strange lines in sky..

[andrewjohn81]

in this case I don't really think it got Worse. I do agree with the others that it's still there though. I think "moving it around" in some cases is good enough for sure. If you can't see it, it's good enough in most work. Just look at the movie Transformers. haha (just kidding, it's better than I could do I'm sure).

But I agree, the only way to get rid of banding is to have noise in the image.

I was just hoping for some way of creating good smooth looking gradients like that without bands.

[simmsimaging]

in this case I don't really think it got Worse. I do agree with the others that it's still there though. I think "moving it around" in some cases is good enough for sure. If you can't see it, it's good enough in most work. Just look at the movie Transformers. haha (just kidding, it's better than I could do I'm sure).

But I agree, the only way to get rid of banding is to have noise in the image.

I was just hoping for some way of creating good smooth looking gradients like that without bands.

Few cases are good enough though. Banding is one of those problems that will come back and bite you in the ass if you try to ignore it. Far safer to get rid of it. That said, if you want to see if it is 'good enough' you should check the individual channels to see how it shows up. The composite channel can be more forgiving in some cases, and you don't want it to look worse on your clients screen...

For sure though: if you have banding in RGB and you are dealing with an image for print you are *going* to have serious problems if you try to blur it/ignore it because separating to CMYK is going to make it way worse almost without fail, and it shows up on paper very quickly.

b

[percydaman]

i dunno, blurring the sky seemed to make it look alot better. The issue for me, is as others stated, it's all the adjustment layers and such that is usually the cause of banding. So collapsing it all isn't a very good option, especially if you have clouds and such, that have their own subtle gradients, hanging around.

[simmsimaging]

I for one never said it wasn't better, but better is not gone, and for me not being able to apply any adjustments is effectively useless, it's too handicapping for my line of work.

It's misleading to think that adjustments 'cause' banding in most cases though. It's a lack of detail information that is exposed by adjustments. Kind of like noise in shadows etc.

There are a lot of little tricks to help avoid exaggerating banding with adjustments, but nothing beats having a banding-free base image to work with. It's one of the few things I really miss about the film days.

b

[jujubee]

Like I said - I spent maybe all of 10 seconds lassoing it to a new layer, holding the selection than blurring. At first I didn't notice the banding wasn't 'completely' gone, but I'm sure I could have nuked it completely with a little more blur than I initially applied.

It is an improvement, and it works in a tight spot - especially if you don't have time to re-render.

From a personal standpoint/preference, I think saving out to .png is always better than .jpg in post. And the higher bits you go, obviously the better.

[MikeeMax]

for info, the output file in the example was a tiff.8 bits. the sky is a vray sky correct with color correct. the output image is banding free, but as soon as you apply correction layers in pshop, the banding shows up, as if it was hidden in here since the beginning. why is there a lack of detail info in vraysky ?

[jujubee]

To the best of my knowledge, the Vray Sky is composed of gradients. Maybe that's just the way it is...

[simmsimaging]

Like I said - I spent maybe all of 10 seconds lassoing it to a new layer, holding the selection than blurring. At first I didn't notice the banding wasn't 'completely' gone, but I'm sure I could have nuked it completely with a little more blur than I initially applied.

It is an improvement, and it works in a tight spot - especially if you don't have time to re-render.

Like I said, 10 secs or 10 hours, blurring only makes the problem different, or worse, but it won't make it go away. It can sometimes make it better, but then again: so can standing further away from the monitor

Anyway, I thought the point was how to eliminate it, but to each his own. It just takes some experience and testing to judge what is 'acceptable' levels of banding so it's safer/better to get rid of it entirely. That cannot be done with blurring IMO and experience.

The VraySky probably bands for the same reason that any gradient created in photoshop without sufficient dithering, or made through blurring image data does - no texture to break up the transitions and not enough transitions in 8bit colour (nor in 16bit quite often).

[jujubee]

Any 8-bit image should be able to more or less show a smooth gradient. I don't think it's so much a problem of it being 8-bit. The only time I've ever seen a nasty crappy gradient banding is coming out of Illustrator or if a gradient is stretched.

And I still don't see how you are saying the problem is worse...

But yeah, I agree it would be much better to eliminate it at the source before getting to that point - I just offered up one post option to consider as not everyone is a total Vray purist.

[simmsimaging]

I actually never said it was worse - must have been someone else. I just said it was still there, and that more of the same blur approach was not going to help. That's all. Anyway, I'm not trying to be pissy about it at all - no skin of my butt either way - just trying to save someone barking up the wrong tree and not knowing why it wasn't working for them.

If you haven't seen this in 8 bit image files then you have been lucky. I cannot count how many times I've worked on removing banding from image files going back Photoshop 2.0 It's a real issue, but more so in the age of digital cameras/backs that lack the forgiving quality of film grain.



b

[jujubee]

Well, if it's going back to the source... I think it may be the fact that he threw the VraySky into Colorcorrect. I tried this many years ago and I think I may recall this happening to me.

[plastic_]

It's just a 8bit issue. 16.7 M colors aren't that much if you think about it. It's only 256 steps hue/saturation/lightness. Large gradients are supposed to show banding in 8bit, it's just too low dynamic range. Blur can work if the banding was exaggerated because of adjustment layers in a 8bit workflow. Like increasing contrast reduces the dynamic range even more (in photoshop, the histogram tool can visualize that).

It should help to convert the photoshop file to 16bit, even if the original file is 8 bit, to get more headroom for adjustment layers.
But you shouldn't save out a raw rendered image with 8bit in the first place, it makes no sense, unless you are very tight on hd space. I save everything 16 bit. Even that can be too low, when composing multiple passes.

With such a large, subtile gradient, there will always be banding in the end, if it's saved as 8bit/jpeg from photoshop. in that case only noise/dither may help.

[andrewjohn81]

Any 8-bit image should be able to more or less show a smooth gradient.

I'd say less in about half my cases, which is a pretty bad percentage.
if you think about it, a gradient IS just a bunch of bands right next to each other, unless you are adding some texture behind that to break up the bands. If there are enough bands, you don't see them. But if your range in color from one end of the gradient to the other is very small the bands must be larger in order to change from one color to the next. There are only so many colors to work from. Unless you are going from fully saturated to completely non saturated color (and this example is about half way there) then you probably don't have too many colors to work from and will see banding in the original render.

I think it is agreed with most people here that the best way to get rid of the banding in subtle color changes would be to use some method of dithering. Now the question I would really like to see discussed is: If your render already has some banding in it what is your method of adding a dithering pattern to the gradient?

I haven't yet come up with a good method for doing that. The closest I can think of is maybe to create a very noisy selection of the area, then blur that. That should offset many of the gradient's pixels, while at the same time leaving some of them in tact thus mixing up the pixels instead of just pushing the bands around. Any other bright ideas? because I don't think that works well.

[ShaunDon]

Nothing more to add here, just wanted to say I'm really psyched everybody's talking about this. I've run into this kind of banding so many times before to unending frustration. Andrew, your explanation makes a lot of sense, I wasn't even thinking about the fact that most skies usually have less than 50% saturation, especially at the horizon, so there aren't nearly as many integer steps between hues to avoid banding. Keep talking guys, I'm learning something here!

[beestee]

If your render already has some banding in it what is your method of adding a dithering pattern to the gradient?

In Photoshop, you can use the Glass distortion type to introduce dithering...but you will still have to manually isolate the problem areas.