Hi,
as far as I see this, we are using a normal gaussian blur in the blur tool. The problem with our own algorithm: it is damn slow. I just played around with the CImg::blur() method and it is like 20 times faster, with an (at least for me) equal result. I think we should consider using CImg (or OpenCV) in more plugins instead of using or own algorithms, if it is not needed. These libraries have been tested a lot and are tuned for performance. What do you think? I can try to patch the blur tool to see if it really makes differences. BTW, I try to implement an "Shadow / Highlight" imageplugin at the moment. This will also use CImg. If we are going to use CImg in more plugins, we should add some classes to DImg to easily convert between the two storage formats. What do you think? Andi _______________________________________________ Digikam-devel mailing list [hidden email] https://mail.kde.org/mailman/listinfo/digikam-devel |
If CImg works fine here, the blur filter doesn't need to be a ThreadedFilter anymore... Right? Andi On Monday 13 April 2009 15:47:59 Andi Clemens wrote: > Hi, > > as far as I see this, we are using a normal gaussian blur in the blur tool. > The problem with our own algorithm: it is damn slow. > I just played around with the CImg::blur() method and it is like 20 times > faster, with an (at least for me) equal result. > I think we should consider using CImg (or OpenCV) in more plugins instead > of using or own algorithms, if it is not needed. > These libraries have been tested a lot and are tuned for performance. > > What do you think? > I can try to patch the blur tool to see if it really makes differences. > > BTW, I try to implement an "Shadow / Highlight" imageplugin at the moment. > This will also use CImg. > > If we are going to use CImg in more plugins, we should add some classes to > DImg to easily convert between the two storage formats. > > What do you think? > > Andi > > _______________________________________________ > Digikam-devel mailing list > [hidden email] > https://mail.kde.org/mailman/listinfo/digikam-devel _______________________________________________ Digikam-devel mailing list [hidden email] https://mail.kde.org/mailman/listinfo/digikam-devel |
I just inserted the blur() method into my "not working" shadow / highlight
plugin. Now you can even use blur in realtime, without the "Try" button. We definitely should switch to CImg here. Andi On Monday 13 April 2009 15:50:07 Andi Clemens wrote: > If CImg works fine here, the blur filter doesn't need to be a > ThreadedFilter anymore... Right? > > Andi > > On Monday 13 April 2009 15:47:59 Andi Clemens wrote: > > Hi, > > > > as far as I see this, we are using a normal gaussian blur in the blur > > tool. The problem with our own algorithm: it is damn slow. > > I just played around with the CImg::blur() method and it is like 20 times > > faster, with an (at least for me) equal result. > > I think we should consider using CImg (or OpenCV) in more plugins instead > > of using or own algorithms, if it is not needed. > > These libraries have been tested a lot and are tuned for performance. > > > > What do you think? > > I can try to patch the blur tool to see if it really makes differences. > > > > BTW, I try to implement an "Shadow / Highlight" imageplugin at the > > moment. This will also use CImg. > > > > If we are going to use CImg in more plugins, we should add some classes > > to DImg to easily convert between the two storage formats. > > > > What do you think? > > > > Andi > > > > _______________________________________________ > > Digikam-devel mailing list > > [hidden email] > > https://mail.kde.org/mailman/listinfo/digikam-devel > > _______________________________________________ > Digikam-devel mailing list > [hidden email] > https://mail.kde.org/mailman/listinfo/digikam-devel _______________________________________________ Digikam-devel mailing list [hidden email] https://mail.kde.org/mailman/listinfo/digikam-devel |
In reply to this post by Bugzilla from andi.clemens@gmx.net
2009/4/13 Andi Clemens <[hidden email]>:
> > If CImg works fine here, the blur filter doesn't need to be a ThreadedFilter > anymore... Right? No. all filters need to use an outside thread to run. and it's yet the case, for example with levels or curves tools. > > Andi > > On Monday 13 April 2009 15:47:59 Andi Clemens wrote: >> Hi, >> >> as far as I see this, we are using a normal gaussian blur in the blur tool. >> The problem with our own algorithm: it is damn slow. Yes, i know the problem... Algorithm is not mine, it come from a Bresilien mathematician. (:=))) >> I just played around with the CImg::blur() method and it is like 20 times >> faster, with an (at least for me) equal result. It's a Gaussian blur like ? >> I think we should consider using CImg (or OpenCV) in more plugins instead >> of using or own algorithms, if it is not needed. >> These libraries have been tested a lot and are tuned for performance. CImg is already in digiKam core, No problem to use it. >> >> What do you think? >> I can try to patch the blur tool to see if it really makes differences. >> yes, of course. >> BTW, I try to implement an "Shadow / Highlight" imageplugin at the moment. >> This will also use CImg. >> ok >> If we are going to use CImg in more plugins, we should add some classes to >> DImg to easily convert between the two storage formats. yes. Code to convert DImg <-> CImg is already in greystorationiface.cpp >> >> What do you think? Fine for me Gilles >> >> Andi >> >> _______________________________________________ >> Digikam-devel mailing list >> [hidden email] >> https://mail.kde.org/mailman/listinfo/digikam-devel > > _______________________________________________ > Digikam-devel mailing list > [hidden email] > https://mail.kde.org/mailman/listinfo/digikam-devel > Digikam-devel mailing list [hidden email] https://mail.kde.org/mailman/listinfo/digikam-devel |
On Monday 13 April 2009 15:56:51 Gilles Caulier wrote: > 2009/4/13 Andi Clemens <[hidden email]>: > > If CImg works fine here, the blur filter doesn't need to be a > > ThreadedFilter anymore... Right? > > No. all filters need to use an outside thread to run. and it's yet the > case, for example with levels or curves tools. > > > Andi > > > > On Monday 13 April 2009 15:47:59 Andi Clemens wrote: > >> Hi, > >> > >> as far as I see this, we are using a normal gaussian blur in the blur > >> tool. The problem with our own algorithm: it is damn slow. > > Yes, i know the problem... Algorithm is not mine, it come from a > Bresilien mathematician. (:=))) > > >> I just played around with the CImg::blur() method and it is like 20 > >> times faster, with an (at least for me) equal result. > > It's a Gaussian blur like ? Yes > >> If we are going to use CImg in more plugins, we should add some classes > >> to DImg to easily convert between the two storage formats. > > yes. Code to convert DImg <-> CImg is already in greystorationiface.cpp I meant: we should add some methods to the DImg class, not classes to DImg :-) > > >> What do you think? > > Fine for me > > Gilles > > >> Andi > >> > >> _______________________________________________ > >> Digikam-devel mailing list > >> [hidden email] > >> https://mail.kde.org/mailman/listinfo/digikam-devel > > > > _______________________________________________ > > Digikam-devel mailing list > > [hidden email] > > https://mail.kde.org/mailman/listinfo/digikam-devel > > _______________________________________________ > Digikam-devel mailing list > [hidden email] > https://mail.kde.org/mailman/listinfo/digikam-devel _______________________________________________ Digikam-devel mailing list [hidden email] https://mail.kde.org/mailman/listinfo/digikam-devel |
2009/4/13 Andi Clemens <[hidden email]>:
> > On Monday 13 April 2009 15:56:51 Gilles Caulier wrote: >> 2009/4/13 Andi Clemens <[hidden email]>: >> > If CImg works fine here, the blur filter doesn't need to be a >> > ThreadedFilter anymore... Right? >> >> No. all filters need to use an outside thread to run. and it's yet the >> case, for example with levels or curves tools. >> >> > Andi >> > >> > On Monday 13 April 2009 15:47:59 Andi Clemens wrote: >> >> Hi, >> >> >> >> as far as I see this, we are using a normal gaussian blur in the blur >> >> tool. The problem with our own algorithm: it is damn slow. >> >> Yes, i know the problem... Algorithm is not mine, it come from a >> Bresilien mathematician. (:=))) >> >> >> I just played around with the CImg::blur() method and it is like 20 >> >> times faster, with an (at least for me) equal result. >> >> It's a Gaussian blur like ? > > Yes > >> >> If we are going to use CImg in more plugins, we should add some classes >> >> to DImg to easily convert between the two storage formats. >> >> yes. Code to convert DImg <-> CImg is already in greystorationiface.cpp > > I meant: we should add some methods to the DImg class, not classes to DImg :-) yes. i have understood. Just take code from greystorationiface.cpp, extend DImg with it and factorize... Gilles _______________________________________________ Digikam-devel mailing list [hidden email] https://mail.kde.org/mailman/listinfo/digikam-devel |
Free forum by Nabble | Edit this page |