GH4 Saturate your profile

B

bunk

Member
This might be of interest to some of you.

YUV smearing kept nagging in the back of my head. Problem of all profiles …a limitation of the GH4 and it’s codec. Happening when an almost desaturated color crosses the U or V axis. Once it’s recorded you can not undo it (I have yet to see an example it can be undone). So best is to find a way to prevent it, before it happens. Those almost greyish colors should not hop the axis.
Solution …Saturate them!

In the test I did it works, with both CineD and Natural in 1080 All-Intra. Worst codec.
‘Works’ is maybe a too big a word, but the smearing is refined, the islands are broken up into smaller parts giving it a better distribution.
Attached my Natural test. Settings : Hue-2 vs Hue +5
I eyeballed the desaturation of the saturated clip and just by the look of it it doesn’t look like there are any disadvantages.

If it works in 1080 All-I it will work in UHD ProRes 422, just not sure how effective that would be …and alas the profile that could use this workaround the most, can’t use it as we have not acces to the Hue settings …bummer. But it might help the attempts to recreate a V-log version with CineD

In the linked movie second half is the saturated part.
Nat_hue_test.jpg



[video]http://www.bunkweb.nl/filmpie/Hue_CineD.mov[/video]
 

Attachments

  • Nat_hue_test.jpg
    Nat_hue_test.jpg
    34.7 KB · Views: 0
Last edited:
mash_gh4 said:
attention: by increasing the saturation your color channels will clip sooner! :(
Click to expand...
It's evening here so I tried with the help of a light. No luck.
blownout.jpg


blownout_hue.jpg

I see no difference except for the better distribution!

blownout_wfm.jpg


Could you show an example? Thanks!
 
didn't use my camera, but a simple simulation in a compositing application should illustrate it:

a.) a simple line from a colorchecker test chart and its waveform using normal saturation

saturation1.png


b.) the same frame with extrem decreased saturation. the difference of the color channels is fading away...

saturation2.png


c) extreme oversaturation -- the increased differences of latitude in the color channels -- the red and blue touch the limits and get partially clipped

saturation3.png


i did only change the saturation, nothing else!

the real world situaition is little bit more complex, because there are not only RGB channels used, but the issue is still the same. :(
therefore i would second joe12souths advice.
 
Last edited:
mash_gh4 said:
i did only change the saturation, nothing else!

the real world situaition is little bit more complex, because there are not only RGB channels used, but the issue is still the same. :(
therefore i would second joe12souths advice.
Click to expand...
Thanks for the example. If you compensate by lowering Contrast it still seems to work ...and then there is always the vectorscope to check. Haven't tested it out in daylight yet as I only thought of it this evening. I think it deserves a change.
 
Bunk, I think that you've shown that your method is sound and the results speak for themselves, especially when looking at the waveforms or those over-saturation comparisons.

But the brick walls of clipping are also boundaries we really don't get to cross.

Your method will work just as long as you don't approach clipping - and that's going to be a lot harder to control in daylight.
Perhaps you could use it safely if you regard it as your "indoors" profile.
 
bunk said:
Could you elaborate a bit, thanks!
Click to expand...
In order to avoid clipping one or more channels you would have to decrease contrast, which introduces a different set of problems.

Some people love to shove everything to -5, you're suggesting pushing towards +5. The reality is that a low DR 8-bit image involves a lot of compromising, and that is almost always best accomplished near the middle. You push to the extremes, either direction and you compromise something.
 
You must log in or register to reply here.
Back
Top