«
»

United against Banding

Dear lazyweb, do you know any method to render subtle gradients in SVGs dithered? Or is there a renderer that outputs 16bit/channel PNGs? Banding is ruining our efforts in GNOME, openSUSE and Fedora!

This entry was posted on Wednesday, August 15th, 2007 at 1:45 pm and is filed under Artwork, Inkscape. You can follow any responses to this entry through the RSS 2.0 feed. You can leave a response, or trackback from your own site.

7 Responses to “United against Banding”

  1. MDK Says:
    August 15th, 2007 at 3:06 pm

    Check http://benno37.livejournal.com/10678.html .

  2. David Schroeder Says:
    August 15th, 2007 at 3:35 pm

    How’s this for a roundabout-enough solution:

    Take whatever image you’re working on, without the fine details. Export that, bring it in to Krita at 16 bits/channel, do a blur on it, then add a little noise, then convert to 8 bits/channel, then add the finer details.

    And it’s reassuring that FOSS isn’t alone with this problem. I’ve run into this exact issue with Illustrator, and the general conclusion I’ve found is there’s no way to help it.

  3. Hans Nieser Says:
    August 15th, 2007 at 3:46 pm

    This is exactly what’s been bugging me with Inkscape. I actually thought there was some kind of Inkscape-specific bug with gradients, since I couldn’t see any banding in gradients created with GIMP, as it turns out GIMP does dithering by default and my LCD screen is to blame. This makes Inkscape somewhat unfit to design (pixmap) graphics for websites with unless one isn’t bothered by the banding that will be seen on most people’s LCD screens.

  4. ajax Says:
    August 15th, 2007 at 4:22 pm

    I’ve been thinking about how to add automatic dithering to cairo for a while now to avoid this problem. Basically you know the, uh, gradient of your gradient at any given point, so if it’s sufficiently flat along either axis that you’d create a band of solid pixels more than (say) ten wide, then add half an LSB of noise to the internal representation of the pixel just before rasterising. If you’re evaluating the fragment color internally at 16bpp before flattening to 8bpp (and I think we are) then this should give you a random dither for basically free.

  5. jimmac Says:
    August 16th, 2007 at 2:03 am

    ajax: show, show, show! :)

  6. M??ir?­n Says:
    August 21st, 2007 at 2:40 am

    Oh this came up on the inkscape-devel list today…


    http://sourceforge.net/mailarchive/forum.php?thread_name=3c3cb8048846fb3293764fd73db0473c%40localhost&forum_name=inkscape-devel

    Wow sourceforge mailing list archives and links to said archives really suck. Sorry about that!

  7. jimmac Says:
    August 21st, 2007 at 12:28 pm

    /me sighs on the note of putting that as a preference switch.

Leave a Reply