3D Scopes in Color

By Alex on October 1, 2009 3:30 PM | Permalink | Leave a Comment(0) | TrackBacks (0)

In addition to the usual waveforms and vectorscopes, Color offers a 3D scope which scatters the pixels of your image through a 3D graph of the gamut of your selected color space. It's great for "client wow factor," but I've still not found a student who can think of a really good practical use for the view. Sure, you can roughly assess the distribution of pixels as you attempt to balance and guarantee consistency across shots -- but the freely-manipulable 3D perspective strikes me as too imprecise to use for much more than you could do with more normal scopes already.

The 3D scopes do offer one feature that almost all of my students agree is useful: if you click any of the three swatches at the bottom of the scope, you can pinpoint the values associated with one specific point in your image (by clicking and dragging on the Preview scope). In the 3D scope, beside the swatch you have selected, you'll see text indicating the specific color values associated with that point in the color space you're viewing. Even cooler (but probably less useful), the 3D scope will use bold lines to "triangulate" the selected point inside the mass of points in the 3D field.

In the remainder of today's article, I'll post quick snaps of the different color spaces represented in 3D -- but in the meantime, do you have any killer practical uses for the 3D scope itself? Leave them in the comments -- the commenter who convinces me that they're uniquely useful on the regular will get a prize.

The RGB graph scatters points inside of a unit cube with each of the three channels (red, green, blue intensity) on its own axis. Notice that it might be easier to think of this space as a diamond rather than as a cube: near the origin (where all of the axes meet at 0), you find darker points; lighter points appear near the opposite corner of the cube, near where a point with high values on R, G, and B would appear.

The HSL graph may or may not strike you as more useful. It's a polar transformation of RGB space, so it consists of a ring around a center line. The ring is sort of like the control wheels in the correction rooms. It represents both hue and saturation: the direction a point falls indicates what hue it has, while the distance from the center axis indicates its saturation in that color. Finally, the center axis (I think of it as "vertical," but you might not) indicates the luma of the point. Respectable color folks disagree about how the "ring" should be theoretically connected to the ends of the line to define the "edges" of the space (does it make a cylinder, or a bicone?), but in practical terms I've found it useful to visualize the limits of the space like Color does: as a bicone, i.e. two cones butted up against each other with their pointy bits up at the ends of the luma line.

The YCbCr gamut is straight-up weird, but it does represent the broadcast-legal gamut more closely than RGB/HSL space. Because of the way composite signal reduces chrominance data to two channels, it can represent a whole heck of a lot of colors (the entire RGB gamut, plus a lot more), but some of those colors don't actually exist when the TV tries to convert them back to RGB space for display. But on the linked page, look at the weird orientation of the RGB cube in YCbCr space: if you move along any of the YCbCr axes, you'll hit the edge of that RGB shape sooner with some intermediate colors than with others. The YCbCr gamut works much the same as the HSL space, except that the goofy hexagon in YCbCr view more reliably represents the legal limits of YCbCr space.

Finally, the IPT gamut works a lot like the RGB gamut, except it's weighted to more closely represent how the human eye perceives different intermediate ranges of the different channels. If you're using the 3D graph to answer a technical question, use the RGB gamut; if you're using it to perform perceptual adjustments on the image as a whole, you may prefer the IPT gamut.