Colour management is not some new concept and has been around for a long time in stills photography and has been especially usefull in demanding RGB to CMYK conversions in commercial printing. Properly implemented as in RCM it provides precise transforms between different colour spaces without the limitations of LUTs which need the user to expose their footage at exactly the same levels that the LUT was created in. (LUT ‘s are in need of ‘clean room’ conditions to work optimally) LUT ‘s have their uses for creating certain looks but if you are after the best system to transform your camera’s colour space and tonal response to a working or output space the RCM is hard to beat ( maybe ACES). V- gamut on the GH5 is just a shoe-in from the Varicam and is not a properly defined gamut like the one in HLG so the only way at present to get it to work is to make a LUT from scratch. Why bother though when HLG has a precisley defined gamut that can be properly interpreted by RCM and the codec uses more of the data range in the first place. There is no advantage in transforming log to log over log to a hybrid log as long as the mathematical model is known and in any case this is a minor issue compared to the complexity of transforming between colour spaces. There has been a small industry in LUT creation in the video world but I’m convinced the LUT’s days are numbered as more NLE’s adopt colour management and as more cameras can work in bigger colour gamut’s than REC709 and more output devices also move beyond REC709.
GH5 Color matching - GH5 and BMPCC4K
- Thread starter npm
- Start date
Colour management is not some new concept and has been around for a long time in stills photography and has been especially usefull in demanding RGB to CMYK conversions in commercial printing. Properly implemented as in RCM it provides precise transforms between different colour spaces without the limitations of LUTs which need the user to expose their footage at exactly the same levels that the LUT was created in. (LUT ‘s are in need of ‘clean room’ conditions to work optimally) LUT ‘s have their uses for creating certain looks but if you are after the best system to transform your camera’s colour space and tonal response to a working or output space the RCM is hard to beat ( maybe ACES). V- gamut on the GH5 is just a shoe-in from the Varicam and is not a properly defined gamut like the one in HLG so the only way at present to get it to work is to make a LUT from scratch. Why bother though when HLG has a precisley defined gamut that can be properly interpreted by RCM and the codec uses more of the data range in the first place. There is no advantage in transforming log to log over log to a hybrid log as long as the mathematical model is known and in any case this is a minor issue compared to the complexity of transforming between colour spaces. There has been a small industry in LUT creation in the video world but I’m convinced the LUT’s days are numbered as more NLE’s adopt colour management and as more cameras can work in bigger colour gamut’s than REC709 and more output devices also move beyond REC709.
joe12south
Veteran
I think we all look forward to that magical, mystical, majestical someday. But it still isn't even here for still photography. I've been working over 35 years and color management has always been complex. It shouldn't be as difficult as it is, but as soon as you get out of theory, it is.
A slightly different scenario, but as a parallel: I'm currently looking at two monitors that are the exact same make and model and just went through the same calibration routine with the same calibrator, and they don't match. They should be very, very close, but, they're just don't. I can beat my head on the wall and pound my fist on the desk and demand that it *should* work, but that doesn't seem to make it so. That sums up color management pretty well.
filmguy123
Veteran
So both Paul and Sage are on this?
You both have my money already.
You both have my money already.
I didn’t say colour management was perfect - it’s just better than LUT’s. I also didn’t say it was simple and know from working in it myself that it’s highly complex and thus even less likely that a sImple LUT can provide a solution. My identical monitors don’t match either if it’s any help and I’ve abandoned calibration tools and now trust factory presets tweaked by eye. This is pertinent as if identical monitors can’t be matched what variations are in camera sensors between identical models........
joe12south
Veteran
Understood, and apologies if it seems like I was putting words in your mouth. My intent is simply to reinforce to the less experienced (e.g. not been burnt yet) that while one *should* be able to simply use RCM or ACES inputs and outputs to match cameras, in practice it just ain't so.
Agree - it still takes a fair bit of knowledge and skill to use it and for many people a LUT is a far simpler and easier solution to get from LOG to a usable image. One thing I have learnt is that trying to get ‘ accurate’ colour is far harder than getting ‘ pleasing colour’ and that accurate colour doesn’t always look correct.
Sage
Well-known member
Look-up tables and color management are different tools to different ends; they may be used in conjunction to a specific end as desired
While a mathematical transform is inherently smooth, look-ups (especially intricate ones) may easily be less so, especially at lower resolutions and paired with trilinear rather than tetrahedral interpolation.
Interesting note, they are not actually tied to Rec709 (unless intended so), and not exposure dependent (for color space transforms). For measured camera output tranforms, these *are dependent on WB
Hope that helps!
While a mathematical transform is inherently smooth, look-ups (especially intricate ones) may easily be less so, especially at lower resolutions and paired with trilinear rather than tetrahedral interpolation.
Interesting note, they are not actually tied to Rec709 (unless intended so), and not exposure dependent (for color space transforms). For measured camera output tranforms, these *are dependent on WB
Hope that helps!
joe12south
Veteran
Yup, LUT's are neither magic, nor evil. They simply are what they are...a flexible tool. There are good one's and bad ones. Useful tool in the right hands, dangerous in the wrong.
filmguy123
Veteran
Sage, what do you assess as the image quality difference between GH5 and BMPCC 4K?
With color conversion, do you think the two cameras would look largely similar in terms of aesthetic?
Let me clarify my question. Is the main IQ difference in the BMPCC 4K simply the color, creating visual tricks to make the image look smoother, less crisp, less digital, more organic, etc. (for lack of more precise terms)? Or is there other mojo going on that no color conversion or LUT may assist with?
Via a conversion LUT, Will the cameras appear to match in more ways than simply color - or will they still look distinct, but with matching color?
With color conversion, do you think the two cameras would look largely similar in terms of aesthetic?
Let me clarify my question. Is the main IQ difference in the BMPCC 4K simply the color, creating visual tricks to make the image look smoother, less crisp, less digital, more organic, etc. (for lack of more precise terms)? Or is there other mojo going on that no color conversion or LUT may assist with?
Via a conversion LUT, Will the cameras appear to match in more ways than simply color - or will they still look distinct, but with matching color?
joe12south
Veteran
Remember that if shooting RAW on the P4K, the color science is not baked into the file, it is applied in post. The GH5(s) color is mostly predetermined by the profile you choose when shooting.
Other visible differences:
NOISE: The GH5 is going to have more noise at higher ISO's, the GH5S and the P4K will be more similar, but even then how much noise, and what it looks like, will differ from camera to camera from ISO to ISO. (LUT's can't change this.)
NOISE REDUCTION: The GH5 always applies some visible noise reduction. The GH5S *really* applies it a higher ISO's. The P4K assumes NR will be done in post. (LUT's can't change this.)
SHARPENING: The GH5(s) always applies some visible artificial sharpening. With native lenses, there can be even more. The P4k assumes sharpening will be done in post. (LUT's can't change this.)
COMPRESSION: The GH5(s) internal codec throws away half of the color resolution, more heavily compresses shadows, introduces color errors (pushing similar colors to be the same), has visible macroblocking and mosquito noise. The P4k recording RAW will have none of those artifacts. (LUT's can't change this.)
All of the above should be manageable, but they are points worth noting as there are some reasonable steps that can be taken to ensure footage intercuts seamlessly.
Other visible differences:
NOISE: The GH5 is going to have more noise at higher ISO's, the GH5S and the P4K will be more similar, but even then how much noise, and what it looks like, will differ from camera to camera from ISO to ISO. (LUT's can't change this.)
NOISE REDUCTION: The GH5 always applies some visible noise reduction. The GH5S *really* applies it a higher ISO's. The P4K assumes NR will be done in post. (LUT's can't change this.)
SHARPENING: The GH5(s) always applies some visible artificial sharpening. With native lenses, there can be even more. The P4k assumes sharpening will be done in post. (LUT's can't change this.)
COMPRESSION: The GH5(s) internal codec throws away half of the color resolution, more heavily compresses shadows, introduces color errors (pushing similar colors to be the same), has visible macroblocking and mosquito noise. The P4k recording RAW will have none of those artifacts. (LUT's can't change this.)
All of the above should be manageable, but they are points worth noting as there are some reasonable steps that can be taken to ensure footage intercuts seamlessly.
Sage
Well-known member
The GH5 will have a different texture; its natively a bit more brittle due to the sharpening (and NR) that is built in. The P4K will have a silk texture akin to a Red, or other cinema cameras (bypassing processing). The GH5 texture is largely reversible via 'digital OLPF' settings, but the P4K will have an edge here from the starting point. Also, it'll have data and compression advantages, which will contribute to its gradability and silk gradients. Its a purist image. They'll be easily matchable sum total, both in color and texture (in juxtaposition with a GoPro or cell phone). Sharper impressions here, in a month or two
@joe12south Right on
Last edited:
filmguy123
Veteran
Cary Knoop
Veteran
Sharpening, like blurring, is a destructive and thus irreversible operation.
joe12south
Veteran
True, but it can be visually mitigated to a greater or lesser degree depending on the tool and who is weilding it. If the goal is to better match a camera that has sharpening with one that doesn't, then it's certainly possible to do a credible job with any NLE.
I'd personally combine low pass and blur to take the edge off the GH5 footage, but one could also add sharpening to the P4K footage if that is more their cup of tea.
Sage
Well-known member
Sure, this is covered in the current GHa Pdf (last page, an email went out about it a few months ago), but this will be revised shortly;
(In Premiere) It involves placing an adjustment layer above the timeline, setting Opacity to 20%, and adding a Gaussian Blur set to 6. This arose from my comparison of the GH5 to the Alexa in SBS with the same framing to compare OLPF texture (the email had a link to the raw footage)
Essentially, this works because it is like an inverse of a sharpening algorithm; sharpening will increase micro-contrast between details (such as edges), and a partial opacity blur distributes the brightness (or darkness) of detail to neighboring regions, reducing micro-contrast. It is critical that the blur size match the sharpening radius, yet not introduce a gauze (too large). The Opacity setting, in turn, is tied to the strength of the sharpening contrast added (20% is for -5 footage)
filmguy123
Veteran
Sage - superb, thank you. I will read the PDF for more.
Joe - what do you mean by you would “combine low pass”
Thanks!
Joe - what do you mean by you would “combine low pass”
Thanks!
joe12south
Veteran
if you simply apply a blur, then you blur everything equally. But the in-camera sharpening only exaggerates high contrast "edges." A low pass filter will filter out the areas of low contrast, leaving only the high contrast areas to be affected by the subsequent blur filter. This will yield a better result, saving a lot of real, high frequency detail that would otherwise get blurred away.
filmguy123
Veteran
Thanks Joe, that makes a lot of sense. How might I do this effectively in Premiere?
I accidentally set my GH5 to zero sharpening and NR the other day and the shots looked very nasty in the fine detail but it was adequately reversed with a node adjustment in Resolve. This is not a recommendation of how to work BTW - stick to -5 on Sharpening and NR!
TheTrickster
Well-known member
Any idea on similar effect in FCPX?