> I guessed from what Sven said, inverting the image colors first might help, and I was right. > I made my second try by: > > * Inverting the image > * Selecting the entire image in the initial 'lasso' pass. > * Disabling 'Contiguous' > * Setting L,a,b sensitivity to 0,707,555 respectively. (a,b was just a guess, but L was 0 to accommodate the harsh brightness contrasts of the widgets) > * Switching to background mode > * Scrawling a bit (~3 strokes) on the windows > > Overall it was quite straightforward, actually. > http://img.photobucket.com/albums/v449/neota/tech/gimp/screenshot-2006-06-21-try2.png > (selected areas marked in blueness -- the window 'shadow' got selected, but otherwise it seems quite satisfactory.) > .. My experience above causes me to wonder if 'BGselect' could be made by inverting the L*a*b values before interpreting them; maybe something to try after 2.4.
I do not quite understand your problems. I am an aloof developer who has serious problems to understand user's problems. Please help me out, maybe I am misunderstanding something? So please do not get me wrong here. What one defines foreground or background is not a matter of the tool but a matter of the human being who is using the tool. I cut out the windows selecting them all with the lasso. See: http://www.gerald-friedland.de/tmp/multiwindow_sel.png Then I disabled contiguos because the Windows in this image are not connected (may be "disconnected" would be a better string?). Then I needed a few foreground strokes, mainly to select the KDE toolbar (or was it GNOME) and to include all the colors shown in this GIMP tool dialog. The result is then: http://www.gerald-friedland.de/tmp/multiwindow_cutout.png If you want to have the screen-background instead of the windows, use "select/invert" and you get the background instead of the foreground... No need to fiddle with color sensitivity....? Greetings, Gerald -- Gerald Friedland Raum 164 Tel: ++49 (0)30/838-75134 Freie Universität Berlin Takustr. 9 http://www.gerald-friedland.org Institut für Informatik 14195 Berlin gfriedland@xxxxxxxxxxxx _______________________________________________ Gimp-developer mailing list Gimp-developer@xxxxxxxxxxxxxxxxxxxxxx https://lists.XCF.Berkeley.EDU/mailman/listinfo/gimp-developer
