More frequently, a grading operation will reduce apparent contrast by increasing the brightness of midtones, perhaps using a gamma operation, which can make artifacts in problematic areas such as graduated sky less apparent. Doing this, though, robs information from one tonal range of the image – say, shadows – and donates it to another –midtones. This means that information from the shadow areas, which tends to be poor in any case, now serves as the new midtone. Shadow contrast must therefore be increased, risking increased visibility of artifacts. This is an unavoidable compromise of contrast manipulation in these circumstances.

The other factor affecting all representation of luminance on DSLRs is that they exclusively record 8-bit images, the greyscale being represented as one of 2⁸ = 256 levels. Equipment intended to maximize grading potential typically uses 10 bits, providing 2¹⁰ = 1024 levels, as well as preprocessing the image using a curve to distribute brightness information evenly among them. This is often referred to as a “log curve” since the shape of part of it may approximate some kind of logarithm, but most camera manufacturers freely define these curves according to the characteristics of the camera’s particular hardware.

These two techniques greatly reduce the visibility of contouring, or banding, the visibility of sharp edges in areas where a gradient should be continuous. Since DSLRs offer neither high bit depth or log encoding, and because contour edges are often made irregular by compression artifacts and can become visible after only very slight contrast increases, this is a particular problem.

Color and saturation

Increasing saturation is dangerous because all DSLRs store chromaticity information at a lower resolution than luminance, simply because the human eye has fewer color-sensitive sites than it does brightness-sensitive ones. When the material is viewed straight out of the camera, this assumption holds up, and the material looks reasonable, and  subsampling (which is used by high end equipment as well) isn’t the main reason that grading up saturation can be a problem for DSLRs. The problem is that these lower-resolution colour channels are as even more likely to see artifacts than the main luminance image, because their lower resolution makes artifacts bigger and potentially more objectionable. Shooting a low-saturation scene may also cause the codec to allocate comparatively little bitrate to colour, exacerbating the issue still further.

Motion

Although it isn’t directly related to colour and brightness, the motion in a scene can affect the gradeability of an image. Compressors are often designed to take advantage of the fact that large areas of a moving image are likely to be moving in the same direction all at once, an assumption which can break down badly in certain cases. The classic example is shooting through a chainlink fence from a car, where the scene represents both the motion of the fence and any action occurring beyond it.

With modern equipment, these situations should not cause an absolute breakdown of the image. However, they may cause the encoder to allocate a lot of bitrate to encoding the motion of the scene as opposed to its brightness and color information.

Mitigation

The simplest workaround is simply to avoid grading, and it is a reality that the DSLR image will always be less suited to extreme manipulation than many other systems.

One saving grace, particularly of Canon’s EOS series, is that there are options to make changes to colorimetry in camera. This is not comparable to the per-shot flexibility or advanced selective grading of a software solution, though it allows a director of photography to achieve a result close to the intended result in-camera, in concert with lighting, optical filtration and production design. Minimizing the requirement for grading is the best option, even at the cost of additional time spent on set.

Where grading must take place, good photographic technique is critical – a well-exposed image that nevertheless avoids excessive clipping will place the majority of the scene’s information in the most lightly-compressed range. A waveform monitor makes this easier, as do correct optical filtration, production design, and lighting. In camera, selecting options such as Canon’s Highlight Tone Priority, or creating a curve that pushes the image toward the intended result, may also help, since this work is done before compression.

It is difficult to recommend the use of “super-flat” or other extremely low-contrast curves unless that is the photographic intent. Producing an extremely low-contrast image with the express intent of recovering that contrast later may produce an image with pleasing tonal range but risks severely exacerbating artifacts.

The other thing we can do on set is simply to avoid using so much compression. This has not historically been an option with DSLR because the cameras lacked means to obtain uncompressed video. However, at the time of writing the Panasonic DMC-GH2, which appears to have uncluttered HDMI output, was being discussed as the potential camera partner to an uncompressed (or less compressed) recorder.

In postproduction, mitigation can start as soon as the image comes off the camera. Usually, we will transcode the camera original file to an intermediate format. Tools such as Rarevision’s 5DtoRGB (which also works on other cameras) do a much better job than Quicktime-based tools at reconstructing full-resolution colour difference channels, using much more careful mathematics.

Finally, if all else fails, we can attempt to manually clean up unacceptable artifacts. In many situations noise is less objectionable than artifacts or contouring, and the addition of a very small amount of noise can obscure problems. This is sometimes termed “error diffusion dithering”, although a more overtly stylistic choice such as simulated film grain can have a similar effect.