It starts at the camera

Modern color correction and grading systems allow you to performance a lot of magic in post. All too often though, producers take the “fix it in post” mantra way too seriously. Since modern cameras offer excellent low light sensitivity, the basics of good lighting and proper art direction get short-changed. If the director, DP and lighting crew are allowed a bit of time to exercise their craft, the results will be so much better in the end. Grading can make up for lighting deficiencies, but often at a cost of noise and color artifacts. Here are three basics to remember that will greatly improve your next production.


Proper exposure is essential, but with improved low light sensitivity and an increased use of log-based gamma curves, operators often have a hard time deciding what the right exposure actually is. I’ve mocked up some images to help clarify some of my points. (Click on any of these images throughout this post for an enlarged view.)

Most log curves mathematically “bend” or “compress” the high-end and low-end range of what the sensor is picking up. It’s a way of squeezing a wide dynamic range into a recordable space. The typical mantra is to “expose to the right”. This can be interpreted as making the image brighter by setting the exposure so it displays in the bright section (the right side) of a histogram. In you expose too dark, you will effectively underexpose the image. When you try to make it brighter during color correction, you will increase noise, especially in the midrange. That’s because there’s not enough of an image to gracefully “stretch” without introducing artifacts, such as noise, posterization and banding.

To illustrate these points, I have taken this RED One shot and created various exposures from the raw file. These were then exported as “baked” ProRes files, which I subsequently graded using FCP X’s color board tool. I took the corner of the image into Photoshop and exaggerated the levels to make the defects in the dark area more obvious. If you review the larger images, you’ll see the one that started with the darkest simulated exposure as having the most posterized look, while the dark areas are smoothest in the shots that represented brighter lighting.

You want to be careful not to go too bright, since the opposite is also true. Highlights on hair make be too bright and you might have difficultly getting the image dark enough in the dark areas of the picture. Ideally, the image wants to cover an area on a histogram that’s roughly the middle third of the scale. On a waveform, skin values want to hit in the 50-60IRE range. This lets a colorist stretch up or down from there, without having to raise skin values “out of the mud”.

Contrast ratio

Next to having the image bright enough for good correction, you need to have some range to work with between the darkest and lightest portions of the picture. A typical waveform would show this as 0-100IRE, but that gets more difficult with log images, which start out considerably more washed out in appearance. Typically a 20-40 IRE spread in a log image, like ARRI Log-C or RED RedLogFilm, will yield excellent grading results, as these images attest.

Hue separation

One of the in-vogue looks is the “orange and teal” style popularized by blockbuster films and emulated by Magic Bullet Mojo and Looks. This is based on the color theory of various color models. But to get the look in a convincing way, you really need to start out with proper art direction. Skin tones tend to be pink-orange. If you shoot a very warm scene with an actor standing close to an orange wall, you’ve effectively set up a monochrome situation where all the hue values are the same and only saturation and brightness are changed. In that example, it’s very difficult to make the flesh tones and the wall color be different from each other. This can be fixed in the beginning by proper art direction and lighting.

A good place to go to understand and play with examples is Adobe’s Kuler website. The interactive color swatch tools are a good way of testing color schemes using scientific models of complementary colors, triads, etc.

©2012 Oliver Peters