HOME | DD

Description
This is why nodes are a sucky way to do things. Making nodes is programming, just program already and stop with trying to protect us from ourselves.

This is my attempt at a 3x3 convolution matrix. It doesn't work right now, in part, because Blender apparently clamps RGB values without telling you. So to make this work, I'd have to separate and recombine RGB for all 9 of the cells. So like another 50 nodes or so.

But I don't think I'm going to fix it. Because there's no "image" input node. So even if I make this a node group, you'd have to change the image field on every cell rather than getting them from a single input.

And besides which, I really don't want a 3x3, I want a 5x5.

I find that there's very little rhyme or reason to how Blender node inputs and outputs work. Is a Tangent type a vector? Apparently not to Blender. Why do mapping transformations have separate settings for Normal and Vector? Or even for anything: if you don't want to translate a vector, then just don't translate it. Can you plug color into value or vice versa? Depends. If you do, what do you get? Depends. Is some particular output type clamped? Depends.

Sometimes I wish that it would just have one type, float. At least then I wouldn't have to use all those separate and combine nodes.

Just irritating because there's a lot you can do with a convolution matrix. Like texture blurs, and edge detection, sharpen and unsharp mask... Even if you don't want to render with it, would be a useful tool for in-Blender texture generation.

Description
This is why nodes are a sucky way to do things. Making nodes is programming, just program already and stop with trying to protect us from ourselves.

This is my attempt at a 3x3 convolution matrix. It doesn't work right now, in part, because Blender apparently clamps RGB values without telling you. So to make this work, I'd have to separate and recombine RGB for all 9 of the cells. So like another 50 nodes or so.

But I don't think I'm going to fix it. Because there's no "image" input node. So even if I make this a node group, you'd have to change the image field on every cell rather than getting them from a single input.

And besides which, I really don't want a 3x3, I want a 5x5.

I find that there's very little rhyme or reason to how Blender node inputs and outputs work. Is a Tangent type a vector? Apparently not to Blender. Why do mapping transformations have separate settings for Normal and Vector? Or even for anything: if you don't want to translate a vector, then just don't translate it. Can you plug color into value or vice versa? Depends. If you do, what do you get? Depends. Is some particular output type clamped? Depends.

Sometimes I wish that it would just have one type, float. At least then I wouldn't have to use all those separate and combine nodes.

Just irritating because there's a lot you can do with a convolution matrix. Like texture blurs, and edge detection, sharpen and unsharp mask... Even if you don't want to render with it, would be a useful tool for in-Blender texture generation.