Banding as a tell-tell sign of postprocessing

Nikolai

Whenever I deal with any type of artificailly created gradient (I use PS, but I don't think it matters), be it a result of a blur or a gradient tool I immediately start noticing banding issue when resizing.
What bugs me the most is that when a seemingly identical gradient comes from the camera (example - light's falling off on a solid color background), the issue doesn't exist at all or only presents itself at extreme compression/reduction values.
What I would like to be able to do is to "revitalize" such artificially introiduced gradients to their original, non-banding-prone quality.
Any ideas?

DavidTO

Add a little bit of noise to the area that's banding.

arodney

Are you working with more than 8-bits per color all the way through?

Are you sure its banding in the image, not on the display?

pathfinder

Both responses are excellent points, Nik.

It does not take very much added noise to defeat the banding in a gradient.

Computer generated gradients are notorious at being displayed with banding, some of which can be display sizing related, and not seen at other image sizes or in prints. But adding a bit of noise should cure it.

Nikolai

David, Andrew, Jim
Thank you for the quick response guys!
Will try the noise!

BinaryFx

Nikolai wrote:

Whenever I deal with any type of artificailly created gradient (I use PS, but I don't think it matters), be it a result of a blur or a gradient tool I immediately start noticing banding issue when resizing.

An edit to a *perfect* CG grad will degrade the gradient (simple matter of input tones to output tones before and after the edit). When making a CG grad, one can choose to add a minor LSB "dither" effect to help avoid banding - although usually if there will be banding then one may need more aggressive dither than what is built into the gradient tool or in colour conversions.

What bugs me the most is that when a seemingly identical gradient comes from the camera.

One is computer generated, perfect.

The other (camera, natural gradient) is far from "perfect" (although a raw may be better than a JPG).

You are comparing apples to oranges.

What I would like to be able to do is to "revitalize" such artificially introiduced gradients to their original, non-banding-prone quality.
Any ideas?

Sure - don't edit them, or if you have to, regenerate them.

The above is not so practical, so this is why some folk make noise about noise.

http://members.ozemail.com.au/~binaryfx/howto_smartnoise.html

Further reading:

http://www.ledet.com/margulis/ACT_postings/ColorCorrection/ACT-Banding-16-20bit-8-bit.htm


Stephen Marsh
http://members.ozemail.com.au/~binaryfx/

Nikolai

Thank you, Stephen!