Question(s) about dithering

[Joe]

I've always wondered this.  Does all of the dithering in a piece have to fit on the same grid to work effectively?  As in, if I have a 50% patch of red and yellow on the bottom-left corner, would it be more subtle if the yellow conformed to the same exact grid as say a 50% yellow and green dither in the top-right corner?

Basically what I'm asking is if it matters, and how much better would it look.  If I'm not being clear, I can definitely expound on this.

[1ucas]

Interesting question. I don't think it really matters unless the dithered boundaries actually meet.

[Joe]

Hm.  I suppose there isn't really a way to prove it.  But when I rapidly switch my eyes back and forth between the top two squares, the one on the left just seems(obviously at the pixel level it is) misaligned, and the set on the right seem to match better.

[Rawsushi]

I want to respond, but I'm not sure I get the question.

Is the idea that you want to combine two 50/50 dithers specifically?
I certainly don't believe that all dithering needs to adhere to a grid.

[Joe]

No, nothing in specific.  Just imagine a color you're using in a particular piece.  What I'm asking is, wherever you place that color in the piece, if you dither with it, is it a more effective dither if it conforms with the same grid over the entire piece, or does it not matter and it's more effective to dither the particular section accordingly with no regards to the dither as a whole.  I use 50% dither as an example because(and correct me if I'm wrong) I consider it to be the basic form from which the other patterns (25%, 75%, etc) are built off of.

[ndchristie]

dithering and other visual technicques are a "see first formulate second" thing.  In general you will be able to see the difference between which is or is not more subtle, and as other have identitied this only matters at the edges of dithering regions, not across a piece.

[Jad]

The eye is very good at spotting patterns in things, but is also prone to misinterpreting patterns and can often be swayed by optical illusions.

So nah, I don't think the eye is good enough to be able to imagine the whole dither spectrum across a piece.

When 50% dither patterns touch, though, it can be important to think in a system of 'dark color grid' and 'bright color grid'

if I remember it right, in the old 'LET'S COME UP WITH TECHNIQUES AND NAMES FOR THEM' days they called this "interlaced dithering" and Peppermint Pig was hot at it O:

Does anyone have any art from him saved anywhere? Can't find anything ):

[Rawsushi]

At the risk of being completely wrong, was it something like this? I wasn't around at the time.

[Jad]

OH MY GOD THE YELLOW CHANGES COLOR COMPLETELY SUCH A SMOOTH GRADIENT

THE HELL!! PIXEL MAGIC

[Gil]

At the risk of being completely wrong, was it something like this? I wasn't around at the time.

No, not at all, I'll try to make an example of interlaced dithering later tonight. Pep used it in skintones a lot.

[Reo]

This is it I think

[Gil]

Yes, that's it exactly.

[Helm]

The exact same principle is applied in Rawsushi's example so I don't know where you get your 'not at all' Gil.

[Joe]

At the risk of being completely wrong, was it something like this?

This is it I think

Oh, wow, I didn't know anything like this existed.  Is there any writing on this?

[Gil]

The exact same principle is applied in Rawsushi's example so I don't know where you get your 'not at all' Gil.

In Rawsushi's the yellow doesn't change at all. The thing that Pep called interlaced dithering is a specific form of gradient dithering that makes colors come back in the gradient multiple times. It's a very specific effect that tends to eliminate banding, while letting two dither regions hug each other.

[Helm]

huh I've used this interlace dithering since forever and I've never seen it that way, as far as I can tell the idea is PRIMARILY to blend on the dither instead of off of it really. But I'm sure you have some reason to disagree

[Gil]

As Jad mentions, this is part of the whole "inventing techniques" era of pixel art. I'm not defending its use or anything, just trying to place it historically.

It's comparable to selout as being a useless technique that anyone who does some examining picks up on his/her own anyway.


See, I don't ALWAYS disagree with you

[Helm]

Here's some things one can do with working on top of the dither as much as they do outside it.

I don't mind the invention of technique names as much as I used to and I'd like to see this sort of dithering used more so if the name helps popularize it that's great. The problem with selout was not the name, it was the concept behind it being considered prescriptive for a period: (without wanting to turn this into another selout thread) I mean 'if you want to be a pro pixel artist, put these broken outlines around your sprite!'.

The whole premise that this helps readability of a sprite against a background color of any type was a mistake in my opinion. Usually used in Capcom sprites, most of them very big sprites, nobody's eyes will stop tracking chun li if she passes in front of a blue background exactly her color even very fast, really.

[Gil]

St0ven had this cool way of positioning his shading diagonally, so the dither turned into almost paint streaks at the edges. I liked that too.

I have to admit though, I'm finding less and less use for dither unless it's for for example chainmail on a sprite.

[Helm]

I find dithering extremely useful personally and every time I decide to go without for some reason or another I feel like someone's taken half of my pixelling skills away. Without it pixel art is almost like low grade cg painting, what's left is just the manual anti-alias and palette control.

[Gil]

I used to have my dithering period, but now it's almost the opposite. I'm still finding out about my own style and techniques.

I used to create stuff like this, with almost 100% dithering:

[Jad]

25% dither looks like complete ass in my opinion, my eye spots the square pattern immediately. If you manage to hide it well with low contrast/clever use of noise or busy areas of art, then congratulations! If it makes things smoother without being seen then it has actually served its purpose!

Also yeah, interlaced dithering for brushed metal/chainmail surfaces looks pretty cool I'd say! I like using it for metal!

Like I did on the hat of that one!

Ok, very little actual interlaced dithering but I think what's there does a lot to give that illusion of contrast between speculars on the uneven surfaces of brushed dark metal

[Joe]

There's some very useful information here, thanks guys.  But it's led me to have more questions, if you don't mind.



For A, or in thinner clusters, is it better to use AA or to dither?

And for B, Jad mentioned how 25% dither looks shitty unless it subtle.  I think that's a pretty universal feeling, although I hate 50% more if it's not subtle.  But my question is, if I only have those two colors, is it better to make a 50% dither cluster, and then dither between all 3 or to do a straight gradient?  A better question would be, what's the best way to create a subtle or unnoticeable dither with very few colors?

[ptoing]

A) Depends how many colours you wanna use. If you do not have a lot of space for your detail, AA is probably better. A good practise also is to try and get even lines (as in regular stepping)
B) Again, depends on what you are going for. If you use more than 50% or better said other than straight mathematical dithers things will look more organic (look at Arachne's work for examples)

As for how to make unnoticeable/subtle dither with few colours = low contrast between the colours you are dithering with. This is the only thing that works universally across all monitors you might view it on. On old CRTs 50% dither usually generates something that looks like a perfectly new colour even if the ditherpair has quite high contrast. On LCD screens and newer CRTs not so much, dither will always be noticeable unless the colours are quite close in value.

[Joe]

Ah, okay.  I will take note of these things, and I will definitely check out his work.  Thanks

[ptoing]

...and I will definitely check out his her work.

[Joe]

his

Whoops.

[crab2selout.png]

The eye is very good at spotting patterns in things, but is also prone to misinterpreting patterns and can often be swayed by optical illusions.

So nah, I don't think the eye is good enough to be able to imagine the whole dither spectrum across a piece.

When 50% dither patterns touch, though, it can be important to think in a system of 'dark color grid' and 'bright color grid'

if I remember it right, in the old 'LET'S COME UP WITH TECHNIQUES AND NAMES FOR THEM' days they called this "interlaced dithering" and Peppermint Pig was hot at it O:

Does anyone have any art from him saved anywhere? Can't find anything ):

Whatever happened to Pep? Did he tire of the pixels? Or did he just have enough of moderating?

[Larwick]

And for B, Jad mentioned how 25% dither looks shitty unless it subtle.  I think that's a pretty universal feeling, although I hate 50% more if it's not subtle.  But my question is, if I only have those two colors, is it better to make a 50% dither cluster, and then dither between all 3 or to do a straight gradient?  A better question would be, what's the best way to create a subtle or unnoticeable dither with very few colors?

I'm unsure why the standard grid-style 25% dither is so popular (apart from the fact it's simple and seems the next logical step from the 50% dither). I can't remember who I learnt this from, but I'm pretty sure it was someone on Pixelation. Not sure what to call it or how to describe it, but I've always tried my best to go with option A rather than B wherever I can. It gives a much smoother gradient, even with very few colours.



Apologies for the exaggerated and bad fading-to-black part of option B, I just rushed that bit heh.

[Ai]

can't remember who I learnt this from, but I'm pretty sure it was someone on Pixelation. Not sure what to call it or how to describe it, but I've always tried my best to go with option A rather than B wherever I can. It gives a much smoother gradient, even with very few colours.

I want to point out that the 'square-ish' dither is more expandable, if you need
the extra dithering levels..
Here, I have smoothed the A and B transitions as much as I could see how:


Is there any way to further smooth A so that it provides as many levels as B??

[Jad]

I don't think there is, but on the other hand depending on the piece you can often get away with A because you won't need more levels to create smoothness.

Pattern B looks like lace!

[Larwick]

Great point! Seems like B would be much better for longer gradients or larger areas - but for smaller things the grid is more visible.

[happymonster]

I've done a bit of work on dithering in the past, both with hand drawn art and in the area of programming to see how to get the best results. I'll post what I came up with in a separate thread shortly as it might be of interest to some people here.