RSS/Atom feed Twitter
Site is read-only, email is disabled

comparison of enlargements produced with two GIMP methods, three GEGL methods, and three candidate GEGL methods

This discussion is connected to the gimp-developer-list.gnome.org mailing list which is provided by the GIMP developers and not related to gimpusers.com.

This is a read-only list on gimpusers.com so this discussion thread is read-only, too.

Nicolas Robidoux
2010-06-03 05:49:52 UTC (over 14 years ago)

comparison of enlargements produced with two GIMP methods, three GEGL methods, and three candidate GEGL methods

I have created 6.11x enlargements with two GIMP methods, three GEGL methods, and three candidate GEGL methods for which patches are provided in

https://bugzilla.gnome.org/show_bug.cgi?id=619314

The enlargements, named a.png to h.png, are found in the 15M tar archive

http://web.cs.laurentian.ca/nrobidoux/misc/catenlargementtests.tgz

To know which method was used to produce each image, read the KEY text file.

One of the three candidate methods is very fast. The total runtime for the creation of the enlargement with upsmooth is only 15% more than with bilinear.
Only one of the three candidate methods is slower than GEGL bicubic (upsharp, by 30%).

Nicolas Robidoux

PS

Warning: the upsize, upsmooth and upsharp samplers already in GEGL are prototypes of the methods showcased in the above enlargement test.

Nicolas Robidoux
2010-06-03 18:33:36 UTC (over 14 years ago)

comparison of enlargements produced with two GIMP methods, three GEGL methods, and three candidate GEGL methods

To illustrate the effect of the samplers on images from the opposite end of the spectrum (text as opposed to natural), I have put together a similar comparison with a text image created with GIMP. It shows 9.73x enlargements with two GIMP methods, three GEGL methods, and three candidate GEGL methods for which patches are provided in

https://bugzilla.gnome.org/show_bug.cgi?id=619314

The enlargements, named A.png to H.png, are found in the tar archive

http://web.cs.laurentian.ca/nrobidoux/misc/textenlargementtests.tgz

To know which method was used to produce each image (and get runtimes for the GEGL methods), read the KEY text file.

Nicolas Robidoux

Sven Neumann
2010-06-03 22:07:58 UTC (over 14 years ago)

comparison of enlargements produced with two GIMP methods, three GEGL methods, and three candidate GEGL methods

On Wed, 2010-06-02 at 23:49 -0400, Nicolas Robidoux wrote:

I have created 6.11x enlargements with two GIMP methods, three GEGL methods, and three candidate GEGL methods for which patches are provided in

https://bugzilla.gnome.org/show_bug.cgi?id=619314

The enlargements, named a.png to h.png, are found in the 15M tar archive

http://web.cs.laurentian.ca/nrobidoux/misc/catenlargementtests.tgz

To know which method was used to produce each image, read the KEY text file.

One thing I noticed when looking at these images is that the GEGL ops all create a black border at the bottom-right edges. Is that something you are aware of and expect to get "fixed" ?

Sven

Nicolas Robidoux
2010-06-04 20:26:15 UTC (over 14 years ago)

comparison of enlargements produced with two GIMP methods, three GEGL methods, and three candidate GEGL methods

To illustrate the effect of the samplers when ROTATING images, I put together a comparative test. It shows the results of rotating a quite sharp image of astronauts near the space station---if I remember correctly---by 6.5 degrees with four GIMP methods, three GEGL methods and three candidate GEGL methods for which patches are provided in

https://bugzilla.gnome.org/show_bug.cgi?id=619314

(The image is very sharp: I actually wonder if NASA sharpened it.)

The results, named one.png to nin.png, are found in the tar archive

http://web.cs.laurentian.ca/nrobidoux/misc/astronautsrotatetests.tgz

To know which method was used to produce each image (and get runtimes for the GEGL methods), read the KEY text file.

Nicolas Robidoux

Nicolas Robidoux
2010-06-05 16:12:38 UTC (over 14 years ago)

comparison of enlargements produced with two GIMP methods, three GEGL methods, and three candidate GEGL methods

In light of some feedback received, me and my students will be making major changes to three of the gegl methods, so you may want to wait until the comparisons are updated.

Nicolas Robidoux

On Fri, Jun 4, 2010 at 2:26 PM, Nicolas Robidoux wrote:

To illustrate the effect of the samplers when ROTATING images, I put together a comparative test. It shows the results of rotating a quite sharp image of astronauts near the space station---if I remember correctly---by 6.5 degrees with four GIMP methods, three GEGL methods and three candidate GEGL methods for which patches are provided in

https://bugzilla.gnome.org/show_bug.cgi?id=619314

(The image is very sharp: I actually wonder if NASA sharpened it.)

The results, named one.png to nin.png, are found in the tar archive

http://web.cs.laurentian.ca/nrobidoux/misc/astronautsrotatetests.tgz

To know which method was used to produce each image (and get runtimes for the GEGL methods), read the KEY text file.

Nicolas Robidoux