On Thu, Jun 4, 2015 at 7:19 PM, Ulrich Pegelow <ulrich.pege...@tongareva.de> wrote:
> Definitely looks like a bug. Same happens here with other images as well. > > The root cause seems to be the "flip" function in crop&rotate. It > produces a white (?) line at the left of the roi when activated with > mode "horizontal", on the top of the roi with mode "vertical" and on > both sides if the two are combined. The shadow&highlights module makes > this white line more visible. > > Here is how to best see: > > 1) open image in darkroom (ideally an image with homogeneous color, e.g. > blue sky) > 2) zoom into image (e.g. 200%) > 3) activate crop&rotate with horizontal flip > 4) switch on shadows&highlights with radius 200 > 5) move the roi around: at each position you see the left side being > lighter than it should > > So we have a likely problem with the roi calculation in the module which > is most complex in this aspect :( > Hi. I have seen similar c&r issue happening because of c&r flip feature: https://www.dropbox.com/s/gm9n6l9dgibmy7e/out-5.ogv?dl=0 And there is even a fix: https://github.com/darktable-org/darktable/compare/master...AlicVB:mask_jump But: [00:07:28] <alicvb> LebedevRI: I've create the branch : https://github.com/AlicVB/darktable/tree/mask_jump but as it can change old image, even at 100% export, I'm really not sure it's ok to push it (i have cc'd alicvb because of the branch) > > Ulrich > Roman. > Am 04.06.2015 um 13:26 schrieb thokster: > > > > Here is an image with blurred foreground. There it is obvious. > > Try toggling on/off crop and rotate with horizontal flip. > > BTW a (horizontal) flip is changing the log./lin. histogram. Wouldn't > > expect that but that's probably some missing knowöedge on my site. > > > > OT: I would expect that only the wavefront histogram is flipping. The > > wavefront changes as expected but only twice when toggling. > > > > > https://dl.dropboxusercontent.com/u/352146/Shadhi%20and%20flip/P1220469.RW2 > > > https://dl.dropboxusercontent.com/u/352146/Shadhi%20and%20flip/P1220469.RW2.xmp > > > > > > regards > > thokster > > > > > > Am 03.06.2015, 19:14 Uhr, schrieb Ulrich Pegelow > > <ulrich.pege...@tongareva.de>: > > > >> Hi, > >> > >> I cannot reproduce. Shadows&highlights its default settings uses a > >> strong gaussion blur. If by some effect the input to this module > >> contains even a 1 pixel defect line at one of the borders, this would be > >> smeared by the gaussian blur to a visible size. > >> > >> Not possible to decide based on what I know. > >> > >> Please open a ticket in redmine and attach an example raw+xmp. > >> > >> Ulrich > >> > >> Am 03.06.2015 um 09:56 schrieb thokster: > >>> Hi, > >>> > >>> using the shadows and highlights module and flip the image horizontal > >>> makes a light border at the left side of the image. > >>> > >>> Is this a known behavior? > >>> Can anybody reproduce it? > >>> > >>> Tested with git-master and different raw files (Nikon, Canon, > >>> Panasonic). > >>> > >>> regards > >>> thokster > >> > > > ------------------------------------------------------------------------------ > _______________________________________________ > Darktable-users mailing list > Darktable-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/darktable-users >
------------------------------------------------------------------------------
_______________________________________________ Darktable-users mailing list Darktable-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/darktable-users
