[Bf-committers] Patch: New compositor nodes

Frédéric van der Essen fvdessen at gmail.com
Mon Aug 29 15:49:27 CEST 2011


On Mon, Aug 29, 2011 at 9:11 AM, j.bakker at atmind.nl <j.bakker at atmind.nl>wrote:

> Hi Frederic
>
> I have checked the nodes to identify any problems with my project
> (Compositor redesign).
>
Thanks, I really look forward to the completion of your project :)


> I don't see any issue Vibrance, the HCL nodes and having a separate gain
> node. The gain function is hidden in the color balance node.
> Posterize: There is just a small difference between Math round and the
> Posterize node. But the math round is only based on values, as this one
> also concerns colors.
>

The posterize node also works on channels, it is the same as having a Math
round on each channel. The real difference is gamma correction, I wanted the
node to give similar results to other packages such as gimp and photoshop,
and they do it in non-linear RGB. So it can be done with math round, but it
would take 7 nodes to do it.

It is the Same for vibrance, it could be done with  HCL sep/comb nodes and a
curve, it's a tradeoff between ui clutter, code bloat, feature
discoverability, and ease of workflow. Not sure where to draw the line ?



> I + these nodes, but we should be aware that implementing ocio in blender
> will in the future change the HCL nodes.
>

No problems about that. Talking about color spaces, when implementing the
rgb_to_lab transform I noticed that one of the rgb_to_xyz transform matrix
were not the same as in the reference documents I found on the subject. I
put a comment in the code next to it. (math_color.c -> xyz_to_rgb() -> case
BLI_XYZ_CIE ) Anybody knows the reason ?


>
> Pixelize+ToonAA: This is more a question if these should be part of a
> compositor. Perhaps someone else can answer this.
>

The Pixelize node seems to be popular in the feedback I recieve. No feedback
on ToonAA yet.


> ToonAA: I see some issues as the algorithms normally are used to upscale
> 2x2 or 3x3 pixels and this node is tends to be implemented as a convolution
> filter.
>

EPX cannot be implemented as a true convolution filter, however it could be
put as a case in the filter node to reduce UI clutter.


>
> What are your thoughts about the last two nodes?
>
> PS. in your comments you mention pixelprocessor, but don't use it. (ToonAA)
>

I am not very familiar with blender's codebase, so I used other nodes as
templates, and there are some parts I don't understand. So I left those
parts as they were, they are good indicators of what I could have done
wrong.


> Jeroen
>
>
> Original Message:
> -----------------
> From: Frédéric van der Essen fvdessen at gmail.com
> Date: Sun, 28 Aug 2011 17:14:22 +0200
> To: bf-committers at blender.org
> Subject: [Bf-committers] Patch: New compositor nodes
>
>
> Hi!
> I'd like to get this patch reviewed:
> It adds a few nodes to the compositor :
>
> Vibrance:
> Increases the saturation but prevents clipping and oversaturation.
> Video: http://frederic.vanderessen.com/videos/node-vibrance.ogv
>
> Hue Chroma Luminance:
> Adds a Hue Chroma Luminance control,separation,combination nodes.
> The HCL Colorspace is similar in spirit to the HSV, but is based on the
> Lab Colorspace, and gives more natural results.
> Video: http://frederic.vanderessen.com/videos/node-hcl.ogv
>
> Gain:
> Increases or decreases the dynamic range of the picture.
>
> Pixelize:
> Pixelizes the picture. The size of the blocks and the method used to
> pick the
> blocks colors can be changed.
> Video: http://frederic.vanderessen.com/videos/node-pixelize.ogv
>
> Posterize:
> Reduces the number of colors in the picture. Similar to Gimp's posterize
> filter.
>
> ToonAA:
> Anti-Aliasing filter for pictures with a small number of colors. Based
> on the EPX pixel art upscaling algorithm.
>
> The patch is in the tracker
> :
> http://projects.blender.org/tracker/index.php?func=detail&aid=28386&group_i
> d=9&atid=127
>
> You can find a Debian Testing 32bit build here
> :http://frederic.vanderessen.com/blender_builds/blender-nodes.tar.gz
>
> Thanks!
> -Fred.
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers
>
> --------------------------------------------------------------------
> mail2web.com - Microsoft® Exchange solutions from a leading provider -
> http://link.mail2web.com/Business/Exchange
>
>
> _______________________________________________
> Bf-committers mailing list
> Bf-committers at blender.org
> http://lists.blender.org/mailman/listinfo/bf-committers
>


More information about the Bf-committers mailing list