massive shift in green color on BMPCC4K

The place for questions about shooting with Blackmagic Cameras.
  • Author
  • Message
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostThu May 21, 2020 9:55 pm

It's disheartening that some people here think that this is not a big deal. For me, this was the reason I stopped using my BMPCC 4k. When I go out and shoot at a park, all the greenery there becomes yellow-ish. To me, this is very bothersome, because I'm after the film look, which usually renders neutral colors (which is how visual artists usually do paintings too: neutral colors, not high key). So with all the trees suddenly turning yellow-ish (as if someone turned them into hue vs luma), and that not being fixable by grading (you can only shift the cast to another color, not remove it), it's a problem. If I turn the yellows into greens using hue vs hue, this also changes my skin tones, because that color range there is all mashed up already by the sensor.

Look at this video at this exact 00:55 time. You will notice the yellowness I'm talking about. And while the S1H uses also a Sony sensor, it's using a newer, actual dSLR sensor, and not a night surveillance sensor (that the BMPCC 4k uses) that happens to be good-enough to be soldiered into cameras.

Heck, in fact I'd say that even the S1H doesn't have the problem fixed, because if you see at 2:48, the S1H SHIFTS the problem (the skintones become purple, as in the Fuji cameras, that ALSO use Sony sensors), but fixes the greens. So basically, IMHO, it seems that if you're using a Sony sensor, you are going to get screwed color-wise in one way or another. This is something that BM should comment on.

Personally, I found refuge at Canon cameras, which suck video-wise otherwise, but they do give me the look I want. I've bought Sony, Fuji, Panasonic and BMD cameras, that all use Sony sensors. These cameras are all gathering dust on my shelf. The only looks I enjoy are the Canon ones, and the original BMPCC (Fairchild sensor).

Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline

John Paines

  • Posts: 3561
  • Joined: Tue Jul 28, 2015 4:04 pm

Re: massive shift in green color on BMPCC4K

PostThu May 21, 2020 10:50 pm

Oh, man ... not jsfilmz again. This is the same guy who got Hong Kong Jack(?) in trouble. With all due respect for the point of view -- I don't see these shifts, but you never know -- there's gotta be a better way than relying on somebody who doesn't know what he's doing.

The OP provided one example, but I can't replicate it among shades of green. Or it's well within the normal range of what you'd expect, for color correction.

As for Canon, that has to be among the least accurate camera lines out there, designed as it is to reproduce pleasing skin tones, sometimes at the expense of fidelity elsewhere.
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostThu May 21, 2020 10:54 pm

He's not doing anything weird in that video, he simply uses the rec709 as given by the manufacturer.

Also, this was just an example. Please don't try to push down the main argument by attacking the guy. The problem is VISIBLE here, on my OWN footage. I don't have to rely on HIS. I simply include his as a reference. The problem IS real.

As for Canon, that has to be among the least accurate camera lines out there, designed as it is to reproduce pleasing skin tones, sometimes at the expense of fidelity elsewhere.


Canon made their sensors originally to emulate Kodak film. So there are probably some inaccuracies. But that's what makes film look like film. I personally want nothing to do with the video look. Canon can be EASILY manipulated to look like film. Sony sensors can not.

There's a reason why so few movies are shot with the Ursa btw, despite being quite cheap for what it does. It just doesn't look like film. Indies prefer to save and rent an Alexa or a RED, than having to deal with an Ursa.
Last edited by Eugenia Loli on Thu May 21, 2020 11:24 pm, edited 2 times in total.
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline

John Paines

  • Posts: 3561
  • Joined: Tue Jul 28, 2015 4:04 pm

Re: massive shift in green color on BMPCC4K

PostThu May 21, 2020 11:04 pm

Eugenia, it feels a bit like what we went through about a week ago, where no credible examples were ever produced for the claims being made on cDNG v braw.

I could post material that doesn't demonstrate the issue at hand, but that doesn't prove it can't happen. If you have demonstrations of the matter (and original frames), minds are open.

As for Canon, and the other generalizations, that's a long discussion.
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostThu May 21, 2020 11:14 pm

The video I posted IS how my greenery footage looks too. Why do you want me to go out right now and shoot side by side BMPCC vs Canon M50? Just so I can convince you? I don't have to do that to prove anything. I'm telling you though, that what is seen in the video above, IT IS what I witnessed here too. My BMPCC 4k footage outdoors while under the sun LOOKS YELLOW in the greenery, while Canon cameras look more green! And I want the yellow and green colors DECOUPLED (because the sensor is delivering a SOUP instead).

As I wrote above, the no1 culprit for that, is probably the CFA filters in front of the sensor. Not the sensor itself exactly, and not the color science of each camera manufacturer. But if Sony can't fix that (or if they're using cheaper filters on their cheaper sensors), that's a problem for all these camera manufacturers. And I've identified problems with all of them: Panasonic: magenta people, Fuji: purple people, Sony: brown galore, BMPCC: yellowness galore. Canon: CLEAN, neutral colors. Alexa: Clean, neutral colors. And that's that.
Last edited by Eugenia Loli on Thu May 21, 2020 11:17 pm, edited 2 times in total.
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline

lee4ever

  • Posts: 310
  • Joined: Tue Jan 08, 2019 9:59 pm
  • Real Name: Aki Lee

Re: massive shift in green color on BMPCC4K

PostThu May 21, 2020 11:14 pm

Eugenia Loli, I fully agree with them. This was apparently the reason why I didn't like the BMPCC4K and I returned it at that time.
Offline

John Paines

  • Posts: 3561
  • Joined: Tue Jul 28, 2015 4:04 pm

Re: massive shift in green color on BMPCC4K

PostThu May 21, 2020 11:23 pm

Eugenia, if you're personally satisfied in the matter, then I guess that's the end of it. No law specifies you have to use or like a BMPCC 4K. But saying that your footage looks like a youtubber's , where even basic grading competence can't be assumed among users of these cameras, won't go very far in convincing anyone else.

If we're going to have a reasonable discussion, as opposed to just announcements, there need to be proofs -- original frames people can examine.
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostThu May 21, 2020 11:30 pm

Well, if you had bothered to actually look at the video's description, JZSFilm has already uploaded downloadable videos right out of the camera: https://drive.google.com/file/d/1yA78s5 ... 5ehlc/view
Feel free to download 22 GBs and compare the BMPCC 4k to the S1H. And no, I don't want to see extreme grading to fix the greens. I want it to look proper right out of the camera (with the extended video lut). You see, if you try to use hue vs hue to "fix" the greens, you WILL be changing other colors that are NOT in the frame, so we can't evaluate what kind of shift your grading generated.
Last edited by Eugenia Loli on Thu May 21, 2020 11:39 pm, edited 2 times in total.
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline

John Paines

  • Posts: 3561
  • Joined: Tue Jul 28, 2015 4:04 pm

Re: massive shift in green color on BMPCC4K

PostThu May 21, 2020 11:38 pm

I did note the 22 gigs for download, and declined to proceed. jsfilmz is not a credible source. It *might* not be a waste of time, but I don't have a few hours to find out.

Again, I'd insist that the burden is on anyone making the case -- if the idea is to have an actual discussion. Otherwise it's just, he says no, she says yes.
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostThu May 21, 2020 11:40 pm

I have personally deleted most of my BMPCC 4k footage, so I don't have it anymore. Since I bought the camera last year, I used it for a couple of months only, didn't like at all of what I saw with the color cast, deleted the SD card and now use it elsewhere, and the BMPCC 4k now is gathering dust. That's the full story. There are videos on the link above for you to play with, they look EXACTLY as my own footage looked like: freaking yellow (with the right white balance, yes). No need to do double the work. That's what I experienced, what you see in these files above. Download the files above, it doesn't take a genius to shoot correctly at iso 400 and at 5600k. You might not like the guy, but his shots are correctly shot. You don't have to take him down for everything if you don't agree on the cdng vs braw issue. He did shoot correctly those shots.
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline

John Paines

  • Posts: 3561
  • Joined: Tue Jul 28, 2015 4:04 pm

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 12:07 am

Whatever this guy came up with, I can't concede your point third hand-- I do have the camera, I don't see "massive" yellow shifts. And I looked for shades of green very like that packaging, based on product displays, pictured above. And no go -- green was green, or the divergence was quite manageable.

Just as a side note, here's "green" from a 35mm movie. How "realistic" or accurate is it? Natural green, meanwhile, is full of yellow, quite unlike this:

green.jpg
green.jpg (272.23 KiB) Viewed 471 times
Offline
User avatar

Dmitry Shijan

  • Posts: 1356
  • Joined: Wed Sep 17, 2014 7:15 pm
  • Location: UA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 1:53 am

Eugenia Loli wrote:My BMPCC 4k footage outdoors while under the sun LOOKS YELLOW in the greenery, while Canon cameras look more green!

Because Pocket 4K sensor produce IR pollution and shifts greens to brown. You need to use IR filter when shoot outdoors. Pocket 6K use stronger IR filter and looks way greener.
Also it may be because because RAW prosessed in YRGB color managed project instead of basic YRGB project. (See my examples in earlier posts here).
YRGB color managed (same as as CST node) apply simple matrix color space input profile to RAW source. This look do not match to ProRes processed in camera.
But YRGB non color managed project setting apply more complicated LUT based input profile to RAW source. Sort of sensor factory calibration. This look match to ProRes processed in camera.
Additional Color Checker correction makes greens even more greener and other colors also look richer.
try this example project viewtopic.php?f=21&t=65149#p537852

Image
Image
Image
All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 2:05 am

I don't use the color managed option.
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline
User avatar

Dmitry Shijan

  • Posts: 1356
  • Joined: Wed Sep 17, 2014 7:15 pm
  • Location: UA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 2:09 am

IR pollution may depend of year season. Closer to equator - stronger IR i guess.
Also ND filters always produce IR pollution.
All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 2:55 am

Not the case. I always shoot outdoors with an IR CUT filter. The yellowness persistent.
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline
User avatar

Robert Niessner

  • Posts: 2951
  • Joined: Thu Feb 21, 2013 9:51 am
  • Location: Graz, Austria

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 5:23 am

Eugenia Loli wrote:But that's what makes film look like film. I personally want nothing to do with the video look. Canon can be EASILY manipulated to look like film. Sony sensors can not.

There's a reason why so few movies are shot with the Ursa btw, despite being quite cheap for what it does. It just doesn't look like film. Indies prefer to save and rent an Alexa or a RED, than having to deal with an Ursa.


Well, if you want to make a good argument, then this is a weak one. According to your thesis it is a problem with the Sony sensor and its color filters, making the footage look like video and that's the reason why indies don't shoot with the Ursa.
The hole in your argument: the URSA series don't use Sony sensors.

I have no idea how many indie filmmakers and their motivation for camera choice you know in reality, but your generalization seems a bit far fetched.
Robert Niessner
LAUFBILDkommission
Graz / Austria
---------------------------------------------------------------------
Blackmagic Camera Blog (German):
http://laufbildkommission.wordpress.com

Read the blog in English via Google Translate:
http://tinyurl.com/pjf6a3m
Offline
User avatar

Dmitry Shijan

  • Posts: 1356
  • Joined: Wed Sep 17, 2014 7:15 pm
  • Location: UA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 5:50 am

I posted this test before, but can't find original thread now, but formally with 12 bit RAW source with ColorChecker correction you can recover colors even from almost 100% desaturated image:
Here is original:
Image

Here is example when :arrow: Saturation was down to 0.1 and next ColorChecker correction node was applied:
Image

Same same attempt to recover colors from ProRes 444 and 422 sources:
Image
Image
All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/
Offline

John Griffin

  • Posts: 576
  • Joined: Sun Aug 23, 2015 4:18 pm

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 5:52 am

Eugenia Loli wrote:I don't want to see extreme grading to fix the greens. I want it to look proper right out of the camera (with the extended video lut). You see, if you try to use hue vs hue to "fix" the greens, you WILL be changing other colors that are NOT in the frame, so we can't evaluate what kind of shift your grading generated.
tweaking a hue shift is not ‘extreme grading’ and it does not change other colour relationships unless you move the whole line on the curve. The whole idea of Hue vs Hue is that it can be controlled very precisely to only change specific colour ranges.
Which is what I said, and even though this is what I was talking about, there is not too much comparison between your actions at the moment compared to some not around here.
Offline

Dune00z

  • Posts: 319
  • Joined: Thu Oct 18, 2018 3:40 pm
  • Real Name: Duane Eues

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 6:03 am

This thread is clearly devolving into old and tired topics.

The Crown, gorgeous cinematography, is shot on a Sony f55. It looks incredible.

I've seen footage that looks identical to 35mm film and Arri from Sony cameras graded by artists.

If you can't get good looking results from cameras shooting decent codec 10bit/12bit log or raw it's more than likely on the user which would be a lack of skill (which is nothing to be ashamed of), and not the camera's fault.
Online

Denny Smith

  • Posts: 12349
  • Joined: Thu Aug 01, 2013 4:19 pm
  • Location: USA, Northern Calif.

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 6:13 am

OK, I retract my previous statement.
Denny Smith
SHA Productions
Offline
User avatar

dondidnod

  • Posts: 228
  • Joined: Sun Nov 25, 2018 7:52 am
  • Location: Castro Valley, CA
  • Real Name: Donald Keller

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 6:28 am

I have a BMPCC 4K and I am more interested in Narrative films and prefer to grade for pleasing skintones. I don't mind sacrificing some green accuracy to get that. You seldom see green in (human) faces.

Here is an indoor example with light from Andoer CFL 5500K softboxes (CRI 80) with a Voightlander 17.5mm F/0.95 MFT lens, at F5 with a Hoya IR cut filter, in Braw 3.1. In Premiere Pro I graded with Cinespace 25, Whitebalance temp -9.5, tint 16.8, exposure -2.4 and Saturation 160.6.
YelExampleV17C4KB31F5IRTk3_C052exp24A.jpg
Color corrected for flesh tones
YelExampleV17C4KB31F5IRTk3_C052exp24A.jpg (641.03 KiB) Viewed 414 times

Here is that shot with the Whitebalance temperature at -42.5. The yellow is more controlled, but still present in the green, and the flesh tone suffers with a more pale complexion. The white chess pieces benefit however as they are less tan and slightly bluish.
YelExampleV17C4KB31F5IRTk3_C052exp24T42A.jpg
Adjusted for less yellow and better green
YelExampleV17C4KB31F5IRTk3_C052exp24T42A.jpg (651.19 KiB) Viewed 414 times

Here they are together with the corrected (yellowish) image on the top and the Whitebalance temperature at -42 to reduce the yellow on the bottom. Scroll down to compare the two.
YelV17C4KB31F5IRTk3_C052exp24Top&T42Btm.png
Corrected for flesh tone - Top & for green - Bottom
YelV17C4KB31F5IRTk3_C052exp24Top&T42Btm.png (801.88 KiB) Viewed 414 times
Offline
User avatar

dondidnod

  • Posts: 228
  • Joined: Sun Nov 25, 2018 7:52 am
  • Location: Castro Valley, CA
  • Real Name: Donald Keller

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 6:36 am

Here is the finished product:
A010_05111044_C052LtA.jpg
Voightlander 17.5mm BMPCC 4K
A010_05111044_C052LtA.jpg (848.7 KiB) Viewed 410 times

Join me to view the final film projects from Caroline Blair's Motion Picture Directing Class at City College of San Francisco on a YouTube live stream, Friday, May 22, 2020 from 7:00 PM to 8:30 PM PST.

This will include the premiere of my new film, The Plague Dilemma:

A Knight returns home from the Crusades in the Holy Land in 1349 to find that the Black Plague is killing thousands daily. He has a crisis of faith, and questions Death as to how a loving, all-powerful God could allow this tragedy. He challenges Death to a chess match to gain some time to seek knowledge. Inspired by Ingmar Bergman's Swedish language 1957 film, THE SEVENTH SEAL.
Produced, Directed, filmed and acted by Don Keller alone in his home studio.

I did the camera work for a scene from Blade Runner 2049 using a Panasonic S1H and Leica lenses in another piece in the show.

Link to the show:
https://bit.ly/cine131live
Offline
User avatar

Dmitry Shijan

  • Posts: 1356
  • Joined: Wed Sep 17, 2014 7:15 pm
  • Location: UA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 6:55 am

Another problem with colors may be due grading in tiny Rec709 color space.

Here is technically bad workflow example, when color correction done in tiny Rec709 color space and gamma:
BMDfilm source -> Transform to Rec709 -> ColorChecker correction, Noise Reduction, Exposure/Gain adjustment, Film emulation LUT

Here is good workflow example, when color correction done in wide gamut color space and log gamma:
BMDfilm source -> Transform to Wide Gamut like ArriWideGamut/LogC or RedWideGamut/RedLog3G10 -> ColorChecker correction, Noise Reduction, Exposure/Gain adjustment -> Transform to Rec709 -> Film emulation LUT
All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/
Offline

John Griffin

  • Posts: 576
  • Joined: Sun Aug 23, 2015 4:18 pm

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 7:27 am

Dmitry Shijan wrote:Another problem with colors may be due grading in tiny Rec709 color space.

Here is technically bad workflow example, when color correction done in tiny Rec709 color space and gamma:
BMDfilm source -> Transform to Rec709 -> ColorChecker correction, Noise Reduction, Exposure/Gain adjustment, Film emulation LUT

Here is good workflow example, when color correction done in wide gamut color space and log gamma:
BMDfilm source -> Transform to Wide Gamut like ArriWideGamut/LogC or RedWideGamut/RedLog3G10 -> ColorChecker correction, Noise Reduction, Exposure/Gain adjustment -> Transform to Rec709 -> Film emulation LUT

This shouldn’t be an issue in the 32bit float working space unless you are applying LUT’s. Even in a REC709 timeline the full colour and tonal data from the original source clip is available.
Which is what I said, and even though this is what I was talking about, there is not too much comparison between your actions at the moment compared to some not around here.
Offline
User avatar

Dmitry Shijan

  • Posts: 1356
  • Joined: Wed Sep 17, 2014 7:15 pm
  • Location: UA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 7:57 am

Yes, if you transform with CST node, all original color data is always there (until you use Luma/Saturation mapping option), but colors are still limited and arranged within tiny Rec709 color space coordinates. Just test by yourself - apply ColorChecker correction inside Rec709 and compare it to ColorChecker correction done inside RedWideGamut color space/Log gamma. The result will be very different. In Rec709 color space Red colors always will shift to more orange side because not enough space.
Another great example - upscaling Pocket4K BRAW BMDfilm v4 with clipped red channel to RedWideGamut color space fixes ugly red channel clipping.
Image
Image
All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 11:14 am

Time is 4 AM here, and I spent all night testing stuff with my ex-BeOS/Google/Microsoft software architect husband.

He found that the problem is indeed the tiny gamut in rec709, as Dmitry Shijan mentioned. While the colors are available to the MATH of Resolve, they are not available to my screen and to the color grading chain under rec709 2.4. I was able to get the exact colors I wanted by using Rec.2020 in the Color Managed YRGB, and by transforming to Arri Alexa's colorspace (and at the very end, back to rec709). My monitor is not a pro video monitor, but at rec2020 I was able to see colors more accurately.

The reality was not that the sensor is not good, as I was claiming, but because it was TOO GOOD. You see, when you capture RAW, you are able to capture some colors that are further away from what rec709 can see. So some colors CLIP in the color grading chain. The greens become yellow, and the blues become cyan! And the reds clip in some red led lights (as you already know from night footages). But the information is actually still there, just not viewable without trickery.

To avoid that, you need to lower contrast MORE than BDMFilm provides, to SOFTEN the image, and to REMOVE saturation. Essentially, we have to EMULATE a LOG h.264 encoding, which removes information compared to RAW! The better the h.264 encoding is (as found in Sony/Panasonic/Fuji sony-sensor-based cameras) the worse the clipping will be outside of SRGB's boundaries! Canon's encoder sucks, which is why it FITS in rec709 without much clipping, giving us "pleasing" colors. I know, it's counter intuitive!

By switching to Rec2020, and going back and forth in the CIE Chromaticity panel to make sure that I'm not clipping much while in SRGB either, I was able to get exactly the colors I wanted. Here's Before (Extended video lut) and After black magic trickery. I also tested that on skin tones and other types of footage from the BMPCC, and there is NO COLOR CAST anymore using my new method!

There is perfect separation of color now, just like as found in Alexa and film!

Image

Image
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline

John Griffin

  • Posts: 576
  • Joined: Sun Aug 23, 2015 4:18 pm

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 11:41 am

Colour space remapping is tricky stuff. Colours can be clipped or compressed (and compressed in different ways) depending on the settings. Both have advantages and disadvantages. As a rule if you don't have a calibrated monitor able to display colours outside sRGB / REC709 then you shouldn't be working in a wider colourspace like REC2020 as even if it looks right on the screen it won't be right at output.
Which is what I said, and even though this is what I was talking about, there is not too much comparison between your actions at the moment compared to some not around here.
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 11:48 am

Well, it's that, or not using the camera at all at this point. I did an export, I didn't see a problem (on a TV).
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline
User avatar

Dmitry Shijan

  • Posts: 1356
  • Joined: Wed Sep 17, 2014 7:15 pm
  • Location: UA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 11:52 am

Eugenia, not sure if i understand all your workflow description but Rec2020 is not a best option as well. It was designed as delivery color space but not editing color space. It also produce some strange color shifts to cyan-green during color correction. I done a lot of tests between Arri Wide gamut, Panasonic V-log, BMDfilm, Rec2020, ACES and RedWideGamut, and end up with RedWideGamut color space as a best universal option.
Try to download my example project shared here viewtopic.php?f=21&t=65149#p537852 ,change CST node gamma/color space input settings to Pocket4K.v4 instead of BMDfilm.v1 gamma/color and see how it looks.

By the way, i didn't test yet if there is a difference in BRAW between YRGB and YRGB color managed project look. BRAW is very different from DNG, it always pre-transforms colors under the hood from original sensor color space to BMDfilm Wide Gamut v4 color space.
Last edited by Dmitry Shijan on Fri May 22, 2020 12:01 pm, edited 1 time in total.
All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 12:01 pm

Unfortunately, there's no way around it. I've tried everything else. The only thing that works is to fit the wide color space of the BMD cameras in rec2020. If we don't do that, there's clipping, which manifests as color shift in rec709 timelines.
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline

John Paines

  • Posts: 3561
  • Joined: Tue Jul 28, 2015 4:04 pm

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 12:03 pm

Eugenia Loli wrote:Unfortunately, there's no way around it. I've tried everything else. The only thing that works is to fit the wide color space of the BMD cameras in rec2020. If we don't do that, there's clipping, which manifests as color shift in rec709 timelines.


How about this? Link to that original frame, which I gather is from the youtubber and his 22 gigs, and your correction in some uncompressed format.

And we'll how much or little it takes, inside a rec. 709 workflow, to get your result.
Offline
User avatar

Dmitry Shijan

  • Posts: 1356
  • Joined: Wed Sep 17, 2014 7:15 pm
  • Location: UA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 12:04 pm

What clipping are you talking about? Can you share BRAW frame?
BMDfilm source -> CST node Transform to Wide Gamut RedWideGamut/RedLog3G10 -> ColorChecker correction, Noise Reduction, Exposure/Gain/Contrast adjustment -> CST node Transform to Rec709 turn Luma/Saturation mapping on -> Film emulation LUT (optional)
All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 12:22 pm

>What clipping are you talking about?

The clipping outside the rec709 boundaries! If you add a rec709 BMD lut on a normal colorspace rec709, and you look at the graph, you will see the right side of it, clips with a hard line! It doesn't fit inside. When that happens, the system pushes the colors closer together, with greens becoming more yellow. That's the "cast" we were seeing.

>Can you share BRAW frame?

It's linked from the beginning of this page. There's no color checker frame there, it's just a tree.

Bah. It's 5:20 AM here. I'm going to bed. Do whatever you want. I've found my solution.
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline
User avatar

Dmitry Shijan

  • Posts: 1356
  • Joined: Wed Sep 17, 2014 7:15 pm
  • Location: UA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 12:36 pm

John Griffin wrote:As a rule if you don't have a calibrated monitor able to display colours outside sRGB / REC709 then you shouldn't be working in a wider colourspace like REC2020 as even if it looks right on the screen it won't be right at output.

Seems you don't get the concept.
If final delivery is Rec709 color space, your CST node or LUT in the end of the chain is always adjusted to transform wide gamut/gamma to Rec709 gamut/gamma. But same time you manipulate with colors and contrast in wide color space and log gamma because there is a lot of free space around and so colors don't produce clipping.

Colors are manipulated inside Large color space, but you see final result in Rec709 in realtime. This is not something special to Resolve. Protoshop or any other color managed image editor works like this. ACES also works like this.
All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/
Offline

John Griffin

  • Posts: 576
  • Joined: Sun Aug 23, 2015 4:18 pm

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 12:37 pm

In a properly setup colour managed workflow there should be no clipping of out of gamut colours when transforming a larger space to a smaller one unless this is a specific setup the user wants. Look at the gamut mapping options in either the color space transform node the colour management setup.
Which is what I said, and even though this is what I was talking about, there is not too much comparison between your actions at the moment compared to some not around here.
Offline

John Griffin

  • Posts: 576
  • Joined: Sun Aug 23, 2015 4:18 pm

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 12:40 pm

Dmitry Shijan wrote:
John Griffin wrote:As a rule if you don't have a calibrated monitor able to display colours outside sRGB / REC709 then you shouldn't be working in a wider colourspace like REC2020 as even if it looks right on the screen it won't be right at output.

Seems you don't get the concept.
If final delivery is Rec709 color space, your CST node or LUT in the end of the chain is always adjusted to transform wide gamut/gamma to Rec709 gamut/gamma. But same time you manipulate with colors and contrast in wide color space and log gamma because there is a lot of free space around and so colors don't produce clipping.

Colors are manipulated inside Large color space, but you see final result in Rec709 in realtime. This is not something special to Resolve. Protoshop or any other color managed image editor works like this. ACES also works like this.

I do get the concept, I have been working in colour managed workflows since Photoshop 5. I remember when it first came in and what a huge change it was.
Which is what I said, and even though this is what I was talking about, there is not too much comparison between your actions at the moment compared to some not around here.
Offline
User avatar

Dmitry Shijan

  • Posts: 1356
  • Joined: Wed Sep 17, 2014 7:15 pm
  • Location: UA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 12:48 pm

It's linked from the beginning of this page. There's no color checker frame there, it's just a tree.
Ok, got it. So example with tree and lake is from jsfilmz video. That dude post really ugly graded tests and seems has no idea how deal with colors and color transforms. So i can not recommend to rely on his YouTube uploaded tests :)
But from the other side it is nice that he shares a source files from different cameras, so people can download them and do their own grading and see real difference between cameras.
All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/
Offline
User avatar

Dmitry Shijan

  • Posts: 1356
  • Joined: Wed Sep 17, 2014 7:15 pm
  • Location: UA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 2:57 pm

Yep, Pocket4K if very brownish outdoors compare to Panasonic examples. I guess this is due IR pollution mixed with BM Color Science. It is impossible to correct with simple WB/tint adjustment. IR filter+ColorChecker correction should fix this. This is not a surprise because same IR pollution effect exists in all BM camera generations except Pocker6K.
For Panasonic i set input node to v-log/v-gamut, but i am not really sure if this exact match to S1H input specs. Is it universal for all Panasonic cameras or each model require some specific input LUT?
Image
Image

Indoor shoots without IR pollution looks way better and very easy to match with simple WB/tint adjustment even with randomly blinking LED lights.

Image
Image
All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/
Offline
User avatar

rick.lang

  • Posts: 13391
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 3:25 pm

I hope Eugenia is sleeping well.

Another fascinating thread on colour and we’re not even in the Resolve forum. Appreciate everyone’s contributions.
Rick Lang
Offline

John Paines

  • Posts: 3561
  • Joined: Tue Jul 28, 2015 4:04 pm

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 3:32 pm

Dmitry, can you link to the two versions of the lake shot (the originals)?

I believe Panny said somewhere or other that the Varicam LUTs are suitable for the SH1 -- it's effectively the same color science as is present on the EVA1, the cameras match. But can't tell you where I saw it.
Offline

Howard Roll

  • Posts: 673
  • Joined: Fri Jun 03, 2016 7:50 am

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 4:15 pm

Dmitry Shijan wrote:Yes, if you transform with CST node, all original color data is always there (until you use Luma/Saturation mapping option), but colors are still limited and arranged within tiny Rec709 color space coordinates. Just test by yourself - apply ColorChecker correction inside Rec709 and compare it to ColorChecker correction done inside RedWideGamut color space/Log gamma. The result will be very different. In Rec709 color space Red colors always will shift to more orange side because not enough space.
Another great example - upscaling Pocket4K BRAW BMDfilm v4 with clipped red channel to RedWideGamut color space fixes ugly red channel clipping.
Image
Image

I don’t know that the accuracy of the color checker tool within different timeline gammas is sufficient to make this point. Your workflow let’s you choose where in the color pipeline the noise reduction happens to get better results, smart. The chart match function is a magic bullet, there is way too much happening under the hood to use this feature as any sort of metric.

Good Luck
Offline
User avatar

Dmitry Shijan

  • Posts: 1356
  • Joined: Wed Sep 17, 2014 7:15 pm
  • Location: UA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 4:58 pm

John Paines wrote:Dmitry, can you link to the two versions of the lake shot (the originals)?

I believe Panny said somewhere or other that the Varicam LUTs are suitable for the SH1 -- it's effectively the same color science as is present on the EVA1, the cameras match. But can't tell you where I saw it.


It is from JSFILMZ youtube test video. He put links to original files in video description or in comments under the video
Here is link to that P4K vs S1H tests (22GB) https://drive.google.com/file/d/1yA78s5 ... 5ehlc/view

By the way, i remember his BMMCC vs Pocket4K test at same location with lake and tree, and BMMCC produced more IR pollution that Pocket4K (I compared original sources, but not his youtube poor graded example) Too many IR pollution could be due water reflecting double amount of IR light. Also both cameras use different lenses, so it could be due some specific lens tint mixed with IR.
I done a lot of summer days shoots with BMMCC before i got OLPF/IR filter and never ever see similar extreme brown grass and trees instead of fresh green. It was some shift in green colors but it was never so huge.
All my custom made accessories for BMMCC/BMMSC now available here https://lavky.com/radioproektor/
Offline

John Brawley

  • Posts: 2360
  • Joined: Tue Aug 21, 2012 7:57 am
  • Location: London UK

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 5:36 pm

Eugenia Loli wrote:And while the S1H uses also a Sony sensor, it's using a newer, actual dSLR sensor, and not a night surveillance sensor (that the BMPCC 4k uses) that happens to be good-enough to be soldiered into cameras.


Seems a bit of an unsubstantiated /rumour / hearsay claim to me, especially now that I think you've acknowledged your workflow was the issue in your OP ?

JB
John Brawley
Cinematographer
London UK
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 6:24 pm

The BMPCC 4k does use a night surveillance sensor, nothing hearsay about that part. And "my workflow" was not "wrong" being the default workflow Resolve employs for their cameras. The way things are done on Resolve must be different with RAW cameras, than what has been suggested so far. Everyone is doing the "wrong" workflow with BRAW by using Rec709 2.4 (which is the default on Resolve).

Also, Dmitry, this is not a problem of IR pollution, since I did provide a picture of the tree without the problem, while you could not fix it under the normal Resolve workflow! How do you explain that?

Rec2020 might be a delivery standard, but so is rec709. I didn't find it to be an issue. For the last time, RAW footage of Blackmagic cameras don't fit in rec709 for processing the way normal 8bit or 10 bit LOG or rec709 footage does. The more saturation/contrast and detail exists, the more its "color bits" spill outside of the visible spectrum or the rec709 spectrum. Raw footage *begs* for big color spaces! And if your argument is that it "should have" fit, well, it doesn't fit gracefully! It clips on the sides of the rec709 triangle. The more it clips, the more colors are mashed together internally when the footage is interpreted. The more I manipulate the footage to stay inside the triangle, the more I can get rid of the generated color cast (not inherent!).

Currently, there are a few monitors that can display more colors, at around $1300-$2000, but they're not doing that much better than the monitor I already have. The only monitor that probably can do much better, and let you work in rec2020 without an issue, is the Apple one for $6k. Basically: we got a RAW camera, but tech hasn't moved fast enough to provide us with the rest of the hardware processing chain required to use that RAW camera to its full capability. So in order to go around these problems, we have to "shrink" the RAW footage to something akin to 10bit h.264 log (where encoding has removed information, rather than trying to mash it together), in order for it to behave in our hardware.
Last edited by Eugenia Loli on Fri May 22, 2020 6:45 pm, edited 1 time in total.
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline

John Griffin

  • Posts: 576
  • Joined: Sun Aug 23, 2015 4:18 pm

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 6:45 pm

Eugenia Loli wrote:The BMPCC 4k does use a night surveillance sensor, nothing hearsay about that part. And "my workflow" was not "wrong" being the default workflow Resolve employs for their cameras. The way things are done on Resolve must be different with RAW cameras, than what has been suggested so far. Everyone is doing the "wrong" workflow with BRAW by using Rec709 2.4 (which is the default on Resolve).

Also, Dmitry, this is not a problem of IR pollution, since I did provide a picture of the tree without the problem, while you could not fix it under the normal Resolve workflow! How do you explain that?

Rec2020 might be a delivery standard, but so is rec709. I didn't find it to be an issue. For the last time, RAW footage of Blackmagic cameras don't fit in rec709 for processing the way normal 8bit or 10 bit LOG or rec709 footage does. The more saturation/contrast and detail exists, the more its "color bits" spill outside of the visible spectrum or the rec709 spectrum. Raw footage *begs* for big color spaces! And if your argument is that it "should have" fit, well, it doesn't fit gracefully! It clips on the sides of the rec709 triangle. The more it clips, the more colors are mashed together internally when the footage is interpreted. The more I manipulate the footage to stay inside the triangle, the more we can get rid of the color cast.

Currently, there are a few monitors that can display more colors, at around $1300-$2000, but they're not doing that much better than the monitor I already have. The only monitor that probably can do much better, and let you work in rec2020 without an issue, is the Apple one for $6k. Basically: we got a RAW camera, but tech hasn't moved fast enough to provide us with the rest of the hardware processing chain required to use that RAW camera to its full capability. So in order to go around these problems, we have to "shrink" the RAW footage to something akin to 10bit h.264 log (where encoding has removed information), in order for it to behave in our hardware.

It sounds like you know what you are doing so what specifically do you need help with?
Which is what I said, and even though this is what I was talking about, there is not too much comparison between your actions at the moment compared to some not around here.
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 6:49 pm

It sounds like you know what you are doing so what specifically do you need help with?


I don't like the tone of what you just wrote, which is essentially "go away" in a passive aggressive manner.

I'm providing a solution to both the OP's question, and to my original comments. Because obviously, there was NEW information to be shared on the exact topic, that goes around the problem.
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
Offline

John Griffin

  • Posts: 576
  • Joined: Sun Aug 23, 2015 4:18 pm

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 7:14 pm

Eugenia Loli wrote:
It sounds like you know what you are doing so what specifically do you need help with?


I don't like the tone of what you just wrote, which is essentially "go away" in a passive aggressive manner.

I'm providing a solution to both the OP's question, and to my original comments. Because obviously, there was NEW information to be shared on the exact topic, that goes around the problem.

I'm sorry you feel that way but that was not my intention.
Which is what I said, and even though this is what I was talking about, there is not too much comparison between your actions at the moment compared to some not around here.
Offline
User avatar

rick.lang

  • Posts: 13391
  • Joined: Wed Aug 22, 2012 5:41 pm
  • Location: Victoria BC Canada

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 7:15 pm

Peace and love, Eugenia. I liked your approach in that earlier post as you stayed up most of the night finding a solution that works. I have that Apple Pro Display XDR that can display a good portion of Rec2020 but I haven’t done real work with it yet. I can shortly... my dog is hanging on by a very thin thread which is my priority sadly.
Rick Lang
Offline

John Griffin

  • Posts: 576
  • Joined: Sun Aug 23, 2015 4:18 pm

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 7:22 pm

Dmitry Shijan wrote:
John Paines wrote:Dmitry, can you link to the two versions of the lake shot (the originals)?

I believe Panny said somewhere or other that the Varicam LUTs are suitable for the SH1 -- it's effectively the same color science as is present on the EVA1, the cameras match. But can't tell you where I saw it.


It is from JSFILMZ youtube test video. He put links to original files in video description or in comments under the video
Here is link to that P4K vs S1H tests (22GB) https://drive.google.com/file/d/1yA78s5 ... 5ehlc/view

By the way, i remember his BMMCC vs Pocket4K test at same location with lake and tree, and BMMCC produced more IR pollution that Pocket4K (I compared original sources, but not his youtube poor graded example) Too many IR pollution could be due water reflecting double amount of IR light. Also both cameras use different lenses, so it could be due some specific lens tint mixed with IR.
I done a lot of summer days shoots with BMMCC before i got OLPF/IR filter and never ever see similar extreme brown grass and trees instead of fresh green. It was some shift in green colors but it was never so huge.

Just downloaded these and put them in Resolve in an ACES CM. I can get the greens to match in the lake shot very easily with just a temperature match and a slight Hue vs Hue tweak. The interior shot matches less well but still pretty good ( I think there may be a slight lighting change between them). Owning both the P4k and S1 ( V-LOG) this matches my own experience and these 2 cameras can be paired pretty well with regards to color. Certainly better than either vs my GH5. I've got the Rawlite OLPF / IR filter but haven't shot anything yet due to the lockdown but I got it for the OLPF mainly as never had any IR pollution issues.
Which is what I said, and even though this is what I was talking about, there is not too much comparison between your actions at the moment compared to some not around here.
Offline
User avatar

Eugenia Loli

  • Posts: 203
  • Joined: Mon May 04, 2015 6:47 am
  • Location: Spokane, WA, USA

Re: massive shift in green color on BMPCC4K

PostFri May 22, 2020 7:32 pm

Thank you Rick, you're a saint! All my love and best wishes!

Regarding the settings I used, to be able to reproduce (I'm not in front of Resolve right now, so it's all from memory, I can provide more detailed info if anyone asks):

- YRGB managed in the project properties
- Rec2020 in all 4 categories below there. Leave saturation/luminance to None (because they make the footage clip!).
- In the raw tab, fix exposure, WB and tint if required.
- Create three nodes. On the second node, do CST to Arri LogC, but leave the saturation/luma unmapped (if I remember correctly).
- Add the arri rec709 lut to the third node. Change that node only to a larger colorspace (I used aces).
- Do adjustments to RAW tab's ISO to fix exposure after having done the arri transformation.
- On the very first node, lower contrast more, put saturation to 30-35, and soften the image to 0.53. That's the part where information is removed and makes it fit in rec709 better. Here you can also fix highlights/shadows etc that might be clipping in your monitor in rec2020.
- Add more nodes after the third node, and do your normal color grading and film emulation.
- At the last node, do another CST to Rec709, with sat/luma mapping (it clips rather gracefully via CST). It seems that doing that via CST (rather than working in rec709 all along) does behave differently.

That's how I got the tree shot looking like it should be.
Collage artist, illustrator, filmmaker: https://vimeo.com/eugenia
PreviousNext

Return to Cinematography

Who is online

Users browsing this forum: Denny Smith, Google Feedfetcher, Stephen Fitzgerald, tclark513, Yannick Willox and 34 guests