077 Color nodes

- [Steve] Hi, this is Steve Wright welcoming you to this week's Nuke Nugget, Color Nodes: the Forgotten Ones. In this episode we'll take a survey of several color nodes that you may not have explored yet. The idea here is to simply understand what they can do, not how to operate them. Once you know there's a node out there that will do what you need you can look it up when you need it. So let's start with the Histogram node, which, of course, plots a histogram of your image. It also has a built-in Photoshop Levels tool.

To demonstrate that I've got my Histogram hooked up to this Viewer here and a Grade node over here hooked up to that Viewer, so we can show you the parallels between the Levels tool and the Grade node. OK, so starting here, this guy right here is the black level, the blackpoint. Here we go, see that. OK, I'll undo this and undo that. This guy is the whitepoint.

I bring this down the image gets brighter and brighter. See, .42. OK, so I'll take the whitepoint down to .42 or so, there you go, OK. Undo that, undo that. And of course, the middle guy is the gamma slider just exactly like in the Levels tool. Now the interesting thing is that the number you see here in the gamma window does not match the same number you'll see in this gamma window, even though they are doing exactly the same thing, they actually do kind of an invert of it.

We won't worry about that right now. And down below here, well, let me put this guy back to default, there we go, down below here this is the lift operation. See, I'm lifting that up. And if I come to by Grad node and do a lift I can get exactly the same look there. So we'll undo that and undo that. And lastly, this is the gain. So as I slide this down I'm gaining down the image and I can take my gain over here and gain it down the same amount and I'll get the same results.

Notice that the Histogram node clamps your blacks and whites, which means you cannot use it for any high dynamic range images. Let's cruise over here and take a look at the histogram eqaulization node. Let me clean up my work space here by closing this pane, closing that, opening this up, and opening up my histogram equalization node.

So the histogram equalization node distributes an image across the full spectrum of your colorspace. I'll open up the Histogram node and you can see that this image at the moment doesn't cover very much of our colorspace, but if we put it through the histogram equalization, which, as you can see, has blown it wide open, you can see it's completely redistributed the histogram for the image. Next, let's take a look at the Posterize.

So we start with this ramp and the Posterize node cuts it up into however many different colors you say. In this case, by default, 16 colors, which turns it into essentially a 4-bit image. And you can see I have 16 lovely bands of what used to be a smooth ramp. We can use that to what they call posterize a picture, like this, and that will introduce all this lovely banding into all the channels of your otherwise pretty picture. Don't know why you want to do that, but if you want to you can with the Posterize node.

MinColor. MinColor finds the darkest pixel in your picture and then adds an offset to lift the whole image up, so that that darkest pixel reaches whatever value you have put in as the target. So let me show you this. This is a film scan here, (mumbling) image, and I have, of course, negative code values. You can see that right here, let's cruise in here. And if you look down below here you can see I got negative code values everywhere I go.

So if I set my target to 0 and then I tell it to find my pixel data what it's going to do is cruise through the image and find that lowest code value and then add a constant offset to lift the entire image up to where that code value will be set to the target of 0. So I'll say Find Pixel Data, OK, then we'll see what we got. There you can see, you can see there's been a little bit of a lift, see, that whole picture got a little brighter there.

And then if we cruise in here again and blown up the gamma I no longer have any negative code values at all. The lowest code value I'll ever find anywhere is the target 0. Put that back. And we'll cruise over here to the SoftClip. So the SoftClip, of course, introduces a soft clip. So I've set up a little test bit for you here. Don't need that guy. What I have is a ramp that goes from zero to two.

In fact, if I gain it down, slide my cursor way up here, watch my code values below, there we go, it's almost at two. Put the Viewer gain back. If I drop the Viewer gamma you can actually see where the picture changes from 1.0 and above. So the SoftClip will fix that, so let me put up my Sampler node here, sample the current frame, get me a little more screen space here.

So this is showing that this ramp goes from 0 to 2. All right, so let's turn on our SoftClip and resample and look what happens. So you can set a maximum value and a threshold, anything below the threshold is untouched. You also get to choose between three different algorithms for your SoftClip, three methods to apply the SoftClip. RolloffContrast.

Again, we have my ramp, zero to two, and we'll put up a Sampler node here. Now this does a similar thing, it's going to apply an S curve to the data. So let's turn on our guy here, click the button, and there you go. So it has applied a SoftClip at the top and the bottom with a nice S curve. This is used to increase the contrast of an image, but avoid clipping.

That's what you want to use it for. So if I bring down my gamma, oh, I want to bring down the gain, there we go. So I cruise through here and the largest code value I'll find is 1.0, but if I go upstream of it you can see how I've got very large code values in the fire. Viewer back to normal and slide this guy back up here and let's go see what else we got. OK, the Toe. This applies a soft clip to the blacks.

So there's my ramp again. Again, a zero to two. Put up a Sampler node, so we can watch the action. Sample the current frame, then we'll turn on the Toe, like so, get my Sampler back. And now we'll tell it to resample, and there you go. And of course, you can see the value for that Toe, how much it ramps off, of course. And we use that in a case like this you can apply a Toe to it and you're not going to really see anything, so that there's no clipped black regions.

It will gently roll your shadows off and avoid that clipped black area. So that's the Toe. MatchGrade, this is interesting. So I'm hooking my Viewer up to this Grade and here's the original clip, to this graded version, and then I have the raw clip over here. Now what this MatchGrade does is it specifically finds the Grade to match the same clip. So if we come down here to mister MatchGrade, bang.

It has graded this clip to look exactly like that one. Now both look identical. If I turn this off I get back to my original clip. It'll also output the Grade itself. That way you could apply the Grade to another part of your Nuke script. This is to me the most magical node of all. I have two utterly different hideous clips. This one was properly exposed, (mumbling) my Viewer for you, this one was hideously underexposed.

Let me clear this out. What this does is it actually remaps the colors from this one into that one, like this. Unbelievable. To me this is really impressive. It is not doing a histogram match. Histogram matches don't work for images with disparate content. The content has to be incredibly similar. So the ColorTransfer node is designed to match two completely different clips, whereas MatchGrade is designed to discover the Grade for the same clip.

So there you have it, Nuke's forgotten color nodes. Be sure to check in next week, so you don't miss out on my next Nuke Nugget.