Sent: 2010-04-04 04:24:41 UTC (almost 2 years ago)

From: Danny Robson

gsoc2010: a collection of filters for GEGL

permalink

Hi all,

I am a PhD candidate at the Australian National University, currently
creating and evaluating performance profiling tools for NUMA systems.
I have always had an interest in the realm of graphics, though I have
only managed minor patches for GEGL/BABL in the past.

As the deadline for gsoc2010 applications is far too rapidly
approaching, I was curious about thoughts on another alternative
proposal: implementing a collection or series of related filters.

I have a selection of ideas which I have briefly investigated, and may
be of interest:

1. While GEGL has excellent support for high depth colour spaces, there
are few examples of filters which take advantage of this explicitly. I
would be interested in implementing (or porting) a collection of tone
mapping operators as native GEGL filters.

2. Image matting techniques with trimaps, such as the closed form of
Levin et. al., have very interesting applications for end users. I
would be interested in implementing one such technique and further
filters applying it (such as haze removal and spatially varying white
balance).

3. The matting of #2 could segue nicely into an implementation of image
completion/inpainting (which is receiving more public exposure due to
similar techniques in the upcoming Adobe CS5).

I'm not quite clear on the level of complexity required for a gsoc
proposal so I have left the extent and combination of ideas open.
I would be open to further suggestions for this style of proposal, and
shall use any input for a more concrete proposal if there is interest.

I'll try to catch some people on IRC (under the nick 'eNGIMa') to
discuss the idea's desirability.

Many thanks for your consideration,
- Danny Robson

_______________________________________________
Gegl-developer mailing list
Gegl-developer@lists.XCF.Berkeley.EDU
https://lists.XCF.Berkeley.EDU/mailman/listinfo/gegl-developer

Sent: 2010-04-04 05:16:49 UTC (almost 2 years ago)

From: David Gowers

gsoc2010: a collection of filters for GEGL

permalink

On Sun, Apr 4, 2010 at 11:54 AM, Danny Robson wrote:
> Hi all,
>
> I am a PhD candidate at the Australian National University, currently
> creating and evaluating performance profiling tools for NUMA systems.
> I have always had an interest in the realm of graphics, though I have
> only managed minor patches for GEGL/BABL in the past.
>
> As the deadline for gsoc2010 applications is far too rapidly
> approaching, I was curious about thoughts on another alternative
> proposal: implementing a collection or series of related filters.
>
> I have a selection of ideas which I have briefly investigated, and may
> be of interest:
>
> 1. While GEGL has excellent support for high depth colour spaces, there
> are few examples of filters which take advantage of this explicitly. I
> would be interested in implementing (or porting) a collection of tone
> mapping operators as native GEGL filters.
This leads me to wonder what kind of operation the GEGL 'tone-map' op
currently performs.

>
> 2. Image matting techniques with trimaps, such as the closed form of
> Levin et. al., have very interesting applications for end users. I
> would be interested in implementing one such technique and further
> filters applying it (such as haze removal and spatially varying white
> balance).
>
> 3. The matting of #2 could segue nicely into an implementation of image
> completion/inpainting (which is receiving more public exposure due to
> similar techniques in the upcoming Adobe CS5).

GMIC and Resynthesizer already do this.
Perhaps they could do it faster or better.

>
> I'm not quite clear on the level of complexity required for a gsoc
> proposal so I have left the extent and combination of ideas open.
> I would be open to further suggestions for this style of proposal, and
> shall use any input for a more concrete proposal if there is interest.
>
> I'll try to catch some people on IRC (under the nick 'eNGIMa') to
> discuss the idea's desirability.
>
> Many thanks for your consideration,
> - Danny Robson

_______________________________________________
Gegl-developer mailing list
Gegl-developer@lists.XCF.Berkeley.EDU
https://lists.XCF.Berkeley.EDU/mailman/listinfo/gegl-developer

Sent: 2010-04-04 05:58:39 UTC (almost 2 years ago)

From: Danny Robson

gsoc2010: a collection of filters for GEGL

permalink

On Sun, 4 Apr 2010 12:46:49 +0930
David Gowers wrote:

> > 1. While GEGL has excellent support for high depth colour spaces,
> > there are few examples of filters which take advantage of this
> > explicitly. I would be interested in implementing (or porting) a
> > collection of tone mapping operators as native GEGL filters.
> This leads me to wonder what kind of operation the GEGL 'tone-map' op
> currently performs.

Different operators produce different effects. It is often desirable to
select one which imparts a specific look to the end image. Having a
selection would be beneficial.

> >
> > 2. Image matting techniques with trimaps, such as the closed form of
> > Levin et. al., have very interesting applications for end users. I
> > would be interested in implementing one such technique and further
> > filters applying it (such as haze removal and spatially varying
> > white balance).
> >
> > 3. The matting of #2 could segue nicely into an implementation of
> > image completion/inpainting (which is receiving more public
> > exposure due to similar techniques in the upcoming Adobe CS5).
>
> GMIC and Resynthesizer already do this.
> Perhaps they could do it faster or better.

Most likely they could do the completion/inpainting operations more
efficiently. However my interest is more in the matting operators, and
subsequently application of them.

I was mostly interested in inpainting as a target application of
matting. However, it's excellent note that it is less attractive to
target this action given resynthesizer and GMIC.

- Danny

_______________________________________________
Gegl-developer mailing list
Gegl-developer@lists.XCF.Berkeley.EDU
https://lists.XCF.Berkeley.EDU/mailman/listinfo/gegl-developer

Sent: 2010-04-04 06:33:52 UTC (almost 2 years ago)

From: Øyvind Kolås

gsoc2010: a collection of filters for GEGL

permalink

On Sun, Apr 4, 2010 at 4:16 AM, David Gowers wrote:
>> 1. While GEGL has excellent support for high depth colour spaces, there
>> are few examples of filters which take advantage of this explicitly. I
>> would be interested in implementing (or porting) a collection of tone
>> mapping operators as native GEGL filters.
> This leads me to wonder what kind of operation the GEGL 'tone-map' op
> currently performs.

That op has been removed, I have some experimental (not commited to
git yet) code that revisits the ideas that originally were employed in
that op, another place where some of the concepts of this op is
manifested is in the gegl:stress op (which is stochastic and noisy for
low iterations/reasonable processing times.)

>> 3. The matting of #2 could segue nicely into an implementation of image
>> completion/inpainting (which is receiving more public exposure due to
>> similar techniques in the upcoming Adobe CS5).
>
> GMIC and Resynthesizer already do this.
> Perhaps they could do it faster or better.

GMIC and resynthesizer do some of these things with some approaches,
netiher of them do it as GEGL ops nor are they usable for GIMP or
other applications as GEGL operations. Thus from a purely GEGL based
view no such functionality is available.

/Øyvind K.

--
«The future is already here. It's just not very evenly distributed»
-- William Gibson
http://pippin.gimp.org/ http://ffii.org/
_______________________________________________
Gegl-developer mailing list
Gegl-developer@lists.XCF.Berkeley.EDU
https://lists.XCF.Berkeley.EDU/mailman/listinfo/gegl-developer

Sent: 2010-04-04 08:47:08 UTC (almost 2 years ago)

From: David Gowers

gsoc2010: a collection of filters for GEGL

permalink

On Sun, Apr 4, 2010 at 2:03 PM, Øyvind Kolås wrote:
> GMIC and resynthesizer do some of these things with some approaches,
> netiher of them do it as GEGL ops nor are they usable for GIMP or
> other applications as GEGL operations. Thus from a purely GEGL based
> view no such functionality is available.
>
> /Øyvind K.
Sorry, I misread the message headers and thought this was the GIMP list.

_______________________________________________
Gegl-developer mailing list
Gegl-developer@lists.XCF.Berkeley.EDU
https://lists.XCF.Berkeley.EDU/mailman/listinfo/gegl-developer