Vision Unlimited/LA News

A few days ago, someone wrote about film color correction to one of the lists that i read…

“The Rec 709 issue is a bit confusing because I’ve had D.I. houses tell me that they color-correct LOG footage in Rec 709 color space but aren’t changing the gamma or limiting the dynamic range.  So does that mean they are just using Rec 709’s color gamut but not its gamma structure?”

Another poster wrote that there is a big difference between the corrections actually being made to the program and the corrections being made for monitoring purposes. She suggested that for film work, most facilities seldom transfer to Rec 709 color and then go back for a film-out.  That may be true for film work, but most video production ends up in Rec 709 space and the limitations of that format are a big part of the post production process. So i thought i should write a little about the subject.

First, a good summary of Rec 709 can be found on Wikipedia.

Rec 709 specifies several parameters for HD images and some of them interact, so it is sometimes difficult to separate them out without knowing the processes that got you to the image that you are working with.

The transfer characteristics are defined for the RGB signals coming from the sensor and include a “gamma” exponent. They also include a spec for maximum gain at blacks, which is about 4.5x. The slope of the linear 4.5x portion at the low end of the curve turns into the gamma curve (0.45 power function) at a specified value around 10% output video level.

If you manipulate the RGB signals together, then you can change the overall transfer characteristics of the system without affecting the color (other than the brightness of the colors – it is useful from this point on to think about the brightness or lightness of the colors separate from the saturation and hue.)

If you manipulate amplitude of the RGB signals individually, then you change the color balance. If you change the gamma of the individual RGB signals, then you will affect specific colors and shades in unique ways and are probably leaving the realm of standard color references and making purely aesthetic decisions. There are reasons to change individual gamma settings to correct for camera setup problems but those are less common than creative changes.

Once you get past the RGB controls, you may have to deal with the transformation equations into color difference space. Color difference signals represent the same information as RGB, but using one luma (brightness) signal and two color signals, which are derived by removing the luma from each of the red and blue signals. They are supposed to be independent of luma but, because of the way that most cameras work, there is some cross-talk among the channels. Because of this cross-talk, and because the luma signal is composed of different amounts of R, G and B signals, changing the colors may affect the brightness, depending on where you make the changes. Color difference signals may also be sub-sampled in spatial resolution, but that is mostly a different issue. In fact, the filtering process of sub-sampling chroma can subtly affect color contrast because the edges are softened, but that is not usually an obvious problem because we can’t resolve as much color detail as luma.

Also, standard definition (SD) colors conforming to Rec 601 have different brightness levels than those of Rec 709, because the two standards use different amounts of RGB signals to make luma. If you want to see an obvious difference, look at SD color bars next to HD color bars, both viewed in monochrome (chroma turned all the way down.) The yellow, cyan and green bars are brighter in HD and the magenta, red and blue bars are darker than in SD. On a waveform monitor, the bars look odd to someone used to looking at SD bars. There is a big jump down between the green and magenta bars, instead of the even steps in SD.

The effect of the luma equation is probably the biggest difference between SD and HD signals and accounts for the fact that people think that HD has a larger “color gamut” than SD. There is a difference in the gamut because of the choice of slightly different red, green and blue primary color references, but the difference in the luma equations makes real colors look more contrasty, increasing the apparent saturation of an image.

If you keep your signals in RGB through processing, then the luma issues may not be obvious in your system, but if you leave the RGB domain and convert to luma and color difference for display, then some of these issues may surface. And of course, going to film out involves transformations beyond the knowledge of this video engineer, so the accuracy of your simulated display maps can make predictions of film results dead on or way off base.

The bottom line is that the mode of correction can most definitely allow changes in transfer characteristics (gamma) and color representation that appear to be independent, and which may create unnatural effects, because the processes are themselves not natural as we think of traditional grading. They can also be implemented in ways that are essentially irreversible, so archiving intermediate elements and using non-destructive corrections can save a lot of time and money if the product turns out to need upstream changes.

crc/30