This chapter is from the book
This chapter is from the book
This chapter is from the book
Different Ways of Using HSL Qualifiers
Image segmentation describes the process of divvying up an image into individually adjustable segments for purposes of performing highly specific color corrections (Figure 4.35). The more you work, the more you'll spot easily segmented regions of the frame, and the faster you'll be able to deploy the most suitable secondary tools for the job at hand.
)
Figure 4.35 At left, the original image. At right, the image has been artificially color-quantized to show how the image is naturally segmented for purposes of creating isolated corrections using HSL Qualification.
Segmented adjustments using secondary corrections is a key strategy, both for addressing specific problems and for stylizing an image. Although you can do a solid grade using only a single primary, it puts the burden upon the cinematographer to have anticipated and created the careful pools of light and dark, color and colorlessness, using a combination of lighting instruments, flags, and art department trickery, and working with the costume department to control the visuals' impact on the viewer.
This kind of control over the practical photography is not always feasible on lower-budget features with limited schedules, or documentary productions where you don't have a whole lot of control over where you shoot, when you get the shot, and what people are wearing. As a result, the colorist's job grows.
The following sections illustrate different ways of using the very same tool, HSL Qualification, to achieve very different creative goals.
Isolating and Adjusting Specific Elements
This is the most basic use of HSL Qualification. In particular, I find the most common application of HSL Qualification in my work is to address problems with saturation.
As we saw earlier in this chapter, a common operation is dealing with subjects that are too distractingly saturated. For example, the image in Figure 4.36 has had a primary correction to increase contrast and boost saturation. The result, unfortunately, is that the balloons in the background have become excessively bright. Using a qualifier, it's easy to isolate the balloons and make them more subdued.
)
Figure 4.36 The balloons and ribbons at left are distractingly saturated. Isolating them with a key lets us slightly reduce saturation so they don't draw our eyes away from the actor.
Another technique for dealing with the opposite issue, a lack of saturation, is to enhance the color contrast of an image by using "contrast of extension" (covered in Chapter 3) to create the illusion of higher saturation by using HSL Qualification to boost the saturation of specific subjects in the frame, rather than boosting the saturation of the entire image.
For example, Figure 4.37 has plenty of saturation but not a lot of color contrast. An easy fix for this is to isolate the woman's sweater in the image, and then increase the saturation. The resulting splash of color gives life to an otherwise blah image.
)
Figure 4.37 Isolating the woman's lackluster sweater lets us pump it up with a bit more color, bringing welcome color contrast into an otherwise monochromatic scene.
Using this strategy, you get to have your cake and eat it too: You can even desaturate an image, yet by selectively boosting the color of one or two objects, make it appear as if it's even more colorful.
Correcting Two Exclusive Areas
We've already seen how you can invert a key to leave one element alone when you want to correct everything else. Another segmentation strategy involves using an HSL Qualifier to key an easy-to-isolate region, and then apply separate corrections to the inside and outside of the matte (Figure 4.38).
)
Figure 4.38 A basic correction is applied with node 1, and nodes 2 and 3 apply separate corrections to the inside and outside of a keyed subject, respectively.
This is an especially useful strategy when you're trying to protect or uniquely grade a specific subject. For example, the woman's green dress in Figure 4.39 is a deliberately specific shade that the client wants graded in a particular way, even though the environment is supposed to be graded very differently (in this case a very warm yellow).
)
Figure 4.39 The green dress needs to be keyed to isolate it from the rest of the image in order to produce the grade the client wants.
Using HSL Qualification, it's an easy thing to key the dress to isolate it from the rest of the picture.
Then, you can apply a correction to the inside (white) of the resulting matte, grade the dress however you need it to appear (in this case darker and more saturated), and apply a second grade to the outside (black) of the matte, grading the environment separately to blow out the highlights and add a burning warmth (Figure 4.40).
)
Figure 4.40 Grading a scene with two exclusive corrections, one applied to the inside of the keyed dress to boost its greens and darken the color, and another applied to the outside of the dress to blow out the environment.
How you approach grading the inside and outside of a matte depends on your grading application:
- In DaVinci Resolve, create an "outside node" from the correction node with the dress qualification.
- In Apple Color, choose Inside from the Control pop-up menu to switch that secondary to adjust the inside of the mask.
- If you're using Autodesk Lustre, click the Inside button.
- In Assimilate Scratch, duplicate the scaffold that has your initial qualification, and invert the mask in this new scaffold.
- In FilmLight Baselight, select the layer with the qualification, and choose an appropriate grade type from the Inside column of the Inside/Outside table.
Many NLE filter-based secondary corrections handle this functionality by duplicating the filter doing the keying and inverting the filter to correct the outside region of the matte.
For a detailed example of using this technique to segment skin tone adjustments and environmental adjustments, see Chapter 7.
Grading Light and Dark Regions of an Image Separately
In many instances, you may be dealing with an image that has areas of overexposure combined with areas of underexposure, such as when an underlit subject is standing next to a window, or sitting in a car in front of a window. Backlit windows are notorious for being difficult to expose, and unless the crew was able to apply sheets of neutral density over the windows, or bounce some fill light onto the subject in the scene to balance the exterior light with the interior, they're often deliberately overexposed.
If, in a situation such as this, the extremely bright region of the picture (the window) isn't overexposed to the point of losing all visible detail and the interior subject isn't too horribly underexposed, you should be able to use HSL Qualification to solve the problem. In particular, you can use the Luma qualifier all by itself to isolate either the highlights or the shadows—whichever portion of the image is more cohesive (Figure 4.41).
)
Figure 4.41 Luma-only key as created in DaVinci Resolve.
Then, you can use the resulting matte to grade the under- and overexposed areas of the image to bring them closer in line with one another (Figure 4.42).
)
Figure 4.42 Keying the darkest parts of the car exterior makes it possible to lighten the deeper shadows, adding detail to the man's face and the dashboard of the car.
Aside from window shots, this technique is effective for any clip where you have to deal with two areas that are exposed very differently. However, be careful of how much the softened edge of the matte overlaps from the shadows into the highlights of the image; too much overlap can result in haloing that will look artificial.
Controlling Shadow Contrast
One of the hardest things to deal with is the adjustment of shadow ratios. For example, it's easy to generally lighten or darken a subject overall, although the result may appear a bit washed out. Unfortunately, if you're matching a shot from late in the day into a scene that was photographed at noon, or if you're asked to change the apparent time of day of a scene, you're likely going to have a shadow ratio mismatch that you'll want to attempt to fix.
However, it's not that easy to alter the ratio of shadow to light that falls on someone's face. In the uncorrected image on the left in Figure 4.43, there's a distinct shadow on the woman's face, with a highlight on the other side, that indicates higher-key, dramatic, afternoon lighting.
)
Figure 4.43 Keying the shadow regions of the image makes it possible to lighten them, gently, in an effort to reduce the shadow ratio for a more noonday look.
One strategy would be to use a luma curve to lighten the midtone shadows, while keeping the deepest shadows to maintain a good sense of image detail. However, you might not be able to make a specific enough adjustment using just curves.
The other possibility is to use the previous technique, trying to carefully create a secondary key using the Luma qualifier to isolate the midtone shadows on the face that you want to lighten. Coupled with an adjustment to cool off the highlights, this does the trick, and we now have a lighter, less high-key shot, as seen on the right in Figure 4.43.
When attempting to lighten the shadows on a subject's face, you need to pay particular attention to the transition between the unadjusted highlights and the adjusted shadows. If this middle-point starts to reveal itself as a solarized border, you'll want to adjust the tolerance and/or feathering that you're applying to the Luma qualifier to try and minimize it. If that doesn't solve the problem, you may need to back off of your correction a little bit.
When you create this sort of correction, make sure you don't overdo the shadow lightening or you'll solarize the image (create lighter areas where they should be darker, and vice versa, for an early-'80s music video look). The other possible consequence is that, if the feathering and brightening isn't done just right, the image could start looking like bad HDR (high dynamic range) photography.
Isolating a Subject with Desaturation
A commonly asked-about technique is that of isolating a color subject within an artificially created black-and-white environment, as used in the movie Pleasantville. From the techniques we've covered so far, this should now be a easy nut to crack, especially if the subject you're trying to isolate possesses a unique range of hue, saturation, or luma.
Basically, key on the subject you want to keep in full color, invert the matte, and then desaturate everything else (Figure 4.44).
)
Figure 4.44 Keying the woman's pink dress, inverting the key, and desaturating everything else.
Doing a Better Job by Combining Mattes
The previous example in Figure 4.44 is fine if we're happy with the one dress being in color and everything else being in black-and-white. However, things are rarely that simple. For example, what if we wanted both women to be in color? Fortunately, some applications let us combine mattes from different keys.
Figure 4.45 shows how, in DaVinci Resolve, we can use four separate corrections to isolate the pink dress, the red hair, both women's skin tones, and the other woman's green shirt, and then combine them using a Key Mixer node.
)
Figure 4.45 Using multiple nodes in DaVinci Resolve to combine four separate HSL keys to use as one. Each individual key can be optimized to isolate one subject well.
The resulting combination matte can be fed into the mask input of a correction (node 6), inverted, and used to desaturate everything we didn't key, all at once (Figure 4.46).
)
Figure 4.46 Desaturating everything outside of the combination key shown in Figure 4.44. I left the imperfections to illustrate just how difficult an operation this is, and the importance of spotting these issues. I count 14. (You'll find a complete list at the end of the chapter.)
Now, the example in Figure 4.46 could obviously be further improved with some more aggressive keying combined with shape/Power Window garbage mattes, but hey, I have a book to finish.
The same technique can be used in the Color FX room of Apple Color by combining multiple mattes created with HSL Key nodes and Shape nodes using the Lighten and Darken nodes. This is worth spending a bit of time on, as the Apple Color node interface makes it explicit how you can combine mattes using blending modes (sometimes called transfer or composite modes) to create different shaped mattes. These blending modes exist in other grading, compositing, and editing applications, and you should be able to adapt these techniques to your own application if they're supported.
Figure 4.47 shows how you can combine two mattes using a lighten blending mode, with the result being that the white interiors of both mattes are merged together to become a single matte. This happens because the lighten blending mode compares both images and preserves the lightest pixels within each.
)
Figure 4.47 Combining shapes using a lighten blending mode to combine two matters in the Color FX room of Apple Color.
Next, we can see in Figure 4.48 how the darken transfer mode can be used to preserve the white interior only where two mattes overlap. This is, in fact, the very method that's used to use HSL keys in combination with shapes/Power Windows to limit the keyed matte to within the area of the shape (we'll see this combination later on in Chapter 5). The opposite of the lighten blending mode, the darken mode compares two images and preserves the darkest pixels from each.
)
Figure 4.48 Combining shapes using the darken transfer mode.
Finally, Figure 4.49 shows how you can use the darken blending mode in a different way, using one matte to carve out the overlapping part of another matte. In the node tree, you can see that Vignette2 has been inverted. Since the darken blend mode compares two images and preserves the darkest pixels in each one, this means that the now black portions of the inverted Vignette2 node cut into the Vignette node.
)
Figure 4.49 Using the darken blending mode with Vignette2 inverted.
If you're having a tough time isolating every part of the subject you need to adjust, you might try using combinations of mattes instead.
Altering Saturation Within a Specific Tonal Range
Although some applications have specific controls for highlights and shadow saturation that let you adjust specific zones at the extremes of image tonality, if you lack these controls, there is a simple way to customize a tonal region for adjustment of saturation.
All you have to do to use a Luma qualifier all by itself is to select a range of image lightness within which you will decrease or increase saturation (Figure 4.50).
)
Figure 4.50 Using the L (luma) qualifier in DaVinci Resolve to isolate a range of image tonality within which you can adjust saturation.
Once you've defined the area of the image to adjust, it's a simple thing to lower or raise saturation to achieve the necessary result. In Figure 4.51, you can see that a much larger area of the highlights have been isolated for saturation adjustment than is typically possible using some applications' highlight saturation controls.
)
Figure 4.51 A custom tonal region of the image can be selected using the Luma qualifier for tailored saturation control, in this case desaturating a broad range of highlights.
Furthermore, you can make separate saturation adjustments to the inside and outside of the matte you've defined. This can be a good way of boosting midtone saturation without needing to subsequently lower saturation in the highlights and shadows for purposes of video legalization.
This is also a good technique for isolating a limited range of highlights or shadows with an extremely soft falloff for subtle adjustments.
Desaturating an Image and Selectively Resaturating It Using HSL Qualification
This next technique is one that Chicago colorist Bob Sliga showed me. It's a bit more specialized, but it comes in handy for both creative looks and the occasional utilitarian adjustment. This technique is only possible on grading applications that let you pull keys off of the original source media, rather than the graded state of the image.
The trick is to desaturate an image completely and then pull a key based on the original colors in the source media, as shown in Figure 4.52.
)
Figure 4.52 Pulling a key based on the original colors in the source media.
Using that key, you can add artificial color back into the scene to "colorize" a subject you want to give a special look to (Figure 4.53). A node tree for creating this effect in DaVinci Resolve can be seen in Figure 4.52.
)
Figure 4.53 At left, the desaturated image with boosted contrast. At right, the portion of the image that was colorized using the key pulled from the original color in the source.
This technique is great for giving an otherworldly look to people in special situations. However, I've also used it for shot blowups that have excessive chroma noise in a brightly colored portion of the image. Desaturating the offending subject (an apron in the example I'm thinking of) eliminated the most objectionable noise pattern, and I could then make it red again by keying and colorizing. The final result was a vast improvement, and the effect was nearly indistinguishable.