, there
> is a diff file against current NR tool implementation from svn trunk here :
>
> http://digikam3rdparty.free.fr/misc.tarballs/nr.diff
>
> To test it, just go to digikamimageplugins/noisereduction directory, and
> apply it like this:
>
> # diff -p0 < nr.diff
>
> make, and install the plugin. You can play with it in digikam. There are
> some noisy images to use here :
>
> http://digikam3rdparty.free.fr/Images_2_Test_DImg/NOISE/
>
> To use it, especially with parameters, I recommend you to read the comments
> of the algorithm author from the original gimp plugin source code here :
>
> http://digikam3rdparty.free.fr/misc.tarballs/dcamnoise2-0.63.c
>
> Nota :
>
> - The most important settings are in General tab.
> - The plugin do not support yet 16 bits images. It's just a question of
> time. I will do it later, if the fedback is right from digiKam users.
>
> Please, let's me hear your viewpoints. Thanks in advance...
>
> Gilles Caulier
> _______________________________________________
> Digikam-users mailing list
> [hidden email]
> https://mail.kde.org/mailman/listinfo/digikam-users

_______________________________________________
Digikam-devel mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-devel

Dnia wtorek, 14 marca 2006 00:03, Caulier Gilles napisał:
>
> I have backported the code this week end, and the result look great :
>
> http://digikam3rdparty.free.fr/Screenshots/new_noisereduction_plugin.png

Looks great! Old algorithm was rather distributing noise more evenly
blurring image. Now, this one actually *removes* noise :)

m.

_______________________________________________
Digikam-devel mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-devel

[Gilles Caulier-2]

The code is in svn now. The plugin support 16 bits images !!! I have polished
the implementation and tested indeep to prevent crash and memory leak.

I have found a big problem in the original implementation from "dcamnoise2"
tool : the blured image created before to applying the NR filter is broken.
This is very difficult to see it without zooming in the target image,
especially in 8bits. To 16 bits this weird effect is very increased ! But
this is not a problem with this new tool now (:=)))...

The settings area need to be polished, and all parameters need a full
contextual descriptions.

Please let's me hear your suggestions about.

I would to thanks all digiKam users who have provides some RAW pictures with
noise to debug this plugin.

Gilles Caulier

On Tuesday 14 March 2006 12:03 am, Caulier Gilles wrote:

> Hi all,
>
> I have recently backported the new despeckle algorithm from gimp 2.2.x to
> digiKam like noise reduction tool with support of 16 bits images.
>
> After to have played with it, i have been unsatisfied... The result are a
> little bit strange, especially with recursive option.
>
> After to have serching on the web the best NR technics, my conclusion is
> that this algorithm is a wrong way. despeckle use RGB color space to fix
> noise and this is a wrong issue.
>
> After to any investigation, i have found a great GPL plugin tool named
> dcamnoise2. The code is really impressive and very well documented...
>
> I have backported the code this week end, and the result look great :
>
> http://digikam3rdparty.free.fr/Screenshots/new_noisereduction_plugin.png
>
> The only one NR tool that i have already used witch provide a similar
> result is Noise Nija plugin for photoshop... If you want to test it, there
> is a diff file against current NR tool implementation from svn trunk here :
>
> http://digikam3rdparty.free.fr/misc.tarballs/nr.diff
>
> To test it, just go to digikamimageplugins/noisereduction directory, and
> apply it like this:
>
> # diff -p0 < nr.diff
>
> make, and install the plugin. You can play with it in digikam. There are
> some noisy images to use here :
>
> http://digikam3rdparty.free.fr/Images_2_Test_DImg/NOISE/
>
> To use it, especially with parameters, I recommend you to read the comments
> of the algorithm author from the original gimp plugin source code here :
>
> http://digikam3rdparty.free.fr/misc.tarballs/dcamnoise2-0.63.c
>
> Nota :
>
> - The most important settings are in General tab.
> - The plugin do not support yet 16 bits images. It's just a question of
> time. I will do it later, if the fedback is right from digiKam users.
>
> Please, let's me hear your viewpoints. Thanks in advance...
>
> Gilles Caulier
> _______________________________________________
> Digikam-users mailing list
> [hidden email]
> https://mail.kde.org/mailman/listinfo/digikam-users

_______________________________________________
Digikam-devel mailing list
[hidden email]
https://mail.kde.org/mailman/listinfo/digikam-devel