Re: [Gimp-user] Fwd: Image resizing based upon image content
Konstantin Svist wrote: Doug wrote: snip Having tried out Liquid Rescale I'm very impressed. But does anyone have an idea about the size of image it can deal with in terms of memory requirements (I presume the Gimp setting Tile Cache size)? I've crashed it with a 700MB jpg setting Tile Cache Size to 3 GB (using Linux Mandriva 2007.0) There should be best/worst-case estimates in the paper, but if not, it probably depends on the algorithm. But 700MB jpeg... that's pretty hard core :) Thanks, I'll go have a look at the paper. The jpeg was a mosaic of some 25 sensible-sized pics ;-) Doug ___ Gimp-user mailing list Gimp-user@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-user
Re: [Gimp-user] Fwd: Image resizing based upon image content
Doug wrote: Konstantin Svist wrote: Doug wrote: snip Having tried out Liquid Rescale I'm very impressed. But does anyone have an idea about the size of image it can deal with in terms of memory requirements (I presume the Gimp setting Tile Cache size)? I've crashed it with a 700MB jpg setting Tile Cache Size to 3 GB (using Linux Mandriva 2007.0) There should be best/worst-case estimates in the paper, but if not, it probably depends on the algorithm. But 700MB jpeg... that's pretty hard core :) Thanks, I'll go have a look at the paper. The jpeg was a mosaic of some 25 sensible-sized pics ;-) Doug P.S. the paper is beyond me, but I couldn't recognise any sort of estimates. Thanks anyway. Doug ___ Gimp-user mailing list Gimp-user@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-user
Re: [Gimp-user] Fwd: Image resizing based upon image content
Hi, On Sun, 2007-10-07 at 16:28 -0700, Konstantin Svist wrote: Having tried out Liquid Rescale I'm very impressed. But does anyone have an idea about the size of image it can deal with in terms of memory requirements (I presume the Gimp setting Tile Cache size)? I've crashed it with a 700MB jpg setting Tile Cache Size to 3 GB (using Linux Mandriva 2007.0) There should be best/worst-case estimates in the paper, but if not, it probably depends on the algorithm. It also very much depends on how the plug-in is implemented. If it transfers the full image data from GIMP to the plug-in, then it's basically unusable with large images. Unfortunately quite a lot of plug-ins are implemented that way. Sven ___ Gimp-user mailing list Gimp-user@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-user
Re: [Gimp-user] Fwd: Image resizing based upon image content
Konstantin Svist wrote: David Gowers wrote: oops, this didn't get sent to the list. Also, I think the problems I'm encountering are because: A) YIQ tends to oversaturate too readily B) the image ICC profile is AFAIK undefined, and is, anyway, disregarded. C) L*a*b* colorspace is not being used. In my experience L*a*b* represents saturation accurately, as well as intensity If this plugin could use L*a*b* space (with GEGL's help?) for the application of colorization effect, it would be more useful. -- Forwarded message -- From: David Gowers [EMAIL PROTECTED] Date: Oct 6, 2007 4:39 PM Subject: Re: [Gimp-user] Image resizing based upon image content To: [EMAIL PROTECTED] [EMAIL PROTECTED] On 10/6/07, [EMAIL PROTECTED] [EMAIL PROTECTED] wrote: Konstantin Svist [EMAIL PROTECTED] said on Oct 04, 2007 15:54 -0400 (in part): Kevin Cozens wrote: Dotan Cohen wrote: There is an amazing video describing an image resizing algorithm that removes (or adds) pixels based upon their importance to the subjecto of the photo. Can this be implemented in The Gimp? Of course it can be implemented in GIMP. It just takes someone with the time and ability to write the code for a plug-in that implements the algorithm. I don't think the one you point to below is the same as Liquid Resizer (see http://schumaml.gmxhome.de/downloads/gimp/ and d/l http://schumaml.gmxhome.de/downloads/gimp/gimp-lqr-plugin-0.1.3.zip for it). snip Having tried out Liquid Rescale I'm very impressed. But does anyone have an idea about the size of image it can deal with in terms of memory requirements (I presume the Gimp setting Tile Cache size)? I've crashed it with a 700MB jpg setting Tile Cache Size to 3 GB (using Linux Mandriva 2007.0). Doug ___ Gimp-user mailing list Gimp-user@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-user
Re: [Gimp-user] Fwd: Image resizing based upon image content
Doug wrote: snip Having tried out Liquid Rescale I'm very impressed. But does anyone have an idea about the size of image it can deal with in terms of memory requirements (I presume the Gimp setting Tile Cache size)? I've crashed it with a 700MB jpg setting Tile Cache Size to 3 GB (using Linux Mandriva 2007.0) There should be best/worst-case estimates in the paper, but if not, it probably depends on the algorithm. But 700MB jpeg... that's pretty hard core :) ___ Gimp-user mailing list Gimp-user@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-user
Re: [Gimp-user] Fwd: Image resizing based upon image content
David Gowers wrote: oops, this didn't get sent to the list. Also, I think the problems I'm encountering are because: A) YIQ tends to oversaturate too readily B) the image ICC profile is AFAIK undefined, and is, anyway, disregarded. C) L*a*b* colorspace is not being used. In my experience L*a*b* represents saturation accurately, as well as intensity If this plugin could use L*a*b* space (with GEGL's help?) for the application of colorization effect, it would be more useful. -- Forwarded message -- From: David Gowers [EMAIL PROTECTED] Date: Oct 6, 2007 4:39 PM Subject: Re: [Gimp-user] Image resizing based upon image content To: [EMAIL PROTECTED] [EMAIL PROTECTED] On 10/6/07, [EMAIL PROTECTED] [EMAIL PROTECTED] wrote: Konstantin Svist [EMAIL PROTECTED] said on Oct 04, 2007 15:54 -0400 (in part): Kevin Cozens wrote: Dotan Cohen wrote: There is an amazing video describing an image resizing algorithm that removes (or adds) pixels based upon their importance to the subjecto of the photo. Can this be implemented in The Gimp? Of course it can be implemented in GIMP. It just takes someone with the time and ability to write the code for a plug-in that implements the algorithm. I don't think the one you point to below is the same as Liquid Resizer (see http://schumaml.gmxhome.de/downloads/gimp/ and d/l http://schumaml.gmxhome.de/downloads/gimp/gimp-lqr-plugin-0.1.3.zip for it). Looks like someone has implemented it already: http://registry.gimp.org/plugin?id=5479 http://zinx.xmms.org/colorize/ Trying to compile the supporting libraries.. it's a PITA, unfortunately. On linux (ubuntu 7.10), at least, it's simple. Install 'libsuitesparse' and the matching 'dev' package, and apply the following patch to the Makefile: --- Makefile.old2007-10-01 05:54:38.0 +0930 +++ Makefile2007-10-06 15:48:22.0 +0930 @@ -6,7 +6,11 @@ GIMP_CFLAGS=`$(GIMPTOOL) --cflags` UMFPACK_CFLAGS=-I/usr/include/suitesparse -I/usr/include/umfpack -UMFPACK_LDFLAGS=-lumfpack -lamd -lcholmod -lm +# FIXED -- works on Ubuntu 7.10; +# Only requirement is libsuitesparse package (make sure it's installed cleanly; +#'umfpack' package conflicts with it. I removed umfpack and reinstalled libsuitesparse +#+ libsuitesparse-dev.) +UMFPACK_LDFLAGS=-lumfpack -lamd -lcholmod -lcolamd -lblas -llapack -lm CFLAGS=-g -Wall -O3 $(GIMP_CFLAGS) $(UMFPACK_CFLAGS) LDFLAGS=$(GIMP_LDFLAGS) $(UMFPACK_LDFLAGS) Using this, I currently have the colorizer plugin installed :) My experiments have not turned out as well as those shown on the page, though. Nice! But on Fedora there's only ufsparse library (as of now, anyway) and SuiteSparse has to be installed by hand :( As for email not making it to the list.. I think it's because whoever maintains this list forgot to set reply-to header. To work around this, just use Reply All and (optionally) delete the user's email, leaving the list email only. ___ Gimp-user mailing list Gimp-user@lists.XCF.Berkeley.EDU https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-user
