Individual Qualifier Controls
This section presents the typical onscreen controls you'll find in most HSL Qualifiers. When used together, each control contributes to a highly selective chroma key, letting you quickly sample a range of colors and refine the resulting key for the best possible effect. When used separately, these controls allow you to perform single-qualifier secondary color correction, isolating portions of an image based on individual characteristics of your image as interpreted using the HSL color model.
Eyedroppers and Color Pickers
Most HSL keyers have an eyedropper tool (Figure 4.5) with which you can click on or drag an image to choose an initial value or range of values to begin creating your key. It's a great way to start out if you're unsure exactly which range of values you need to select (which can be tricky with subtle ranges of color), even if you plan on immediately switching to manual adjustment of the qualifiers after you've found your starting point.
Figure 4.5 The eyedropper found in Apple Color.
Some applications have just one color sampling tool, with keyboard modifiers enabling you to use the same tool to add to or subtract from the key by scrubbing different pixels of the image. Nucoda Film Master and Colorista II both use bounding-box sampling. Draw a box over the region of picture you want to sample, and the matte is started.
Other applications, such as DaVinci Resolve, have separate onscreen controls for explicitly sampling a primary value, expanding and subtracting from the range of values, and expanding or narrowing the range of softness (tolerance) of the matte (Figure 4.6).
Figure 4.6 The Color Picker options presented by DaVinci Resolve. From left to right: center color selection, add to matte, subtract from matte, add softness, and remove softness.
However your controls are set up, once you click or drag with your application's eyedropper/color sampling tool, you should see the Hue, Saturation, and Luma qualifiers expand or contract to reflect the color vales of the pixels you're sampling.
Choosing an initial value or range of values with which to begin your key can be a tricky business. The noise, grain, and variation in detail inherent to any image will produce a range of color, even from pixel to pixel. That means that even if you think you've just clicked on a light part of the image, you might have inadvertently clicked on the one dark pixel of noise in that area, giving you a disappointing result. Don't worry, this is bound to happen, just sample other pixels until you find one that gives you the best initial result.
Here are two general tips for choosing a good initial value:
- Choose a color value that's right in between the lightest and darkest values you're trying to key, starting at the middle and working your way outward.
- If the subject you're trying to key has an uneven border with another subject—for example the sky against the horizon—and you're trying to key the sky, choosing a value that's near the border at the horizon is a good way to start building the best part of your key where it's most important.
Viewing the Mask While You Work
Most color correction applications provide you with three different ways of viewing the key you're creating as you work. This is essential to creating high-quality keys. If you only view the final result, is can be difficult to see problem areas like unwanted holes in the matte, excessive noise, or chattering edges. These three modes are as follows:
- A high-contrast grayscale matte-only view (Figure 4.7). I find this to be the most useful preview of the matte I'm creating, as it's drop-dead obvious to see the quality of the white areas that represent the keyed area (the inside of the correction), and the black areas that represent the unkeyed area (the outside of the correction).
Figure 4.7 A high-contrast matte, as seen in nearly any color correction application capable of HSL Qualification.
- The keyed area in color against the unkeyed area in grayscale or solid color (Figure 4.8). This can be a useful view for seeing how your key is shaping up, while also seeing which parts of the inside or outside of a key correspond to shadows or highlights in the subject you're isolating that you may or may not want to include in your key. The high-contrast mode can make these kinds of judgments difficult if you don't remember the image's details.
Figure 4.8 A saturated vs. matte mode, as seen in DaVinci Resolve, that shows you exactly which parts of the image are being included in the key. Other applications show the unkeyed area as simple grayscale, rather than a flat matte.
- The final effect. For some applications this mode is redundant, as the matte and final effect are viewed simultaneously on different onscreen portions of the UI. However, if you're viewing the key matte on your output display, you'll need to switch to this viewing mode to go back to making the correction once you've finished refining the key.
Ideally, it's good to create and adjust your key while simultaneously viewing the matte you're creating in your application's canvas or preview window, or even output to your broadcast monitor. Surprisingly, I've found that clients love watching these mattes being made, so monitor output isn't a bad thing.
The Individual Qualifiers
Once you've sampled an initial range of values to get your key's starting point, the next step in the process is to fine-tune the matte using the qualifier controls. As seen previously, these controls typically follow some common conventions across different applications. In general, each color component has a set of handles and parameters that allow you to adjust one of two sets of values:
- Range handles let you enclose a base set of values that contribute to the hardest, whitest part of the key.
- Tolerance (or softening) handles let you include an additional set of values that add a gradual falloff around the edge of the key, which appears as a range of gradually diminishing grayscale values surrounding the hardest part of the key. Widening the tolerance yields softer edges and a more inclusive key, and narrowing them yields harder edges and a more restrictive key. If the edges of the key are gentle enough due to the softening that tolerance provides, you may not need to blur the resulting matte.
Enable/Disable checkboxes let you manually decide which qualifiers you want to use for a particular keying operation (Figure 4.9).
Figure 4.9 The GUI of a typical qualifier control, as seen in Apple Color. Each set of controls typically corresponds to one or two control panel knobs.
Additionally, there are two general methods used for manipulating controls. The methods used depend primarily on how these controls are mapped to the rotary knobs on a control surface.
- Centered controls work using three knobs per color component. The first knob moves the center point of both pairs of range and tolerance handles all together to the left or right, a second knob expands and contracts the range and tolerance relative to the center, and a third knob expands and contracts the tolerance relative to the current position of the range handles.
- Asymmetrical controls use four knobs to move the left- and right-hand sets of controls of each qualifier separately. Two knobs for Low and Low Soft adjust the left-hand range and tolerance, while two more for High and High Soft adjust the right-hand range and tolerance separately.
If you're using a GUI rather than a control surface, you might actually be able to manipulate the qualifiers in either of these ways, using keyboard modifiers while dragging the appropriate handles.
The Hue qualifier lets you select a portion of the color spectrum to isolate a range of color within an image. The spectrum is continuous, wrapping seamlessly from the left of the control to the right of the control (Figure 4.10).
Figure 4.10 The Hue qualifier graph as seen in DaVinci Resolve.
If you turn on the Hue control without turning on the Saturation control, you'll select all color values within a particular range of hues without regard to intensity or lightness.
The Saturation qualifier lets you select a range of saturation, or color intensity, within the image. The black or colored end of the ramp represents 100 percent saturation; the white or gray portion of the ramp indicates 0 percent saturation (Figure 4.11).
Figure 4.11 The Saturation qualifier graph as seen in DaVinci Resolve.
If you turn on the Saturation control without turning on the Hue control, you'll select all color values within a particular range of saturation without regard for the actual colors themselves.
In a keyer, this qualifier lets you isolate a range of luma, the lightness component of an image (the Y' in Y'CbCr). The black end of the ramp represents 0 percent luma; the white end of the ramp indicates 100 percent luma. In some implementations, there's an additional region of this control that represents the super-white range from 100 to 110 percent (Figure 4.12).
Figure 4.12 The Luma qualifier graph as seen in DaVinci Resolve.
If you turn on the Luma control without turning on the Hue or Saturation controls, you'll select all areas of an image within a particular range of lightness, without regard to the color, similar to creating a luma key. This frequently yields the greatest amount of edge detail with highly compressed media, and it is a commonly used technique for segmenting an image that we'll see later in this chapter.
Post-Key Mask Utilities—Blur and Edge Manipulation
Once a key has been pulled using the eyedropper/color sampling and qualifier controls, the result is a grayscale image. Most HSL keyers provide additional filtering operations that you can use to manipulate this image to make it more usable in situations where a key was difficult to pull, or when the key has noise, chatter, or other undesirable artifacts that would give the resulting correction away.
Although these operations lend themselves to quick fixes, make sure that you've exhausted the possibilities of the qualifier controls to pull the best possible key, and be careful not to use settings that are too aggressive. If you overdo these types of operations, you may wind up with halos or other problems that might actually be worse than the artifacts you're trying to avoid.
Softening simply blurs the keyed matte. This function is useful for eliminating spurious areas of keyed detail that you don't want affecting the final correction, as well as for minimizing the visible effect of animated buzz and noise in the matte (Figure 4.13).
Figure 4.13 A matte, before and after adding blur to soften the edges and minimize marginal portions of the key.
Oversoftening the matte and then creating a correction could result in the same correction being applied to an area surrounding the intended subject, producing an unwanted glow where the correction spills onto the over-feathered area (unless you're intentionally using a secondary to create a glow operation, a process covered in Chapter 10).
Some applications also have a Shrink parameter (or whatever the corresponding control in your application is named). This is usually an averaging control similar to the matte choke function found in compositing applications (Figure 4.14).
Figure 4.14 Left, the original mask; center, the matte after using the Shrink parameter to expand it, filling out holes; right, the matte after using the Shrink parameter to contract it, eliminating unwanted marginal areas of the matte. These mattes were created in DaVinci Resolve.
As seen in Figure 4.14, this control expands or contracts the edges of the matte:
- Expanding the edges of a matte is useful for filling in small holes appearing in a troublesome matte.
- Contracting the edges of a matte is useful for eliminating spurious pixels and unwanted details in the matte, as well as for shrinking the overall matte.
Using shrink all by itself can sometimes result in blockier edges, which is seldom desirable. In these cases, adding some blur will smooth out the result (Figure 4.15).
Figure 4.15 A shrunken matte that's been blurred to address the blocky edges.
Inverting the Key
Color correction applications deal with whether corrections apply to the white portion (inside) or black portion (outside) of a matte in one of two different ways:
- Applications such as DaVinci Resolve and Assimilate Scratch let you invert the keyed matte, switching which part of the image a correction affects.
- Applications including Autodesk Lustre, FilmLight Baselight, and Apple Color let you explicitly set whether a correction affects the "Inside" or "Outside" portion of the matte.
- DaVinci Resolve's node-based correction interface also lets you create an "outside node" that you can use to grade the opposite portion of a key relative to the original node you created.
Inverting a key is an ideal approach if the subject to be excluded from a color correction operation is uniform enough in color and lightness to be selected with an application's Limit Effect controls, and you want to apply the secondary color correction to everything else.
For example, if we wanted to adjust everything within the frame except for the dress, we can still pull our key off of the dress since it's the easiest thing to do (Figure 4.16).
Figure 4.16 At left, the original key pulled off of the woman's dress. At right, the inverted matte, useful for adjusting everything except for the dress.
Then, using whatever control inverts the matte, you can reverse the black and white regions, so now the dress becomes the area that won't be affected by the secondary correction, while the rest of the image is now set up for the correction you want to make.