This IS the Avid L2: Re: [Avid-L2] Lots of LUTs?

Friday, October 17, 2014

Re: [Avid-L2] Lots of LUTs?

It's important to consider what compression will be applied to the video
as a part of the decision. If you're working uncompressed or lightly
compressed, protecting the dark areas and highlights with something like
LogC is probably a great choice. If a lot of compression will be
applied to the recorded (pre-color correction) video, protecting those
blacks and whites might be a poor choice. Data compression schemes are
asked to do some pretty heavy lifting. Good compression schemes look
for ways to throw away bits that no one will notice. At some point, the
information in the shadows and in the highlights become a reasonable
target, allowing the compression to treat the mid (skin) tones a little
more delicately. By pushing the shadows up and the highlights down, we
are telling the compression scheme, "all this stuff is equally
important". It may all be important but if the compressors job is to
squish that full bandwidth HD video into 25 mbps, a lot of data has to
be tossed out. The compression will not be lossless. LogC may be
making your compression more lossy. --J.B.

Disclaimer: I'm far from an expert on data compression. I try to be an
informed consumer of the technology but I get lost pretty easily. I've
started reading Khalid Sayood's book on data compression several times
but I've never made it past the first few chapters. It would be great
reading for people with a stronger electrical engineering background
than mine.

Mark Spano cutandcover@gmail.com [Avid-L2] wrote:
> The way I look at it, which may or may not be accurate but works for
> me, is that you're not leaving 3/7th of the detail out. If that same
> shot was shot using the linear mode, the meat of the signal would get
> captured but the shadows and highlights would potentially be crushed.
> Let's say you're capturing a large dynamic shot. Something that if a
> sensor was unlimited, would land between -200 and 1000 mV. Shooting
> that scene linearly, you'd have to sacrifice top/bottom/or both. If
> the log profile is used, the signal is squeezed into the range the
> camera sensor can react to. So potentially you've got some bits from
> outside the linear range, though it's skewed. So you saved things that
> would have been crushed otherwise. It's not perfect, because in order
> to recover it, you've got to stretch that back out in a logarithmic
> manner - that's where the LUT comes in. The LUT maps log values along
> the curve to linear values, and can help recreate a more light
> accurate shot. But as you've seen, you don't *need* a LUT to recover a
> lot of what was captured, because it's relative signal math. Through
> normal color correction technique, you can transform the shot back
> into linear space. The LogC profile's main characteristics are the
> squeeze of luminance values with a squeeze of saturation/color values
> as well. To recover these without a LUT, you usually drop the Lift,
> raise the Gain, tweak the gamma, and increase the saturation. But
> these are not hard and fast rules, because every shot is different,
> and available light influences what was shot along with where the
> camera operator landed the meat of the log signal.
>
> After all that, your best bet is to use your scopes and use your eyes.
> Like it's always been. If you're usually hands on with color
> correcting source footage, gradually you'll start to be happier to see
> log shot footage roll in instead of linear, because it usually can end
> up looking better.
>
> _