Phase One & Sony Trichromatic Bayer

[Matt W.]

I'm not very informed on the finer aspects of optics, but are these Tiffen filters, particularly the Enhancing one, relevant to this discussion:

https://tiffen.com/color-fx/

That's an interesting filter, but not a particularly useful one IMO. It's the same idea but more primitive. I think it contains didymium (a mix of two rare earth elements) and was used as protective glass because it blocks certain orange wavelengths that blown glass emits, just how Enchroma's color blindness glasses were originally used as laser safety notch filters.

In my experience, the filter looks slightly magenta under some light and slightly green under other lights. Even under the same color temperature, depending on if it's LED or fluorescent or incandescent. Just a weird filter but to me it usually seems to just to impart either a green or magenta tint. In theory it's similar, but I think something more aggressive and targeted would be better.

 

[Matt W.]

Adding in a removable notch filter would therefore be the best approach - it's not permanent and you could get different grades or strengths to tailor things for different uses or effects. That'd be pretty cool to play with.

Graeme

Thanks for the feedback. I just need to find someone who's willing to manufacture one... Enchroma wouldn't go for it. Not sure who else might.

 

[Hrvoje Simic]

Which is why I think some sort of "notch ND" designed to filter out the overlap would be quite a big seller for landscape photographers.

That is a) unnecessary, b) highly unlikely.

And attempting it would be a waste of time and resources because many things can be compensated in signal processing with today's sensors with high imaging performance.

(I'd say)

 

[Matt W.]

That is a) unnecessary, b) highly unlikely.

And attempting it would be a waste of time and resources because many things can be compensated in signal processing with today's sensors with high imaging performance.

(I'd say)

Fair enough, although I'd argue that while digital cameras are easier to match than film stocks, it's not always possible to match them perfectly and that some level of color performance is baked in at the raw stage just by virtue of what filters are used in the BFA. I have observed differences in color saturation etc. that I couldn't time out without introducing too much noise or certain areas where color pollution (particularly infrared) needed to be notched out with, say, an IRND.

Do you find the "Trichormatic" camera to be a waste of resources, too? Or IRNDs? Do you think it's trivial to match the performance of the "Trichormatic" filter by applying a LUT onto a standard camera's raw image? What about with IR pollution? When I shot film I remember that I couldn't get other fast film stocks to look as saturated as Velvia (which has published spectral acceptance curves that are very narrow), nor Velvia to look as good as Velvia 100 (which reaches way too far into IR). Granted my photoshop skills are rudimentary, but nothing I did could perfectly compensate. And no digital system I've used yet has produced colors as vivid as Velvia 50.

While I am no expert on this, I'll still play devil's advocate. Going back to the example Karim posted earlier, would you consider a didymium filter (generally not that useful) pointless for astrophogrpahy:

https://www.lonelyspeck.com/hoya-in...-light-pollution-filter-for-astrophotography/

Its biggest notch is in the same exact area as the sodium vapor wavelength pair (I believe around 589 nm) and the results speak for themselves. (Granted, I usually find this filter pointless, but here is a very good use for it.)

The reason I ask isn't to disagree with you. I'm not an expert in this field and you are. But because if that level of granular control were available in post where you could notch out wavelengths on a spectral curve based on some sort of math comparing nearby photo sites, there would be a lot of power toward cleaning up color pollution from street lights and low quality LED and fluorescent lights far better than is currently available as well as dramatically improving the color performance of cameras, which have suffered in recent years as BFAs have gotten thinner. (Compare the F5 and F55, for instance. The F55 is slower but has better color due to thicker BFA filters and that's why it's approved for rec2020 color and the F5 isn't and Netflix wouldn't accept the F5. If this improved color performance be achieved exclusively with software transforms, that would elevate the F5 considerably.)

Even if the hardware is indeed pointless, the goal is a valid one. I can't be the only photographer who wishes digital had richer color rendering, or else Sony and Phase One wouldn't have introduced this product. But if there's a lossless and more flexible software solution that allows reverse engineering notch filters from local BFA deltas or something, all the better. I just can't imagine how it's possible.

 

[Shane Betts]

It makes sense that Phase One are doing this. Landscape photographers are a major market segment for them and I guess they're trying to give people something by way of Velvia. More listening than telling...

 

[Hrvoje Simic]

Fair enough, although I'd argue that while digital cameras are easier to match than film stocks, it's not always possible to match them perfectly

"Perfectly" is a qualification relative to criteria.

If by "perfectly" you mean "unnoticeable colour difference to audience with properly shot intercut material" - that is possible.

If by that you also mean re-inventing missing DR and compensating performance in signal extremes and compensating missing spectrum coverage and non-existent tonal separation (due to specific optical filter/lens/CFA/raw processing ) - that isn't.

and that some level of color performance is baked in at the raw stage just by virtue of what filters are used in the BFA.

Yes of course, significant level.
But a lot happens after that as well.

I have observed differences in color saturation etc. that I couldn't time out without introducing too much noise or certain areas where color pollution (particularly infrared) needed to be notched out with, say, an IRND.

IRND is another case.
You are cutting off/shelving a lower bottom of captured spectrum, not in-between wavelengths of usable range. When you do the latter all kind of mess can happen.

General colour saturation case depends on quite a few variables, many of which rely on principles and limitations of image's base construction (base colour science) and post production image manipulation tools. Some image aspects are beyond their reach but the signal can still be particularly shaped in other ways nevertheless.

Do you find the "Trichormatic" camera to be a waste of resources, too?

It has just been introduced with zero useful info on what is actually achieved and without test and comparison examples.
Pure logic suggests the conclusion that some imaging professionals invested resources into something sensible and meaningful.

Do you think it's trivial to match the performance of the "Trichormatic" filter by applying a LUT onto a standard camera's raw image?

Pretty sure it is not trivial but whether possible or not, I'd have to thoroughly analyse that camera and that other one to make a valid statement.

I wouldn't advise targeting to match another unused camera but to get the most out of the used one.

When I shot film I remember that I couldn't get other fast film stocks to look as saturated as Velvia (which has published spectral acceptance curves that are very narrow), nor Velvia to look as good as Velvia 100 (which reaches way too far into IR). Granted my photoshop skills are rudimentary, but nothing I did could perfectly compensate. And no digital system I've used yet has produced colors as vivid as Velvia 50.

Yes, that is understandable and well founded.

I'd enjoy giving it a go, though.

While I am no expert on this, I'll still play devil's advocate. Going back to the example Karim posted earlier, would you consider a didymium filter (generally not that useful) pointless for astrophotography:

https://www.lonelyspeck.com/hoya-in...-light-pollution-filter-for-astrophotography/

Its biggest notch is in the same exact area as the sodium vapor wavelength pair (I believe around 589 nm) and the results speak for themselves. (Granted, I usually find this filter pointless, but here is a very good use for it.)

We are not talking about astrophotography but general photography. Where visible spectrum has to be fully covered and with minimal wavelength coverage oscillations.

So not like this:

http://www.hellma-analytics.com/text/1245/en/didymium-liquid-filter.html

The reason I ask isn't to disagree with you. I'm not an expert in this field and you are.

My ignorance is vast though constantly decreasing.

But because if that level of granular control were available in post where you could notch out wavelengths on a spectral curve based on some sort of math comparing nearby photo sites,

Wavelengths are a measurable property of light.
Digital domain deals with quantized data.

there would be a lot of power toward cleaning up color pollution from street lights and low quality LED and fluorescent lights far better than is currently available as well as dramatically improving the color performance of cameras, which have suffered in recent years as BFAs have gotten thinner.

It is not just the wavelengths i.e. colour tonalities achieved by their capture.
It's also how the colours behave and interact throughout the signal range.

Colour pollution caused by poor light quality is not something camera should clean/filter.

(Compare the F5 and F55, for instance. The F55 is slower but has better color due to thicker BFA filters and that's why it's approved for rec2020 color and the F5 isn't and Netflix wouldn't accept the F5.

I don't know which methodologies are used to qualify final imaging performance and gamut coverage but pretty sure F5 results can be further improved.

Gamut is determined by signal mapping, not a sensor.
Sensor determines color purity. Which gets mapped to a gamut.

If this improved color performance be achieved exclusively with software transforms, that would elevate the F5 considerably.)

Every camera's final image can be further improved.

In the scope dependent on:
- Sensor performance
- Base color science
- Quantization precision and recording format properties

Even if the hardware is indeed pointless, the goal is a valid one.

The goal is valid and hardware is not pointless. There is room for improvement there as well.

Just saying that not all things necessarily have to be achieved on a sensor level. And with some cases and image properties they are better left avoided at that stage.

I can't be the only photographer who wishes digital had richer color rendering or else Sony and Phase One wouldn't have introduced this product.

Sony camera. Factory route.

Sony camera. Custom route.

But if there's a lossless and more flexible software solution that allows reverse engineering notch filters from local BFA deltas or something, all the better. I just can't imagine how it's possible.

It cannot be fully lossless due to nature of data processing but losses can be beyond perceivable levels.

It is not about reverse engineering anything. It is about additional engineering in digital domain.

 

[Matt W.]

Thanks, that's given me a lot to think about. I'm still going to try to get my hands on one of these filters but I'm in no rush to try to sell it or distribute one. It sounds like its utility is limited, but I'm still very curious.

I do agree that in-camera processing is underrated. The images you posted look great, but I'm sort of curious about something special purpose for otherworldly landscapes. I also agree that IRND and 589nm sodium vapor light are extreme examples, but interesting nonetheless. I think using an "enhancement" filter (which I normally find a bit useless) for astrophotography is brilliant.

 

[David Mullen ASC]

It's interesting Graeme what you are saying about 3-strip Technicolor. That process passed light through a prism and half was sent through a green filter to an orthochromatic b&w negative stock, and then the other half of the prism passed the light through a magenta filter and first hit a blue-sensitive b&w stock that was dyed red, to filter the light hitting a panchromatic b&w stock bi-packed behind the blue-sensitive stock. (The blue-sensitive stock had no anti-halation backing to allow an image to pass through to the panchromatic stock and one of the artifacts of 3-strip photography was magenta halation in the final image around bright points of light or a hot sky.)

But the dyes used for printing each record were yellow, cyan, and magenta, being a subtractive process.

So how would you say that the color cyan was reproduced? I assume a cyan color in a subject exposed some of the blue record and some of the green record? So some crosstalk was allowed?

 

[Hrvoje Simic]

Thanks, that's given me a lot to think about. I'm still going to try to get my hands on one of these filters but I'm in no rush to try to sell it or distribute one. It sounds like its utility is limited, but I'm still very curious.

I do agree that in-camera processing is underrated. The images you posted look great, but I'm sort of curious about something special purpose for otherworldly landscapes. I also agree that IRND and 589nm sodium vapor light are extreme examples, but interesting nonetheless. I think using an "enhancement" filter (which I normally find a bit useless) for astrophotography is brilliant.

It is important to make a distinction between special purpose and creative intent and general purpose and overall usability and picture accuracy. I encourage new routes for creative expression and specific purposes. Development requires exploration. And explorers have to venture beyond the safety of known.

Processing which determines the final result coming of the sensor can be segmented to:
a) in camera (electronic processing > gathering data),
b) base colour science (interpretation and structuring of data to get a usable image),
c) RAW to RGB (or any colour space) conversion,
d) color signal optimization
e) post production image manipulation

C and D are best done in one step and every step pre-determines the result of the following one.

So in the context of "achieving richer colours" there is a huge amount of inter-dependent and consecutively ordered parameters, defining in which way, to what extent and type of result that can be accomplished. Under the hood getting much, much more complex than it may seem with hands on image manipulation tools and eyes on the monitor.

Undesirable effects of specific types of light sources in digital acquisition are in great part caused by signal deviations of digital quantization and typical signal processing and can be significantly reduced through special treatment of camera's signal. Skilled colorists can achieve this to an extent but post production tools also bring specific limitations of principles they are based on, and best results are achievable by a more elaborate and complex process done before the image gets a post treatment.