Vision Unlimited/LA News

Someone wrote in to a forum last week, asking about the “proper” video levels for other-than-Caucasian skin tones. The subject of where to place skin tones on a waveform monitor is an oft-discussed one and most folks don’t know what the drivers are for making a technical decision. Moreover, the technical decision is not the most important one, even if the image that results is technically compromised in some way. EVEN TECHNICALLY IMPERFECT VIDEO MAY BE THE RIGHT CHOICE FOR A PARTICULAR SCENE. Only the persons responsible for the product can decide that.

In the early days of analog video, creating video signals outside technical standard limits could cause problems in recording and transmission, and could get a broadcast station in hot water with the Father of Communication Corrections (FCC) in Washington, DC. Those days are behind us and most broadcasters will not kick back productions for video violations; they are more likely to kick back a production if they can figure out that it was shot with a “sub-standard” camera, and the FCC is more worried about “wardrobe malfunctions” than blanking width. However, if the subject is compelling, the show will sell, and that is another oft-discussed subject.

So, about that skin tone thing…

Conventional wisdom dictates that Caucasian skin tones should reside around 70-80% video on a waveform monitor. Ignoring the aesthetic issues concerning the rest of the scene, the mood of the scene, and the video or filmic “look” of the show, where did that number come from? It came from a desire to keep skin tones higher than the average background video level, while still trying to prevent any clipping or even flattening of detail in the skin tones from over-exposure.

The appearance of over-exposure in a scene can come from at least four different variables in a video camera system, other than the obvious iris or gain settings.

White clip is the setting that determines the white limit of your video and in a digital video signal, it is somewhere below 109% of reference white. White clip can be set as low as 100% for some applications and pushing skin tones that high could obviously cause flattening of skin tones, but you don’t have to go all the way to clipping to see flattening effects.

Gamma is the term applied to the brightness response of a camera to light. We use gamma to more closely match the tonal response of a camera to that of the human visual system. If we tried to capture light linearly, like many raw camera sensors do, we would need much more information in dark scene areas than we do in bright areas. Our visual system responds roughly logarithmically, stretching the dark values and compressing the bright values, so that we perceive a 1% change in brightness about the same, regardless of value. A 1% change at 1 lux (dark) is much smaller than a 1% change at 100 lux (bright), so to be able to see the dark change, we would have 100 times more information than we can use in the bright area. The effect of gamma on skin tones is that as we move up the brightness curve, the detail in the skin gets smaller, compared to that in the original scene. Hopefully, the gamma curve matches that of our eyes, but if you have made significant changes in the gamma settings of your camera, that might not be the case.

Knee point and slope settings are used in many cameras to reduce the brightness sensitivity of a camera above a preset value (knee point), at a controllable rate (knee slope) so that all detail in very bright parts of the scene is not lost to hard clipping. The reduction of the gain in the bright areas causes the remaining detail to be reduced from what it would be if you simply lowered the overall exposure. Of course, if you lowered the exposure, the entire scene would get dark; using the knee control just affects the brightest areas. The catch is the setting of the knee point – if you set it too close to your skin tone value, you will start to flatten the skin tones and get a “pasty” look.

But the overall knee setting might not be too low and you can still get that pasty look. Why? Because the knee function operates on all three color channels before they are combined into a luma (brightness) signal, and we know that the values of the three color channels are not equal in any of the luma equations. That’s right, there are more than one. The equation used in modern HD cameras is not the same one that is used in older SD cameras. Standard converters should compensate when using one format with another, but the values can be significantly different. For example, SD white represents about 60% green, whereas HD white contains 70%, that leaves less for red and blue. That’s another catch – if you increase the saturation of your skin tones, either by adjusting the color balance or the overall saturation (both of which are possible in most cameras), then the red content can become a more significant percentage of the luma value of the skin, and it is easier to push the red channel into the limits of knee or even clip.

So, even if you are just concerned about the technical aspects, you must look at the picture monitor for clues to problems that might be occurring because of your camera setup. Some will be too subtle to see on the waveform monitor but will show up in your images.

The original post referred to non-Caucasian skin tones, which present different issues based on the same variables as discussed here. As the luma value of the skin tone goes down, the skin may be harder to distinguish from the rest of the scene and the saturation value may go up, so pushing exposure to make the person stand out more may give similar problems to those experienced with Caucasians.

The bottom line here is that lighting is the tool that separates subjects in a scene, not the engineer’s exposure knobs.

cr/30