A different monochrome method

[michaeladawson]

There have been threads here that talk about the wish that Fuji release a monochrome camera. The standard response is pretty much always that it is a niche product and wouldn't sell enough to be profitable.

There seem to be two major advantages of a monochrome camera that get pointed out. 1) There is no interpolation of pixels to arrive at a final image. In a color digital camera the value for each pixel is derived by interpolating it with the other pixels around it. The result is a less accurate image. 2) Because there is no CFA all pixels capture photons from all wavelengths. More photons is better, better in lower light.

Instead of coming out with a true monochrome camera why doesn't Fuji opt for monochrome mode similar to what the Pixii camera does? Pixii engineers know what the absorption values are for the different RGB filters on the CFA. It is a simple matter to calculate what the full spectrum capture of a pixel would have been if there had been no color filter. For example, if the red filter blocks 66.6% of "non-red" wavelengths then you can multiply the captured value of the red CFA pixels by 3 to get the non-filtered value.

I know, I know. You are still calculating a value and therefore what is recorded isn't really the "true and accurate" value. But I have to believe that it would be a really close value that would not really be discernible in the resulting photo. On top of that, B&W film stocks are well known to have spectral responses. So any inaccuracies with the Pixii method could be thought of as its own spectral response curve.

In any event, this would allow for not having to interpolate the pixels to produce a final image. This should result in sharper images. The Pixii method, while not perfectly accurate, seems like it would produce images a cut above standard interpolations. This seems like a good option for B&W mode on a camera without resorting to a monochrome camera.

In fact, this method would not have to be in the camera at all. Since it is a demosaic function it could be provided as an option in any raw image processing software. All that is needed is an accurate CFA table for each supported camera.

 

[Jeff Biscuits]

It is a simple matter to calculate what the full spectrum capture of a pixel would have been if there had been no color filter. For example, if the red filter blocks 66.6% of "non-red" wavelengths then you can multiply the captured value of the red CFA pixels by 3 to get the non-filtered value.

Well, it’s more complicated than that. If it’s only red light that’s falling on that photosite then you’d be mistaken to compensate for green and blue, since there is none of either. Obviously, there are adjacent green and blue photosites to infer information from, but fundamentally you’re back to interpolation.

I know, I know. You are still calculating a value and therefore what is recorded isn't really the "true and accurate" value. But I have to believe that it would be a really close value that would not really be discernible in the resulting photo.

Indeed. The question is really: is it almost as good as a monochrome sensor, or is it only marginally better than a B&W conversion of an already demosaiced image?

On top of that, B&W film stocks are well known to have spectral responses. So any inaccuracies with the Pixii method could be thought of as its own spectral response curve.

Yes, and it allows that response to be configurable, as with B&W conversion of colour images.

In any event, this would allow for not having to interpolate the pixels to produce a final image. This should result in sharper images. The Pixii method, while not perfectly accurate, seems like it would produce images a cut above standard interpolations. This seems like a good option for B&W mode on a camera without resorting to a monochrome camera.

In fact, this method would not have to be in the camera at all. Since it is a demosaic function it could be provided as an option in any raw image processing software. All that is needed is an accurate CFA table for each supported camera.

And the fact that no-one seems to have done it perhaps tells us how effective it really is.

To be honest I think the Pixii Method seems like a convenient way to commit to B&W as a photographic process—which is fine—but I’m sceptical of its technical merits. (By which I mean I’d love to see evidence, not that I’m dismissive of it.)

 

[michaeladawson]

It is a simple matter to calculate what the full spectrum capture of a pixel would have been if there had been no color filter. For example, if the red filter blocks 66.6% of "non-red" wavelengths then you can multiply the captured value of the red CFA pixels by 3 to get the non-filtered value.

Well, it’s more complicated than that. If it’s only red light that’s falling on that photosite then you’d be mistaken to compensate for green and blue, since there is none of either. Obviously, there are adjacent green and blue photosites to infer information from, but fundamentally you’re back to interpolation.

I know you are right in that it is more complicated than I describe. And in fact I may be describing the Pixii method completely wrong. I wonder if they publish a more technical description of their method than I was able to read in their marketing materials.

 

[Banderras]

Instead of coming out with a true monochrome camera why doesn't Fuji opt for monochrome mode similar to what the Pixii camera does? Pixii engineers know what the absorption values are for the different RGB filters on the CFA. It is a simple matter to calculate what the full spectrum capture of a pixel would have been if there had been no color filter. For example, if the red filter blocks 66.6% of "non-red" wavelengths then you can multiply the captured value of the red CFA pixels by 3 to get the non-filtered value.

I have 0 knowledge about Pixii, but if that's their marketing material, then it's cold pressed snake oil. Color filter absorption values are not known constants, they are % attenuation of the signal.

I think, what you describe as a new method is exactly what is currently happening with the RGB -> BW conversion. First you use bayer interpolation to restore the original "full spectrum" pixel value, then you convert it to grayscale.

 

[saltydogstudios]

Instead of coming out with a true monochrome camera why doesn't Fuji opt for monochrome mode similar to what the Pixii camera does? Pixii engineers know what the absorption values are for the different RGB filters on the CFA. It is a simple matter to calculate what the full spectrum capture of a pixel would have been if there had been no color filter. For example, if the red filter blocks 66.6% of "non-red" wavelengths then you can multiply the captured value of the red CFA pixels by 3 to get the non-filtered value.

I have 0 knowledge about Pixii, but if that's their marketing material, then it's cold pressed snake oil. Color filter absorption values are not known constants, they are % attenuation of the signal.

I think, what you describe as a new method is exactly what is currently happening with the RGB -> BW conversion. First you use bayer interpolation to restore the original "full spectrum" pixel value, then you convert it to grayscale.

My question is - how difficult would it be to create an RGB + Monochrome sensor?

Our modern screens are all capable of both monochrome and color output - a bit of charge across each pixel changes the color.

Why couldn't sensors use the same technology in reverse? (for detecting light instead of projecting it)

Obviously there's a scale difference and potential reliability issues. But in theory, it could work.

--
"no one should have a camera that can't play Candy Crush Saga."
Ye olde instagram: https://www.instagram.com/sodiumstudio/ (will probably still be around after April 10th)

 

[Doppler9000]

There have been threads here that talk about the wish that Fuji release a monochrome camera. The standard response is pretty much always that it is a niche product and wouldn't sell enough to be profitable.

There seem to be two major advantages of a monochrome camera that get pointed out. 1) There is no interpolation of pixels to arrive at a final image. In a color digital camera the value for each pixel is derived by interpolating it with the other pixels around it. The result is a less accurate image. 2) Because there is no CFA all pixels capture photons from all wavelengths. More photons is better, better in lower light.

Instead of coming out with a true monochrome camera why doesn't Fuji opt for monochrome mode similar to what the Pixii camera does? Pixii engineers know what the absorption values are for the different RGB filters on the CFA. It is a simple matter to calculate what the full spectrum capture of a pixel would have been if there had been no color filter. For example, if the red filter blocks 66.6% of "non-red" wavelengths then you can multiply the captured value of the red CFA pixels by 3 to get the non-filtered value.

I know, I know. You are still calculating a value and therefore what is recorded isn't really the "true and accurate" value. But I have to believe that it would be a really close value that would not really be discernible in the resulting photo. On top of that, B&W film stocks are well known to have spectral responses. So any inaccuracies with the Pixii method could be thought of as its own spectral response curve.

In any event, this would allow for not having to interpolate the pixels to produce a final image. This should result in sharper images. The Pixii method, while not perfectly accurate, seems like it would produce images a cut above standard interpolations. This seems like a good option for B&W mode on a camera without resorting to a monochrome camera.

In fact, this method would not have to be in the camera at all. Since it is a demosaic function it could be provided as an option in any raw image processing software. All that is needed is an accurate CFA table for each supported camera.

I suspect the Pixii approach reflects the inability or unwillingness to order sufficient quantities of monochrome chips or some other scale issue.

Monochrome sensors that don’t have CFAs are more efficient and have a less-obstructed optical path compared to cameras that rely on post-processing monochrome approaches.

 

[JNR]

I suspect the Pixii approach reflects the inability or unwillingness to order sufficient quantities of monochrome chips or some other scale issue.

Monochrome sensors that don’t have CFAs are more efficient and have a less-obstructed optical path compared to cameras that rely on post-processing monochrome approaches.

I hate to ponder a concept based on a possible technology that is well beyond my overall understanding, but...

Wouldn't it be quite the trick if you could design a sensor that allowed you to flip the X-trans or Bayer array out of the way to allow for an actual monochrome sensor underneath?

In any event, Pentax/Ricoh should be proud that they've managed to prompt a relevant, modern tech conversation after all these years of treading water. Yes, Leica headed in that direction some time ago, but that had nothing to do with actual mass market potential.

 

[Doppler9000]

I suspect the Pixii approach reflects the inability or unwillingness to order sufficient quantities of monochrome chips or some other scale issue.

Monochrome sensors that don’t have CFAs are more efficient and have a less-obstructed optical path compared to cameras that rely on post-processing monochrome approaches.

I hate to ponder a concept based on a possible technology that is well beyond my overall understanding, but...

Wouldn't it be quite the trick if you could design a sensor that allowed you to flip the X-trans or Bayer array out of the way to allow for an actual monochrome sensor underneath?

I think the mechanicals required would be a manufacturing challenge, given the 3-4 um dimensions involved.

In any event, Pentax/Ricoh should be proud that they've managed to prompt a relevant, modern tech conversation after all these years of treading water. Yes, Leica headed in that direction some time ago, but that had nothing to do with actual mass market potential.

Pentax has always had the courage to swim upstream.

 

[Jeff Biscuits]

Wouldn't it be quite the trick if you could design a sensor that allowed you to flip the X-trans or Bayer array out of the way to allow for an actual monochrome sensor underneath?

I think the mechanicals required would be a manufacturing challenge, given the 3-4 um dimensions involved.

Indeed. It would be far more viable to take the Ricoh GXR approach and have interchangeable sensor units.

 

[JNR]

Wouldn't it be quite the trick if you could design a sensor that allowed you to flip the X-trans or Bayer array out of the way to allow for an actual monochrome sensor underneath?

I think the mechanicals required would be a manufacturing challenge, given the 3-4 um dimensions involved.

Indeed. It would be far more viable to take the Ricoh GXR approach and have interchangeable sensor units.

Agreed, but I'm not fond of the lens-attached-to-sensor concept, as what's left behind it doesn't add up to all that much and the deal gets really expensive fast when buying several lenses.

Now, if you could somehow be able to slip an interchangeable sensor into a perfectly registered slot from underneath behind the mount, that would be really something. You could offer various IR versions, as well. Pretty sure you'd have to live without IBIS though (as good photographers had endured for a few centuries until recently).

--
JNR

 

[Doppler9000]

Wouldn't it be quite the trick if you could design a sensor that allowed you to flip the X-trans or Bayer array out of the way to allow for an actual monochrome sensor underneath?

I think the mechanicals required would be a manufacturing challenge, given the 3-4 um dimensions involved.

Indeed. It would be far more viable to take the Ricoh GXR approach and have interchangeable sensor units.

Agreed, but I'm not fond of the lens-attached-to-sensor concept, as what's left behind it doesn't add up to all that much and the deal gets really expensive fast when buying several lenses.

Now, if you could somehow be able to slip an interchangeable sensor into a perfectly registered slot

The pixel pitch on 26 MP APS-C sensor is 3.76 um. 70 um is the thickness of a human hair.

from underneath behind the mount, that would be really something. You could offer various IR versions, as well. Pretty sure you'd have to live without IBIS though (as good photographers had endured for a few centuries until recently).

 

[michaeladawson]

Going back and looking at the Pixii site it is (purposefully) unclear with respect to how they are doing monochrome.

One HUGE part of what I didn't really appreciate the first time is that their claims for a much improved B&W image is that they compare what they do to a "typical camera" that captures 8-bit JPEG monochrome photos. Their big boast is that they capture 16-bit raw files.

They are quite vague on how they reverse the CFA filter colors. As someone suggested, it is quite possible that they really aren't doing anything more than interpolation of the neighboring colored pixels. I'm not saying they are, just that it really isn't clear what they are doing.

So perhaps my enthusiasm for their advertised B&W mode is a bit watered down at this point.

 

[Heritage Cameras]

Now, if you could somehow be able to slip an interchangeable sensor into a perfectly registered slot from underneath behind the mount, that would be really something. You could offer various IR versions, as well. Pretty sure you'd have to live without IBIS though (as good photographers had endured for a few centuries until recently).

There has been at least one camera with interchangeable filters in front of the sensor... the Mamiya ZD. This had a slot in the base through which two IR different filters ("sealed in a protective cartridge") could be inserted - one with a low pass filter and one without. Presumably they could have offered a plain glass without IR-blocking, but I don't believe they did.

https://www.dpreview.com/articles/3664677826/mamiyazd

I doubt if this system could be used for the Bayer filter or even whole sensors for the option of colour or true mono, but I have been surprised before!

 

[JNR]

Wouldn't it be quite the trick if you could design a sensor that allowed you to flip the X-trans or Bayer array out of the way to allow for an actual monochrome sensor underneath?

I think the mechanicals required would be a manufacturing challenge, given the 3-4 um dimensions involved.

Indeed. It would be far more viable to take the Ricoh GXR approach and have interchangeable sensor units.

Agreed, but I'm not fond of the lens-attached-to-sensor concept, as what's left behind it doesn't add up to all that much and the deal gets really expensive fast when buying several lenses.

Now, if you could somehow be able to slip an interchangeable sensor into a perfectly registered slot

The pixel pitch on 26 MP APS-C sensor is 3.76 um. 70 um is the thickness of a human hair.

from underneath behind the mount, that would be really something. You could offer various IR versions, as well. Pretty sure you'd have to live without IBIS though (as good photographers had endured for a few centuries until recently).

Of course, you would have the sensor mounted in a solid casing that would include contacts to body electronics. This really isn't a far-fetched idea, especially so if you could design precision internal rails to locking mechanism, and some kind of user casing that prevents the sensor from being touched or harshly bumped. As it is, I've done hundreds of sensor cleanings with no damage, so they are somewhat less fragile than commonly assumed.

 

[JNR]

Going back and looking at the Pixii site it is (purposefully) unclear with respect to how they are doing monochrome.

One HUGE part of what I didn't really appreciate the first time is that their claims for a much improved B&W image is that they compare what they do to a "typical camera" that captures 8-bit JPEG monochrome photos. Their big boast is that they capture 16-bit raw files.

They are quite vague on how they reverse the CFA filter colors. As someone suggested, it is quite possible that they really aren't doing anything more than interpolation of the neighboring colored pixels. I'm not saying they are, just that it really isn't clear what they are doing.

So perhaps my enthusiasm for their advertised B&W mode is a bit watered down at this point.

Your skepticism is well placed. Any type of unnecessary filter (array) in the path of light seems like it should be disqualifying. As it stands, regular color filter arrays are "only" reducing the potential monochrome light gathering by about 1.5 stops... so you're probably improving the image quality and shadow noise by an amount that is much closer to FF equivalent (but still a bit better) than the 33x44 MF color sensors vs. monochrome. Hope that makes sense as far as relative improvement comparisons are concerned.

Leica is still using older sensor technology for reasons I can't fathom, and that's why the economical Pentax offering is so impressive despite the DSLR and lens limitations involved.

 

[Doppler9000]

Going back and looking at the Pixii site it is (purposefully) unclear with respect to how they are doing monochrome.

One HUGE part of what I didn't really appreciate the first time is that their claims for a much improved B&W image is that they compare what they do to a "typical camera" that captures 8-bit JPEG monochrome photos. Their big boast is that they capture 16-bit raw files.

They are quite vague on how they reverse the CFA filter colors. As someone suggested, it is quite possible that they really aren't doing anything more than interpolation of the neighboring colored pixels. I'm not saying they are, just that it really isn't clear what they are doing.

So perhaps my enthusiasm for their advertised B&W mode is a bit watered down at this point.

Your skepticism is well placed. Any type of unnecessary filter (array) in the path of light seems like it should be disqualifying. As it stands, regular color filter arrays are "only" reducing the potential monochrome light gathering by about 1.5 stops... so you're probably improving the image quality and shadow noise by an amount that is much closer to FF equivalent (but still a bit better) than the 33x44 MF color sensors vs. monochrome. Hope that makes sense as far as relative improvement comparisons are concerned.

Here is a test of the Q2M vs the 100S.

Leica is still using older sensor technology for reasons I can't fathom, and that's why the economical Pentax offering is so impressive despite the DSLR and lens limitations involved.

 

[JNR]

Here is atest of the Q2M vs the 100S.

Helpful to those who primarily are interested in shooting Siemens star charts.

Leica is still using older sensor technology for reasons I can't fathom, and that's why the economical Pentax offering is so impressive despite the DSLR and lens limitations involved.

As for real-world shooters, I'm inclined to have more faith in this balanced analysis:

https://www.dpreview.com/forums/post/67080745

 

[Doppler9000]

Here is atest of the Q2M vs the 100S.

Helpful to those who primarily are interested in shooting Siemens star charts.

If you actually believe people like Jim Kasson spend hours testing using Siemens stars to help people interested in shooting Siemens stars, and you want to improve your understanding of fundamental optics principles, then I would recommend reading more of Jim’s website - it is a remarkable resource.

If you are just trolling, then troll away.

Leica is still using older sensor technology for reasons I can't fathom, and that's why the economical Pentax offering is so impressive despite the DSLR and lens limitations involved.

As for real-world shooters, I'm inclined to have more faith in this balanced analysis:

https://www.dpreview.com/forums/post/67080745

You don’t have to choose one analysis over another. The Siemens star analysis is helpful to real-world shooters who have even a rudimentary understanding of the principles it elucidates.

 

[Doppler9000]

Wouldn't it be quite the trick if you could design a sensor that allowed you to flip the X-trans or Bayer array out of the way to allow for an actual monochrome sensor underneath?

I think the mechanicals required would be a manufacturing challenge, given the 3-4 um dimensions involved.

Indeed. It would be far more viable to take the Ricoh GXR approach and have interchangeable sensor units.

Agreed, but I'm not fond of the lens-attached-to-sensor concept, as what's left behind it doesn't add up to all that much and the deal gets really expensive fast when buying several lenses.

Now, if you could somehow be able to slip an interchangeable sensor into a perfectly registered slot

The pixel pitch on 26 MP APS-C sensor is 3.76 um. 70 um is the thickness of a human hair.

from underneath behind the mount, that would be really something. You could offer various IR versions, as well. Pretty sure you'd have to live without IBIS though (as good photographers had endured for a few centuries until recently).

Of course, you would have the sensor mounted in a solid casing that would include contacts to body electronics. This really isn't a far-fetched idea, especially so if you could design precision internal rails to locking mechanism, and some kind of user casing that prevents the sensor from being touched or harshly bumped. As it is, I've done hundreds of sensor cleanings with no damage, so they are somewhat less fragile than commonly assumed.

I guess we have a different definition of “far-fetched”. Do you honestly believe your notions could be commercially viable? How much would this complex precision system cost compared to a mass-produced sensor? I would hate to work in the warranty service department when people bring in their cameras with interchangeable precision rail-sensor-casing issues.

 

[Jeff Biscuits]

Wouldn't it be quite the trick if you could design a sensor that allowed you to flip the X-trans or Bayer array out of the way to allow for an actual monochrome sensor underneath?

I think the mechanicals required would be a manufacturing challenge, given the 3-4 um dimensions involved.

Indeed. It would be far more viable to take the Ricoh GXR approach and have interchangeable sensor units.

Agreed, but I'm not fond of the lens-attached-to-sensor concept, as what's left behind it doesn't add up to all that much and the deal gets really expensive fast when buying several lenses.

I meant more like the GXR M-Mount module, which was a sensor plus a mount for a lens. This meant you had a body, a sensor, and a lens, all interchangeable with one another.

Which basically gives you this:

Now, if you could somehow be able to slip an interchangeable sensor into a perfectly registered slot from underneath behind the mount, that would be really something. You could offer various IR versions, as well. Pretty sure you'd have to live without IBIS though (as good photographers had endured for a few centuries until recently).