Have we had DVD colors wrong this whole time?

[DS9Redefined]

Hey everyone, new to this forum but been working on a DS9/Voyager project for 5 years now. I did a quick search and didn't find a thread discussing this, so sorry if this is old news. It's new news to us. 😅 If anyone has any inside knowledge and either we're off base or there's some reason that somehow no one ever noticed this, by all means. We're at a loss as to how this has gone on for over 20 years.

One of our team members figured out something really simple that drastically improves the colors on not just Trek, but for every 80s and 90s American TV show that's been transferred to DVD that we've tried so far. Suddenly these shows look like what we remember them looking like when broadcast.

Basically, the DVD video streams are tagged as BT.601, which seems right because it's been the standard for SD digital video since the 80s. However, while these shows were primarily edited using digital tape systems, they were formatted for analog color tv broadcast, which never used BT.601. They used NTSC 1953, aka FCC, or BT.470-2 System M.

So, if you either re-flag the metadata and play in a player that supports BT.470m, or do a full conversion to Rec.709 assuming BT.470m input, well, the results speak for themselves.

https://pbs.twimg.com/media/GsKynKVaUAI7COX?format=jpg&name=small

https://pbs.twimg.com/media/GsKq1rrbUAAisHK?format=jpg&name=small

https://pbs.twimg.com/media/GsKrk0RaUAUVh4g?format=jpg&name=small

So, pick a show and try it for yourself!

Reflag:
ffmpeg -i "input.mkv" -c copy -colorspace fcc -color_trc gamma22 -color_primaries bt470m -color_range tv "output.mkv"

Convert (replace libx264 with your codec of choice):
ffmpeg.exe -sws_flags "spline+accurate_rnd+full_chroma_int" -i "input_file.mkv" -vf "zscale=matrixin=fcc:transferin=bt470m:primariesin=bt470m:rangein=tv:agamma=false:dither=error_diffusion:primaries=bt709:matrix=bt709:transfer=bt709:range=tv" -c:a copy -c:s copy -map 0 -c:v libx264 -crf 18 -preset ultrafast -flags "+ildct+ilme" "output_file.mkv"

You can read more about it here:
https://x.com/jonpoeschek/status/1928307054365790559

[wonkey_monkey]

Just because it's brighter and bolder, doesn't mean it's more "correct". It's just personal preference. You can take any image and find all kinds of tweaks that make it look "better" to you.

but for every 80s and 90s American TV show that's been transferred to DVD

Well yeah, if it increases the saturation on one image, it will increase it on all images. Doesn't mean it's right, though.

Maybe you just all had the colour turned up way up on your TVs back in the day. And you still can, if you want.

[Blue_MiSfit]

I think the white balance is much more correct. Skin tones and neutral greys were very green on the left.

My intuition tells me the conversion is actually correct.

[wonkey_monkey]

neutral greys were very green on the left.

Which parts do you think are meant to come out as neutral grey, and how do you determine that?

The curved section of wall above Tuvok, for instance, or the area to the left of B'Elanna, are closer to grey on the left than the right, where they have turned quite a lot more blue.

I tried the conversion method on an episode and, frankly, I don't think Chakotay is meant to be the colour of a sweet potato (it's even more pronounced in an earlier corridor scene):

https://i.imgur.com/KwcnG0L.png

The wall and the Okudagrams behind Seven look far too cyan.

I note also that on Twitter, it's been stated that an additional RGB filter was applied, which again makes this all sound like a personal preference for more saturated images (which humans tends to prefer anyway, all else being equal).

Someone on Twitter did the conversion directly in MPV to produce this, which looks even worse to me (flipped to keep the oversaturated conversion on the right):

https://i.imgur.com/RWw4Wbh.png

[Selur]

There is always the question of color correction vs color grading and whether the transfer to medium xy was botched up.
To know for sure, you would need to ask the director or color artist that colored the original version before transfer (to medium xy) what colors really were intended.

Cu Selur

[huhn]

ahhh this scares me: "spline+accurate_rnd+full_chroma_int"

[Z2697]

ahhh this scares me: "spline+accurate_rnd+full_chroma_int"

swscale default have lower quality without some of these flags.
But actually, the filter used in the command is zscale, and it doesn't use these flags.

[kurkosdr]

There is always the question of color correction vs color grading and whether the transfer to medium xy was botched up.
To know for sure, you would need to ask the director or color artist that colored the original version before transfer (to medium xy) what colors really were intended.

Cu Selur
If I may add, before HDR10 came along (and brought wide color gamut aka Rec.2020 with it), you couldn't assume a given colorist knew or cared about colorspaces. I mean, most TVs didn't care, not even in Cinema mode, so there was no advantage in hiring a colorist that does.

Even today, there is no consensus whether upscaling from 720x480 to 1920x1080 should include a conversion from Rec.601 to Rec.709 or you just take the YCbCr values as-is and resize.

[huhn]

If I may add, before HDR10 came along (and brought wide color gamut aka Rec.2020 with it), you couldn't assume a given colorist knew or cared about colorspaces. I mean, most TVs didn't care, not even in Cinema mode, so there was no advantage in hiring a colorist that does.
no calibration was taken serious at all times. calibration is not something new. and special "TV" where used just like today that where made to show the image as is. because TVs these days have an FMM mode or other stuff doesn't mean they are calibrated.
Even today, there is no consensus whether upscaling from 720x480 to 1920x1080 should include a conversion from Rec.601 to Rec.709 or you just take the YCbCr values as-is and resize.
because it does not matter.

you can take bt 2020 hdr convert it using bt709 then convert it back using bt 709 followed by b2020 and get pretty much the same image as you would get with using bt 2020 or nothing in the first place.

[flossy_cake]

[deleted - thought the thread was about The Next Generation]

[flossy_cake]

Tip: if you are playing back with MadVR you can simply tag a file or folder with [primaries=470M] to tell MadVR to decode it as 470M

For reference the relevant MadVR tags are:


[primaries=170M] or [primaries=SMPTE-C] for the default NTSC decoding

[primaries=470M] for the alternate 1953 NTSC decoding (what OP recommends)


There is also a [matrix=601] and a [matrix=470M] tag, but both 170M and 470M use the same 601 coefficients so there is no difference between the two.

Colour definitions: https://linuxtv.org/downloads/v4l-dvb-apis-old/ch02s06.html

The above definitions say there is also another NTSC standard called 240M which was "used during the early days of HDTV (1988-1998)", coincidentally when DS9 was made (1993-1999). This 240M standard has the same primaries and white point as 170M, but different matrix, so you could try previewing it with [matrix=240M].

I would agree in the OP's examples, the 470M looks subjectively more pleasing than 170M, however I'm not sure it's definitely "correct". You would have to know the whole chain the colourist was using to know the correct answer. On some other scenes 470M seems a bit off to me with skin tones looking a bit too pushed (this could depend on your display a lot too) but then again the 170M decoding has its own issues. So in the end I dont know, so I'll just stick to 170M for now.

[huhn]

if the dvd is not correctly displayed with the tags it has then the colorist has failed as easy as that.

it does not matter what is was at what ever point only thig that matter is the end result and a matching tag.

[flossy_cake]

Even today, there is no consensus whether upscaling from 720x480 to 1920x1080 should include a conversion from Rec.601 to Rec.709 or you just take the YCbCr values as-is and resize.

The conversion from 601 to 709 must take place otherwise you'll end up with 601 colours being decoded as 709 by the device doing the decoding, and the resulting colour will be wrong, and it is not a small error either. Basically the colours have to be decoded in the same way they were encoded.

For upscaling PAL to HD 709 you can get away with not bothering with a gamut conversion because PAL and 709 use very close gamuts (only the green primary is slightly different - see document of colour definitions previously linked to). However the matrix formula (RGB <-> YCbCr formula) is still very different between PAL and HD 709, and that one cannot be ignored unless you want noticeably incorrect colour.

NTSC (170M/SMPTE-C) uses the same matrix as PAL so that needs conversion too when upscaling NTSC to HD 709, plus the gamut is different to PAL and 709 as well, notably the red is less saturated.

[flossy_cake]

Just because it's brighter and bolder, doesn't mean it's more "correct". It's just personal preference. You can take any image and find all kinds of tweaks that make it look "better" to you.

Well yeah, if it increases the saturation on one image, it will increase it on all images. Doesn't mean it's right, though.

Maybe you just all had the colour turned up way up on your TVs back in the day. And you still can, if you want.

After more consideration, I think I will have to agree with you on this point.

I will say this: when trying to determine what the intended colour was, it is ENTIRELY IRRELEVANT which one looks better to me. Why? Because colourists make horrible choices with colour grading all the time! This is occurring at truly epidemic levels lately, so it shouldn't surprise me at all when I see gross colour grading with puke skin tones and overuse of undersaturation.

However, in OP's defense this was less of an issue in the 1990s - colour grading was generally nicer back then imo. Still, there's plenty of horribly graded stuff from that era too, but usually it's due to signal loss by analogue devices , for instance DS9/Voyager has dot crawl and rainbows baked into it, so it all went through a yellow composite cable at some point and colours get muted by that.

[flossy_cake]

Another consideration is that 470M is a larger gamut than 709, so in your "convert" example in post#1 where you are shrinking it to 709 with zscale, that will introduce clipping or other artefacts depending on the rendering intent used by zscale which performs the gamut shrink.

The only way to avoid such artefacts is to just flag the file as 470M like in your "reflag" example and display it on a wide gamut display with a media player renderer like MadVR that sees the 470M flag and can properly convert it to the display's wide gamut like P3/2020 which is capable of displaying 470M.

[flossy_cake]

I think the white balance is much more correct. Skin tones and neutral greys were very green on the left.

470M white point is different tho - more blue

470M white: x=.310 y=.316
170M white: x=.313 y=.329

I'm struggling to think how we might use this fact to determine the intended white. Even if we had some reference white pixels somewhere in the show, I'm not sure how that would tell us what colour should be at the display. Because that colour is going to be the result of how we interpret it, and we don't know how to interpret it yet. All we know is we have a pixel colour that decodes to RGB parity. Until we know what colour that parity means, we're clueless. So I don't think this can be solved by looking at pixel values.

I think we would need an actor in a scene holding up a reference white colour swatch under some known lighting conditions illuminated by a bulb with known characteristics. Then if we display that on a 170M calibrated display and the colour is less blue than it should be, that would indicate 470M.

[Sunspark]

We are in luck here, because props from the shows including uniforms worn by the cast have been auctioned at various times so we can look at them taken by cameras.

Picard's TNG uniform season 1-2:

It looks bright because it was.

Season 3-7 they changed the red to a darker shade.

DS9 and Voyager used late TNG red.

The lighting and fabric used in the shows changed which also affected appearance.

https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi8nRcAJhqvA6DukdF885LCD71MVYviyzp8fOaHXuoTX2bCLHvEKhmVdKQimmGzKuJe9YYBe4xj6PCmWEC3xPRnZ18ctwx1KAwr2APR1bwGhibkZQBpdJj_ySh8LB7NO0pdKuESycxxT_o/s1600/89905_Captain+Picard%2527s+Patrick+Stewart+Starfleet+Uniform_1b.jpg

This one here is a jacket and pant set worn by Janeway on Voyager:

https://web.archive.org/web/20160120011448im_/http://startrekpropcollector.com/trekauctions/imgsrv.pl?b4395d28581058bc90bc8037b52a7ede

Do the comparison images on the right look a bit hot, yes, but I used to watch these shows on aerial broadcast on a CRT and the images on the left are NOT how they looked on a real TV of the day.

[Sunspark]

This one from TOS:

https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhHvxFso4pVKNYAKqmwumPm7QQllMkMblDrVra8y72n3YImfp189BAZ51m9odqnIxtw1oUlJrhs_ZpSWsfi2d253yTgXxX035X1wOTnVY7MN-ps7Yn1OLTD9QVr-bHRrb2N1gr51KlCCu4/s1600/img10.jpg

https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgDUIh-MR6irx-3B3uOea7OJn3w5egKWdUjdZywAUSHl1lUoNbd_IwMwlOGySoJiSPBYXI5rTQKmv-3MWPyPDpovAHKn-XLEJ5NVl-Q_mYs0MeijxRcaqWA-4WXiR1GL0FSr6AmIQE1BVo/s1600/img09.jpg

[Sunspark]

Oh, these links are amazing..

This person uses pantone chips and stuff in various lighting conditions:

Gold jumpsuit: https://startrekcostumeguide.com/2024/07/screen-used-ensign-sito-tng-jumpsuit-examination/

Red jumpsuit: https://startrekcostumeguide.com/2024/09/screen-used-male-extra-tng-jumpsuit-examination/

[Z2697]

Photos of the props is not definitive ground truth either. The lighting condition, and cameras' settings are very complex.

[flossy_cake]

Photos of the props is not definitive ground truth either. The lighting condition, and cameras' settings are very complex.

And colour grading of the photos

[Sunspark]

Part of the reason I like Mitchell-Netravali AR for chroma, is because it is friendly to this type of material.

The left photos didn't look like that in broadcast and it doesn't look like that in real life either. You want the colours to have a good dynamic range without being hot or dulled.

True about the lighting condition, part of why I found the pantone chips next to the uniforms in different light conditions interesting. These shows all had outdoor daylight scenes, so it IS possible to see the uniforms under our sun.

[flossy_cake]

The left photos didn't look like that in broadcast and it doesn't look like that in real life either. You want the colours to have a good dynamic range without being hot or dulled..

I would agree that there does appear to be something "off" with the colours when shown in the standard NTSC 170M gamut (left images) as opposed to the wider 470M gamut from 1953 (right images). I would speculate this is maybe due to signal corruption from being through multiple analogue format conversions, possibly been converted to analogue tape at some point as well. Each of these devices imbues its "colouring" of the signal in the same way different analogue audio formats like vinyl vs magnetic tape vs digital. Possibly some hue shifting as well, although I played around with hue control on several scenes with the madvr hotkey for hue and didn't really see much improvement there - rotating hues towards red just made faces look pink and like beetroot and generally unnatural.

Also my QDOLED S90D has low colour luminance in SDR so that just makes these Star Trek examples look even worse , but I've since improved that with the custom colour space calibration. Especially getting the red luminance tracking linearly through the saturation sweep is very important to have healthy skin tones. "Red push" used to be a common problem and I reckon that is maybe why the skin tones looked more healthy on old CRTs which suffered from red push. If you have an old CRT somewhere try playing the DVD version on it and see what it looks like. The last CRT I had was a Samsung and that one had juiced skintones. Even one of my DVRs has skin tone boosting in its video processor that needed to be disabled with Putty.

[flossy_cake]

But just to clarify my opinion - I don't think 470M is the right way to correct whatever happened to the colours. 470M improves it subjectively, and might even be getting it closer to what the colour was supposed to be, but ultimately I'm seeing stuff on scenes with 470M that doesn't appear right either. So I think it needs some correction, but 470M is not the "correct" way to correct it.

[flossy_cake]

[deleted - thought the thread was about The Next Generation]

Oh the OP's tweet says it applies to TNG as well

I don't think it's right...

DVDrip 470M
https://i.imgur.com/9fc3LWj.png

DVDrip 170M
https://i.imgur.com/72b3Ofi.png

BDrip 709
https://c.l3n.co/i/DoFs11.png


I would say both the 170M and 470M decodings have significant hue/saturation errors in the skin tone, so neither is really "correct" imo

[wonkey_monkey]

One could assume more care was taken with the blu-rays, so if you can get the same shot (which episode is it from?) from there, it might go at least some way to clearing up the confusion.

[Sunspark]

The BDrip looks much better and more like what I remember. The DVDrip examples are both wrong, way too much magenta.

Does it make a difference setting the primaries to EBU/PAL? 470 is a Brazil thing and has a different chromaticity, gamma and white point.

[DS9Redefined]

The above definitions say there is also another NTSC standard called 240M which was "used during the early days of HDTV (1988-1998)", coincidentally when DS9 was made (1993-1999). This 240M standard has the same primaries and white point as 170M, but different matrix, so you could try previewing it with [matrix=240M].


It looks like 240m was pretty much only used in Japan except for a few experimental things in the US, and definitely not with the D2 and Digibeta systems of the time. HOWEVER, I was playing around some more with this and found if I set matrix and transfer to 240m and kept primaries as 470m, the skin tones look better, less pink. So... perhaps the monitor calibration of the time was a bit different than what we see on our modern LCDs? Or maybe it was just the style at the time. 👴🏻

https://i.imgur.com/Bne0tPCm.png https://i.imgur.com/IcjA8S7m.png https://i.imgur.com/VetsReTm.png

With these settings, the TNG DVD looks closer to the Blu-ray, just blues and blue-greens are still a bit more Cyan (maybe just a difference in grading choice between OG and BD). Is there any possible way in the workflow of the time they could have ended up with 240m matrix and transfer and 470m primaries?

https://i.imgur.com/3gdgrLMm.png https://i.imgur.com/qU0B5rKm.png https://i.imgur.com/2nQO7fUm.png

[Sunspark]

I found out that for the TNG Bluray, they re-scanned the original 35mm film so that's why the colours are better, there are no other intermediate devices/steps in the chain.

[Sunspark]

Ok this is interesting.. there is a 2018 DS9 documentary "What We Left Behind: Looking Back At Star Trek: Deep Space Nine" where they remastered some scenes for it. Maybe that would help with colour matching.

Full movie: https://www.youtube.com/watch?v=3u3WwFMgy5A

Trailer for it: https://www.youtube.com/watch?v=D4LPDX6uqSM

[DS9Redefined]

I found out that for the TNG Bluray, they re-scanned the original 35mm film so that's why the colours are better, there are no other intermediate devices/steps in the chain.

Correct. I need to look more into it, but I seem to remember a featurette that explained how the film came with liner notes with color correction specs. I'm curious if they stuck to those or improved on it, but it is interesting to note how much closer 470m is to the Blu-ray than 601.

[DS9Redefined]

Ok this is interesting.. there is a 2018 DS9 documentary "What We Left Behind: Looking Back At Star Trek: Deep Space Nine" where they remastered some scenes for it. Maybe that would help with colour matching.

Yeah, we look at the rescans from the doc often, and they're on a whole other level. I think the studio that did those scans probably just made them look great to their taste rather than following liner notes.

[GeoffreyA]

I found out that for the TNG Bluray, they re-scanned the original 35mm film so that's why the colours are better, there are no other intermediate devices/steps in the chain.

One problem with Voyager and, I think, DS9 is that they were edited and finished on tape. TOS must have been a straightforward scan and regrading; the TNG, they rebuilt the CGI. For Voyager and DS9, it seems unlikely that we'll get a proper remastered version from the negatives. For one, do they still exist in their entirety, and where? Then, rebuilding the edit and CGI will be another cost.

[DS9Redefined]

One problem with Voyager and, I think, DS9 is that they were edited and finished on tape.

I don't want to get too far off topic here but I'll clarify this. DS9 and Voyager can be remastered exactly like TNG was. The film is all in the same place the TNG film is. But yes, there is a large amount of CG that would need to either be re-rendered or recreated. The people that did the CG and VFX are ready and willing. The only thing standing in the way is Paramount's willingness to spend the money, as well as the complication of the sale of Paramount.

[GeoffreyA]

I don't want to get too far off topic here but I'll clarify this. DS9 and Voyager can be remastered exactly like TNG was. The film is all in the same place the TNG film is. But yes, there is a large amount of CG that would need to either be re-rendered or recreated. The people that did the CG and VFX are ready and willing. The only thing standing in the way is Paramount's willingness to spend the money, as well as the complication of the sale of Paramount.

Thanks for clarifying. It's sad that money is the obstacle because fans want this, I'm sure. My favourite is certainly Voyager.

[Sunspark]

I'm not an expert, I just enjoy exploring stuff here and there.

So I have learned that 240M is the same as sRGB except for gamma in that it uses broadcast compatible gamma (2.2-2.4) without sRGB's linear toe. Which makes sense because it doesn't look as "dark".

The 240M/470M combo looks decent enough in the sample screenshots and might be as good as it's going to get for DVD.

The blue in Crusher's uniform looks better, the other two look "muddier" and there is one very visible difference. If you look at the light above Picard's head, with the bluray 709 primaries it is blue, but with 470M it is cyan, a completely different shade, which tells me that the light was a different colour since this thing was able to detect that.

I know you guys said not to use photos as a reference, but if we look at this Dr. Crusher photo here, you can see the uniform actually is blue, and not green or muddy. The hair in the photo also looks closest to 240M/470M.

https://i.pinimg.com/originals/ea/d9/4b/ead94b3dcbfd5939bd98be6836f32893.jpg

[wonkey_monkey]

If you look at the light above Picard's head, with the bluray 709 primaries it is blue, but with 470M it is cyan, a completely different shade, which tells me that the light was a different colour

Was a different colour to what?

since this thing was able to detect that.

Nothing is being "detected". Some colours are affected more than others, that's all.

[Sunspark]

Was a different colour to what?

Software claims my monitor is 100% sRGB coverage. I don't know if it's getting the values from the monitor, a dongle, or something else. This thing is 15 years old. The only way I would know for sure would be if I measured it with an instrument.

On my screen, the halo of that light in the bluray version looks blue like aquamarine. But in the 470M versions, the halo has a lot more green in it, I guess you could call it turquoise.

[wonkey_monkey]

Still don't know what "the light was a different colour" is meant to mean...

[Sunspark]

If you can't see it your regular display may have a poor gamut.

How does this look on your display?

https://www.warbyparker.com/learn/wp-content/uploads/2023/02/types-of-colorblindness.jpg

[wonkey_monkey]

I can see it looks more cyan in the right two images compared to the left image. I don't know what you mean by "which tells me that the light was a different colour", because that just seems like a tautology. When you say "the light" are you referring to the physical light present on the set during shooting? Or the image of the light in the final pixels? Nor do I understand what you mean by "this thing was able to detect that."

[huhn]

none of this matters anyway.
even if the prop is red and the image you see is blue then the blue doesn't have to be wrong because if the colourist says it is blue it is blue.

[flossy_cake]

none of this matters anyway.
even if the prop is red and the image you see is blue then the blue doesn't have to be wrong because if the colourist says it is blue it is blue.

Yeah, for all we know the colours we see on the DVDs of DS9/Voyager are what the colourist saw as well, and approved it.

However: what if the colourist made a mistake and didn't know the colours were getting corrupted in the signal chain? Then it could be argued the colourist's decision was actually wrong!

edit: but my theory is because there are rainbows and dotcrawl baked into the DVD version, I don't think the colourist was looking at that "composite video (yellow cable)" copy of the video. So possibly that analogue bodging took place AFTER the colourist was finished? But did they even have colourists back then? I thought "O Brother Where Art Thou" (2000 film) was the first movie to actually be "colour graded"? I don't know how they graded things before then, whether it was things done to the celluloid itself or if they run it through other colour modification analogue tools.

[flossy_cake]

In MadVR we can also tag the file [primaries=DCI] or [primaries=2020] for a "medium" or "high" boost in saturation respectively.

Looking at S1 of DS9, it's interesting how much more balanced DCI-P3 looks, even though I know it can't possibly be correct.

But 170M looks wrong to me too, there is obvious desaturation and greenish/yellowish hue to skin tones. It's quite gross, and I'm pretty sure it's a mastering error of some kind.

I might actually screen this series in DCI-P3, it seems to look better than just increasing the saturation via MadVR's saturation control. Probably because that uses a different formula like eg. HSL model and boosting the S component, whereas a gamut transform probably stretches in a more perceptually uniform way (I'm just guessing here)

edit: but MadVR saturation in the range of +10 to +15 seems to suit S1 of DS9

[Sunspark]

I can see it looks more cyan in the right two images compared to the left image. I don't know what you mean by "which tells me that the light was a different colour", because that just seems like a tautology. When you say "the light" are you referring to the physical light present on the set during shooting? Or the image of the light in the final pixels? Nor do I understand what you mean by "this thing was able to detect that."

Ok, in the bluray that on-set light that I am talking about appears blue as do the larger on-set panel lights. If you went by that one would assume they were all relatively the same. With the other images where one of those on-set lights is a different colour it means the wavelength was different from the larger blue panels and the formula was able to draw or elicit that more as opposed to making them more alike. The thing meaning the matrix/transform formula in tandem with the primaries formula.

It made me wonder if you could see it so I put up that dot image. If you can't see the number in it, you have colourblindness. If you do, then you are fine and it was just a communication issue.

[Sunspark]

I don't know how they graded things before then, whether it was things done to the celluloid itself or if they run it through other colour modification analogue tools.

I don't know, but I do know that on Star Trek the Original Series, they put coloured films over the lights to change the atmosphere or mood in the room/set.

[wonkey_monkey]

If you went by that one would assume they were all relatively the same. With the other images where one of those on-set lights is a different colour it means the wavelength was different from the larger blue panels and the formula was able to draw or elicit that more as opposed to making them more alike.

That's a bit of a stretch, not just because video doesn't encode wavelength. One light is brighter than the other; perhaps the change turns bright lights more cyan than it does dimmer lights.

In any case, it doesn't go any way towards determining what it's "supposed" to look like.

[DS9Redefined]

Alright here's some more to consider. Here's a shot that's got a good number of colors to compare from the DVD of S02E01 The Homecoming.

https://i.imgur.com/atnnhtjl.png

Here is the LaserDisc, captured via Domesday Decoder, saved straight out of ld-analyse.

https://i.imgur.com/YSgsUYQl.png

Here's the ld-chroma-decoder result opened in After Effects. This video file is ProRes 422. It has no color standard flags so After Effects interprets it as Rec.709.

https://i.imgur.com/ZR72eBTl.png

Now copied with ffmpeg adding flags for 470m.

https://i.imgur.com/h8Emjg5l.png

And now with 470m primaries and 240m matrix and transfer.

https://i.imgur.com/i97al9Vl.png

[StainlessS]

Do whatever blows your skirts up, other than that, it dont really matter a damn.

https://www.youtube.com/watch?v=fIh6HDeXKGY

[flossy_cake]

https://i.imgur.com/h8Emjg5l.png


Yeah on that shot it looks much better at 470M but I can guarantee you on other shots and episodes there are gonna be skin tones that will not look right. But like I said I agree with you that 170M decoding looks bad too, desaturated and hued yellow/green. So in the end pick your favourite way of counteracting that I suppose.

Raising saturation in the media player (MadVR in my case) corrects the desaturation but the skin tone is still hued yellowish/greenish. Rotating hue in MadVR couldn't fix that in a way that looks right to me. Using a different gamut like P3, 470M or 2020 seems to increase saturation AND subjectively improves the hue of skin tone to my eyes on my display. 470M mucks up the blues though cause its blue primary is so radically different to all the other gamuts in terms of hue. P3 doesn't suffer from shifted blues, but there are still scenes where P3 looks wrong. So maybe in the end I'll just keep the default 170M and use saturation 10 or 15 clicks in MadVR. I wish I knew which formula MadVR uses for saturation though, cause if it uses HSL or HSB then I don't really want to use it because those models are very crappy and non-perceptually uniform and would just introduce non-linearities in the colour tracking that could show up on certain scenes and ruin those scenes