GH4 V-LOG L noise reduction and color accuracy

So I just tested the HD 200mbps mode and V-LOG L is actually WORSE on that, even though you would think it's less compression and half the pixels to compress. Yikes.

That's because the 200 mb/s mode is an intra-frame codec. Each frame is a stand-alone frame. It doesn't use neighboring frames to interpolate an in-between frame like an inter-frame codec, such as the 100 mb/s mode. It's actually less efficient.

We had this issue in the GH3 days when the inter-frame codec was 50 mb/s and the intra-frame codec was 72 mb/s. Concensus was that although 72 mb/s is more than 50, it had worse image quality due to having to spread 72 mb across 24 individual frames than 50 mb over some subset of incomplete frames. Now with the gh4, the concensus seems to be that 200 mb/s is still not enough for quality intra-frame 1080p.

Also, not that I mean to correct you, but 1080p is actually 1/4th of the pixels of 4K. Amazing to think that 200 mb spread across 24 individual 1080p frames in a second still looks worse than only 100 mb over some subset of incomplete 4K frames.
 
That's because the 200 mb/s mode is an intra-frame codec. Each frame is a stand-alone frame. It doesn't use neighboring frames to interpolate an in-between frame like an inter-frame codec, such as the 100 mb/s mode. It's actually less efficient.

We had this issue in the GH3 days when the inter-frame codec was 50 mb/s and the intra-frame codec was 72 mb/s. Concensus was that although 72 mb/s is more than 50, it had worse image quality due to having to spread 72 mb across 24 individual frames than 50 mb over some subset of incomplete frames. Now with the gh4, the concensus seems to be that 200 mb/s is still not enough for quality intra-frame 1080p.

Also, not that I mean to correct you, but 1080p is actually 1/4th of the pixels of 4K. Amazing to think that 200 mb spread across 24 individual 1080p frames in a second still looks worse than only 100 mb over some subset of incomplete 4K frames.

200 mbits is more than enough for HD. Especially if it is 24p. ProRes422 HQ (the high quality version) only uses 176 mbits/s for 24p and that is 10bit and uses 4:2:2 color. 200 mbits is way more than enough for 24p HD on the GH4. Normal Prores 422 is only 117 mbits/s. An intra-format format is only bad if each frame is bit starved. 72 mbits was not enough bits and each frame suffered. Inter-frame encoding is considered about 3x or 4x as efficient as intra-frame encoding. So 72 mbits was like a normal H264 video at 18mbits while the 200 mbit version is more like a normal H264 video at 50 mbits. There is a huge quality difference between 18 mbits and 50 mbits when using inter-frame encoding.

This may have more to do with how the GH4 uses binning to convert the full sensor down to HD resolution and just a by product of starting with HD resolution instead of 4k resolution so the pixels and macro blocks are larger and easier to see.
 
How did of the beta "testers" not see this? It leapt out the very first test I shot. What a cock-up! In the name of compatibility with cameras that none of us use, Panasonic crippled the feature on the camera we're paying to upgrade.
 
How did of the beta "testers" not see this? It leapt out the very first test I shot. What a cock-up! In the name of compatibility with cameras that none of us use, Panasonic crippled the feature on the camera we're paying to upgrade.

How come none of you saw it in the footage posted by the beta testers?
 
Spawn EVEN graded.jpg
it's all about careful exposure and interpretation.

There's a weird issue with the luma data. When stretching or compressing them past a certain level, they separate at the chroma level and the blocking appears. If interpreted on the correct curve, this does not occur, and the blocking remains hidden.

There's an issue with blocking, yes, but, there was always an issue with blocking hiding in the color channels with panny cams, if you recall, and I somehow managed to circumvent that, and show peeps how to not destroy your footage during grading. I'm in the process of doing the same, here. I noticed a magenta and yellow blocking issue the first time I shot with a stock GH4 with Cine D, I had to adjust accordingly.

V-log with the internal codec is not the best option for landscape and nature shooters, it's not for the available-light-only crowd it's not a true 12 stop upgrade. So don't use it that way. On the GH2 we had 5 useable stops (+2/-2), The stock GH4; 6 stops (+2.3/-2.6) and with V-log it's 8 stops (+3/-4) If you are unwilling to accept these limitations, you'll need to spend more money on 10 bit recorders, or higher end cameras. But for those who have been working with 5 and 6 stops, 8 stops is a luxury.

verdict: So if V-log has these problems, who would want it? Why bother buying V-log at all? One word - falloff. The thing you pay $100 for is falloff. This upgrade is for filmmakers who understand and utilize lights, bounces, etc to carefully craft and expose shots who want that lovely falloff you get with the wider DR cams. Use it for that, and either switch back to another profile to shoot your sunsets, etc. - Or get an external recorder. What did you expect for $100? I feel i got my money's worth.

I said this would happen. I predicted the mass cries of "V-log sucks!" I love V-log. CG Pro users will come to love it too. I have 2 interpretation LUTs complete but they don't work in every situation. It will require probably 2 or 3 more to cover every scenario. Working with natural light - point-and-shoot-don't-think-about-it footage so far. Still trying to nail down a couple of LUTs that encompass a wider DR. It's possible, but takes time. On carefully exposed stuff, these LUTs are spectacular!

Updated flicker feed with tests using SOLID and LIMIT LUTs on both internal codec and (one of Driftwood's 10bit shots which is straight up SOLID interpretation LUT and nothing else). The denser grades have had fuji and kodak looks applied. The Less dense shots are just interpretation and white balancing.

https://www.flickr.com/gp/76276262@N04/3X2a0U
 
Last edited:
FCP X to Resolve is the easiest thing in the world. Export fcpxml, import into Resolve, grade, and export from Resolve. Also FCP-X is 32bit. But, without the Log conversion you're playing with fire. Maybe someday Apple will get their s*%t together, and have these options in the base package and not need a 3rd party LUT Utility

Lol I agree that having to use a 3rd party plugin to load a LUT being wild out of line.. But let me ask you, when round tripping a fcpx file to resolve for grading, at which point in my editing process should that happen? I've done this before with some success, but by the time I got back to FCPX after grading or somewhere in between there was a few audio sync problems. Should I lock my entire edit (with transitions and fx etc) first, THEN grade? Or should I just make all the cuts as they need to be, grade in resolve then finalize in fcpx? Also, have you experienced a difference in color when you grade a clip in resolve and then evaluate it in FCPX? I definitely see a difference in the richness of the colors/blacks. Thanks in advance!
 
Lol I agree that having to use a 3rd party plugin to load a LUT being wild out of line.. But let me ask you, when round tripping a fcpx file to resolve for grading, at which point in my editing process should that happen? I've done this before with some success, but by the time I got back to FCPX after grading or somewhere in between there was a few audio sync problems. Should I lock my entire edit (with transitions and fx etc) first, THEN grade? Or should I just make all the cuts as they need to be, grade in resolve then finalize in fcpx? Also, have you experienced a difference in color when you grade a clip in resolve and then evaluate it in FCPX? I definitely see a difference in the richness of the colors/blacks. Thanks in advance!


You should always lock before grading. i used to round trip, but I love the results coming out of Resolve so much, I've just started finishing there instead.
 
Last edited:
You should always lock before grading. i used to round trip, but I love the results coming out of Resolve so much, I've just started finishing there instead.

as expected. the question that remains for me is- how I would I get my transitions/effects that are exclusive to FCPX to translate into resolve when I'm exporting/importing from final cut to resolve?
 
as expected. the question that remains for me is- how I would I get my transitions/effects that are exclusive to FCPX to translate into resolve when I'm exporting/importing from final cut to resolve?

You have to render and export those separately, which is a pain. I've done round tripping and had few problems if it's just straight cuts, and dissolves, and then applied FCP FX at the end. You're just gonna have to play with it. There is no one right way to do things.
 
You have to render and export those separately, which is a pain. I've done round tripping and had few problems if it's just straight cuts, and dissolves, and then applied FCP FX at the end. You're just gonna have to play with it. There is no one right way to do things.

Understanding overstood!! Thanks a lot bro.. Link to your work?
 
And I just did this tonight, after a double of Johnny Walker. Details are in the Vimeo description along with a download link for some of the raw footage for y'all to play with.


 
Last edited:
I don't care what LUT you use, it's not going to be able to magically remove this nasty magenta/green encoding error. Look at this sample I just shot this morning and processed with Art Adam's LUT:

http://blog.josephmoore.name/wp-content/uploads/2015/09/roof.png

Some LUT's may exaggerate it more or less, but the error is baked into the compressed 8-bit image. What a huge disappointed. That $99 upgrade is going to cost me another $1,300 to be usable.
 
And I just did this tonight, after a double of Johnny Walker. Details are in the Vimeo description along with a download link for some of the raw footage for y'all to play with.


Well, its noisy as hell which leads me to think you used NR in post. But the magenta is still there. It's just hard to see because you have very little gradients to judge from. Go outside and film some solids (walls, streets, sidewalks, or such) and then apply your LUT.

Screen Shot 2015-09-19 at 11.02.38 AM.jpg

Sincerely,
Idiot at DVXuser
 
Well, its noisy as hell which leads me to think you used NR in post. But the magenta is still there. It's just hard to see because you have very little gradients to judge from. Go outside and film some solids (walls, streets, sidewalks, or such) and then apply your LUT.

View attachment 105304

Sincerely,
Idiot at DVXuser

see previous flicker link - already done that. No NR, never use it, never need to. You'll learn eventually; check my body of work, google me, whatever you need to do... Not since the GH2 got hacked have I ever used NR in camera or in post. If you use my methods there's never a need. if you don't you suffer from hi-end gear envy. And as I posted earlier, the blocking is always there in every profile you shoot, it's usually just hidden. (go look in the color channels) My interpretations keep it hidden even under duress, because with V-log trying to push the luma value (green) too far causes the color channels to separate revealing the blocking. It's not just about how I expose, but how I grade.
 
Last edited:
see previous flicker link - already done that. No NR, never use it, never need to. You'll learn eventually; check my body of work, google me, whatever you need to do... Not since the GH2 got hacked have I ever used NR in camera or in post. If you use my methods there's never a need. if you don't you suffer from hi-end gear envy. And as I posted earlier, the blocking is always there in every profile you shoot, it's usually just hidden. (go look in the color channels) My interpretations keep it hidden even under duress, because with V-log trying to push the luma value (green) too far causes the color channels to separate revealing the blocking. It's not just about how I expose, but how I grade.

It's nice that you are contributing over here as lots of us are looking for the potential in the VLog and Gh4. However, you as are I are nothing special other than contributors. I don't want to google you. I would like to see your "LUT" outside filming gradations as I described above. I'm not gonna nor want to get into a pissing match about body of work or IMDB self entries. So, to prove yourself as you claim, go outside and film a general scene with some gradient such as pavement, sidewalk, walls, sky, etc. Throw some skin in there if you would like. Because a toy on a desk just don't reveal what you claim to have been doing. And by film, I mean IN MOTION. Not some flicker grabs.

Sincerely,
Idiot at DVXuser
 
Last edited:
see previous flicker link - already done that. No NR, never use it, never need to. You'll learn eventually; check my body of work, google me, whatever you need to do... Not since the GH2 got hacked have I ever used NR in camera or in post. If you use my methods there's never a need. if you don't you suffer from hi-end gear envy. And as I posted earlier, the blocking is always there in every profile you shoot, it's usually just hidden. (go look in the color channels) My interpretations keep it hidden even under duress, because with V-log trying to push the luma value (green) too far causes the color channels to separate revealing the blocking. It's not just about how I expose, but how I grade.


You are very arrogant, therefore it makes me happy to say that the problem exists in the images you have in the flicker. For example the picture of the "white wall" with the Limit LUT applied is the worst offender.
Edit: Also check out the one with the sky. Never seen purple clouds before.
 
Last edited:
Back
Top