Welcome to our community

Be a part of something great, join today!

Log in
  • Hey all, just changed over the backend after 15 years I figured time to give it a bit of an update, its probably gonna be a bit weird for most of you and i am sure there is a few bugs to work out but it should kinda work the same as before... hopefully :)

Phase One & Sony Trichromatic Bayer

Shane Betts

Shane Betts

Well-known member
Joined
May 5, 2007
Messages
1,560
Reaction score
0
Points
0
Age
67
Location
Melbourne Oz
http://bit.ly/2jwBNVh

This could have interesting ramifications for all digital camera manufacturers but, with Red's willingness to tread the bleeding edge, it could lead to some exciting stuff.
 
Okay...so concluding by three websites copy/pasting PR, Phase One website and official brochure....so far it is known that there is a new, bayer-related solution named Trichromatic.
Named after three primary colors. Unlike other three primary colors based Bayer sensors. This one is better and special though.

Waiting for something more serious about a $50K product, not smelling like marketing fluff.
 
Phase One make pretty high-end gear. Don't dismiss it yet.

What I read is that, rather than pure RGB, their Bayer pattern has three colours more aligned with the response of the human eye. Kinda makes sense in a way. A, perhaps, more analogue response than one based purely on mathematics. I have no idea personally but I'm wondering if this kind of thinking could open Bayer sensors to more nuanced imaging, kinda like the way different film stocks "colour" an image? A less sterile "digital" look maybe?
 
So this is where I get confused.... "In essence, it’s designed around the concept of mimicking the dynamic color response of the human eye." but the pictures are showing that the spectral response of a typical Bayer CFA has overlap between the red, green and blue sensing pixels and their's has no overlap. Yet the human eye has significant overlap between the spectral response of the long, medium and short cones. This certainly fits in with their lowered base ISO as such a CFA with no spectral overlap will let less light hit each pixel.

Then there's the question of what happens to pure colours that will inevitably "hit" only one of the pixels. On a normal CFA, a pure cyan will hit green and blue. On this system I can see how cyan would hit only green or only blue depending on its wavelength. What colour will the system render it at?

Graeme
 
I think you're right Graeme (of course you are). Great to hear your thoughts.

I think they may be looking at something beyond "perfect". I remember shooting products on tranny and there were some colours you simply couldn't reproduce. Not always ideal but the filmstock brought that "character" to the image. Perhaps they're searching for something like that? Or, perhaps, they're just searching for a marketing angle over Hasselblad.
 
It sounds to me that it might look (depending on amount of overlap - I can't believe there's none as the pictures indicate, but I could be wrong) like 3 strip Technicolor. Now that's a pleasing enough look, but it's hardly accurate.

Graeme
 
Graeme Nattress said:
So this is where I get confused.... "In essence, it’s designed around the concept of mimicking the dynamic color response of the human eye." but the pictures are showing that the spectral response of a typical Bayer CFA has overlap between the red, green and blue sensing pixels and their's has no overlap. Yet the human eye has significant overlap between the spectral response of the long, medium and short cones. This certainly fits in with their lowered base ISO as such a CFA with no spectral overlap will let less light hit each pixel.

Then there's the question of what happens to pure colours that will inevitably "hit" only one of the pixels. On a normal CFA, a pure cyan will hit green and blue. On this system I can see how cyan would hit only green or only blue depending on its wavelength. What colour will the system render it at?

Graeme
Click to expand...

We'll see! :biggrin:

I think the whole RGB principle (and it's iterations) is flawed because it's - RGB, THE chromatic aberration per se. Then again, what do I know...:Angel_anim:
 
I wonder if Hasselblad are saying they have optimised out-of-band filtration of the B & G CFA pixels to more precisely reconstruct the Rod response curve around 498nm - And that they have new research into the human perception of dark colours that make use of this?

Cone-response-en.svg


Or perhaps they are intimating that they have a new optical filter material that creates perfect 'house roof' shape optical responses that only overlap 0.01 nm?

AJ
 
"Or perhaps they are intimating that they have a new optical filter material that creates perfect 'house roof' shape optical responses that only overlap 0.01 nm?" :-)

Graeme
 
Graeme Nattress said:
"Or perhaps they are intimating that they have a new optical filter material that creates perfect 'house roof' shape optical responses that only overlap 0.01 nm?" :-)

Graeme
Click to expand...

Hey Graeme, I agree that their marketing is misleading (and probably pretty inaccurate).

I've talking with Enchroma (they make the color blindness-mitigating glasses that they based off of notch filter glasses for lasers) about possibly making some sort of notch filter for landscape photography. Like an ND except it only cuts out the wavelengths where traditional BFA arrays overlap most.

Even if Phase One's marketing is full of shit, their product should provide subjectively richer colors at the expensive of being less sensitive to light, no? Is it possible to make a screw on filter that would provide some of the same benefits by notching out the overlaps? Not with the aim of accuracy, but for landscape photographers looking for slightly bonkers colors like we had with Velvia and willing to sacrifice a few stops of sensitivity? (Velvia has VERY narrow spectral acceptance curves.)

Thanks!
 
So there's multiple approaches to this - filtering as you suggest, adjusting the spectral sensitivity of the CFA dyes or doing an algorithm in software. All three ways could I think get you to a very similar place. I have a 3 Strip simulation 3D LUT for IPP2 that does just this - removes the colour overlap and it gives you that very saturated look. That's really just what a traditional saturation control is doing, but there's many many different ways of achieving saturation control.

The issue on the software side is the overlap removal adds noise. But so does removing the overlap optically as you're lowering sensitivity. It's no-win which-ever way you look at it. But my concern with the optical approach is that it's undoable, whereas in software you can tweak and tweak again without worry.

Graeme
 
These are all good points Graeme. The rush on vintage glass and film grain filters tell you not everyone wants just accuracy and that, sometimes, organic-style limitations, errors and artefacts are more desirable. I agree that having it locked into your sensor is a bit permanent, but then so is monochrome...
 
Shane Betts said:
Phase One make pretty high-end gear. Don't dismiss it yet.
Click to expand...

I'm not dismissing anything. Just saying the marketing approach is "not the best". This is a serious product with stated serious competitive advantage, with zero info provided about this solution. Which would actually help in sales. Basically saying "This is great and that's it."

Shane Betts said:
What I read is that, rather than pure RGB, their Bayer pattern has three colours more aligned with the response of the human eye. Kinda makes sense in a way. A, perhaps, more analogue response than one based purely on mathematics.
Click to expand...

Yes, this suggests a custom CFA.
You can go to Sony sensor division and ask for custom made sensor based on your specific preferences. Which is what Nikon does, for example.

Shane Betts said:
I have no idea personally but I'm wondering if this kind of thinking could open Bayer sensors to more nuanced imaging, kinda like the way different film stocks "colour" an image? A less sterile "digital" look maybe?
Click to expand...

More nuanced imaging is already available today with high quality sensors, wide gamut color spaces and 16 bit sampling.

"Digital look" is mainly caused by principles of digital quantization and data processing. Which can be compensated in the signal chain, making imagery look not just natural and "organic" (film emulations can give that impression) but also as you see it with your own eyes.

And also is available today, being already implemented with...let's see....15 or so cameras.
 
All good points. As Graeme has pointed out, make a LUT. And maybe it's just "New and improved" but I thought it interesting, particlularly coming from Phase One.
 
Graeme Nattress said:
So there's multiple approaches to this - filtering as you suggest, adjusting the spectral sensitivity of the CFA dyes or doing an algorithm in software. All three ways could I think get you to a very similar place. I have a 3 Strip simulation 3D LUT for IPP2 that does just this - removes the colour overlap and it gives you that very saturated look. That's really just what a traditional saturation control is doing, but there's many many different ways of achieving saturation control.

The issue on the software side is the overlap removal adds noise. But so does removing the overlap optically as you're lowering sensitivity. It's no-win which-ever way you look at it. But my concern with the optical approach is that it's undoable, whereas in software you can tweak and tweak again without worry.

Graeme
Click to expand...

Interesting... I still think the hardware solution is preferable to the extent that you're using it in situations where you have the light to combat the decreased SNR (if you look at the chip's specs it's significantly slower than the standard BFA sensor) and in fact for landscape photography I prefer much slower sensor to emulate large format. (I know. It's weird.)

I agree that this is very special purpose. You would never shoot portraiture (or people at all) on Velvia, but Velvia is the gold standard for color landscape photography. So special purpose isn't bad. Yes, you can do most of that work in post but I'd argue not all of it. Some newer generation digital cameras seem a little flatter with color than older ones, even. Or it could just be me. But the thinner and thinner filter arrays seem to do something. And the more you play with color in post the more the image gets thin and noisy.

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. You're getting a more saturated looking image without punching up saturation to the clipping point when you expose and without introducing noise after the fact, in a filter (ND) that many landscape photographers use anyway for long exposure. Plus, better looking color and tonality.
 
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
 
Bear in mind as well that most modern sensors have a "native" ISO around 800 or more when still photographers at least have traditionally worked with 100 or even 64. 200 is fast.
 
Shane Betts said:
Bear in mind as well that most modern sensors have a "native" ISO around 800 or more when still photographers at least have traditionally worked with 100 or even 64. 200 is fast.
Click to expand...

ISO800 is an invented figure to give cinematographers some kind of reference and a more balanced DR distribution around mid grey. And it is pushed rating which often leads to underexposure.

Cameras like Red and Alexa benefit from lower rating, like ISO 640, just like R1 M gave better results at 250 not official 320, which although redistributes available DR towards the right, gives an overal thicker negative and a healthier signal.

Which is why Venice is rated at 500, with stated 6 stops above and 9 below mid grey. Although the sensor can be pushed more and they could have easily gone with stating ISO1000 for 7/8.
 
Isn't this how Arri achieves their color superiority? Sacrifice low light for better color reproduction with a stronger CFA.
 
You must log in or register to reply here.
Back
Top