Hey guys, i’m having problems with getting rid of banding created by monitor/internet display limitations.

Now i’m aware that there are dithering setting that can be applied to a render.

However the problem comes from adding a gradient like effect after the render.

In my image, its is a vignette. So is there anyway to “treat” the vignette to improve the gradation?


use the ‘dither’ setting. it’s somewhere in render buttons I think. don’t quote me on that.

I don’t think you read my post :slight_smile:

right you are, I did not read the entire post. sorry about that. what about creating the vignette as an image file with an alpha channel in GIMP, then compositing it in?

i’m looking for fix within blender so that I don’t have to do that.

plus that wouldn’t be option for animation for example :slight_smile:

At some point your grad is to subtle and there aren’t enough colours to spread over the distance of pixels. The only solution is to introduce noise in the composite, like adding fake dither. Be aware that if you compress the delivery format like sending to YouTube you may get the banding back.

hi 3point, is there a proper technique for doing what described?

all i’ve done now is just mix a noise texture (via mix node) with the end result of the blur node, and the results are already much better. I want to know if that is the proper way though.

thanks a lot for the explanation

Nope you just want to vary the value for the colors to simulate variation over a field (area) of the image. It’s fake interpolation, relying on the eye to blend the colors more smoothly. I guess you may want to target the area that has the banding and not the rest of the image though. Can you pull a mask from the common area to apply the noise in that place?

I understand, thanks. I did exactly that, a vignette is made from a mask anyway, so I just applied the noise to the non masked area.
I’d share the results but unfortunately it’s a private project. It can look, well, noisy, but with a bit of tweaking can probably be perfected. Also depends how large the image is viewed

many thanks. It’s a decent workaround.