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

Image restoration plugin

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.

4 of 4 messages available
Toggle history

Please log in to manage your subscriptions.

Image restoration plugin Thierry Pierron 14 Sep 22:10
  Image restoration plugin GSR - FR 14 Sep 22:48
  Image restoration plugin Tor Lillqvist 15 Sep 03:13
   Image restoration plugin Thierry Pierron 15 Sep 18:31
Thierry Pierron
2004-09-14 22:10:13 UTC (over 19 years ago)

Image restoration plugin

Hello all of you,

I'm new to this list, and I wish to know if somebody here is already involved in
integrating some aspect of the CImg library inside a gimp plugin.

That library has a very efficient algorithm for restoring lossy-compression image format. But the most impressive part in the image in-painting. You can restore picture by removing some part of the image (usually text, logos over the image or whole objets inside the picture, like glasses on a face).

Interrest of that feature is obvious, and I wonder if anyone isn't already involved in a such project ?

If no, I'm extremely motivated to write that plugin (or help people that are writing one). But, well, it could be fine if we can avoid redundant project.

C.f http://cimg.sourceforge.net/ or http://www-sop.inria.fr/odyssee/research/tschumperle-deriche:02d/appliu/

-- Thierry Pierron

GSR - FR
2004-09-14 22:48:45 UTC (over 19 years ago)

Image restoration plugin

tpierron@b2b2c.ca (2004-09-14 at 1610.13 -0400):

If no, I'm extremely motivated to write that plugin (or help people that are writing one). But, well, it could be fine if we can avoid redundant project.

None I know, I raised the topic some time ago.

Another suggestion about this lib: fake motion blur. That is one of the reason I checked it, wanted a quick way to get motion blur given a sequence of images (for example, a 3d rendering). One of that lib's capabilities is interpolating between two similar images. With that new group of images and via composition, you can get acceptable blurs in really few time (compared to 3d apps that do it by rendering again and again, I mean). I did a quick test, but only works with gray images (so RGB would mean separation, three passes and joining, if the lib has no method for RGB, the demo app did not).

From my POV, some kind of wrapper to write multiple plugins would be

indeed interesting (I can think of some tasks that could use it: noise cleaner, hole filler, motion blurer, texture maker, zoomer/scaler...).

Thanks in advance for any effort you do about cimg and gimp.

GSR

Tor Lillqvist
2004-09-15 03:13:01 UTC (over 19 years ago)

Image restoration plugin

Thierry Pierron writes:
> That library has a very efficient algorithm for restoring > lossy-compression image format.

Hmm, do you maybe mean effective and not efficient?

The application of this that I would be most interested in is denoising. I tried the pde_TschumperleDeriche2D example program on the parrot image with some Gaussian noise added, and it is rather slow. It would probably take way too long to filter out noise from a realistical size (megapixels) digital camera image?

Also, the result has an eerie "artistic" or "oilpaintingish" look to it. But I am sure all denoising algorithms always have some artefacts.

But anyway, sure, it is very interesting. It wouldn't be hard at all to make a GIMP plug-in using the CImg library.

For those too lazy to try it our themselves, below are (links to) the original parrot image, with added noise, and after 170 iterations (whoch took 25 minutes on my aging 450 MHz P3, scale appropriately for modern machines).

http://www.iki.fi/tml/parrot_original.png http://www.iki.fi/tml/parrot_output_000000.png http://www.iki.fi/tml/parrot_output_000170.png

--tml

Thierry Pierron
2004-09-15 18:31:03 UTC (over 19 years ago)

Image restoration plugin

Selon Tor Lillqvist :

Hmm, do you maybe mean effective and not efficient?

Also, the result has an eerie "artistic" or "oilpaintingish" look to it. But I am sure all denoising algorithms always have some artefacts.

Yes, it's true that those algorithms are quite slow, even on fast machine. Moreover, denoising part often requires some tests before having acceptable results (which is somewhat subjective). Fine-tuning parameters on slow machines is really painful (some may say unusable).

I'm for instance focussed on inpainting removal, since it gives rather good results, and it's simple to use.

Regards, --
Thierry Pierron