X-Pro1 all-green mono conversion?

[Dante_Stella]

Ok, looking at the X-Pro1 mosaic pattern, has anyone else thought that if it only read green pixels and used the red and blue to recover blown highlights, it would make a really killer 7mp mono camera (minimum)?

When you replicate this pattern, you see that you have 2x2 green fields surrounded by RGB "threads." Though my Fourier transformations are long behind me, I would tend to think that it should be relatively easy to decode this in a way that (when appropriately downsampled) would create a top-flight mono image.

The beauty is that you should be able to do this even after the fact - though you would essentially be applying a green filter to everything. I guess you could do the same thing with a standard Bayer pattern.

Dante

 

[Keith]

I never mastered the Rubik's Cube sorry! 😀

 

[jsirevaag]

So, you want a Leica M9 Monochrom like I do! Silliness aside, it does seem that such an idea would work although you would be stuck (more or less) with a single contrast rendering. An interesting idea.

 

[semilog]

I'd not be surprised if this is what's already being done in-camera.

 

[rxmd]

Should be easy to simulate in Photoshop using the channel mixer (which is how better-quality BW conversion has been done for ages).

 

[dougi]

Should be easy to simulate in Photoshop using the channel mixer (which is how better-quality BW conversion has been done for ages).

What Dante is proposing is probably a bit different, unless Photoshop is channel mixing the pixel sites before de-mosaicing. I assume photoshop is mixing channels post de-mosaicing?

Ie create a green scale image, with a new de-mosaicing algorithm that uses the red and blue photo-site to assist in highlight recovery of the green sites only.

 

[nikoperugia]

Very intriguing. If you know your way around a computer (which I suppose you do), have a look at dcraw (or some of the modified sources available on the web, like libraw). There is one option to output 4 tiff files containing all the pixels -before- demosaicing, one for each channel. Assuming it works correctly for the relatively new xpro files, you only need to figure out how to match the channels back together.

 

[rxmd]

What Dante is proposing is probably a bit different, unless Photoshop is channel mixing the pixel sites before de-mosaicing. I assume photoshop is mixing channels post de-mosaicing?

Ie create a green scale image, with a new de-mosaicing algorithm that uses the red and blue photo-site to assist in highlight recovery of the green sites only.

Yeah, I was mainly referring to his last paragraph about doing this even after the fact (while essentially applying a green filter to everything). Doing it post-demosaicing you throw away 1/3 of the resolution, of course, mapping three possible pixels to two.

If you have access to something like dcraw, it gets easier; work in 16 bit, take the green channel for your picture, and where it is maxed out look at the red and blue channels and see if you can squeeze out an extra bit or two of dynamic range. I think that is what Dante is proposing.

Now the next interesting question is how difficult it is to rip out the colour filter. Seeing that there are homemade IR conversions of DSLRs, maybe there'll be black-and-white conversions in the future, too.

 

[Dante_Stella]

Right - I am proposing pulling the green channel out before Bayer decoding and using the other channels to reconstruct highlights if necessary.

Dante

 

[BobYIL]

Ok, looking at the X-Pro1 mosaic pattern, has anyone else thought that if it only read green pixels and used the red and blue to recover blown highlights, it would make a really killer 7mp mono camera (minimum)?






The beauty is that you should be able to do this even after the fact - though you would essentially be applying a green filter to everything. I guess you could do the same thing with a standard Bayer pattern.

Dante

- It'll again be a 16.3MP camera; mono or color..There is no binning there.

- DR may suffer a stop once a green filter is applied (The DR of an RGB sensor is the sum of the individual DRs of each color not overlapping onto each other as they have different responses through different quantum efficiencies..)

- Green filter will not turn R or B into color-blind but into brown and turquoise; with different frequency responses this time.

- The highest DR of any sensor is the one obtained through combining the individual ones coming from R, G and B.. they do complement each other; monochromatic uses only the overlapped zone.

 

[Macmook]

Wow, my head hurts. But even following this thread at a superficial level has increased my understanding in several areas.

Thanks

 

[Dante_Stella]

Bob:

I meant that the green filters would be the ones on the sensor. I wasn't suggesting doubling up by putting one on the lens. But that does raise an interesting point, since you'd probably want to use contrast filters of other colors.

Dante

- It'll again be a 16.3MP camera; mono or color..There is no binning there.

- DR may suffer a stop once a green filter is applied (The DR of an RGB sensor is the sum of the individual DRs of each color not overlapping onto each other as they have different responses through different quantum efficiencies..)

- Green filter will not turn R or B into color-blind but into brown and turquoise; with different frequency responses this time.

- The highest DR of any sensor is the one obtained through combining the individual ones coming from R, G and B.. they do complement each other; monochromatic uses only the overlapped zone.