Question about raw files and monochrome conversions

[JMC]

I'm definitely no expert on raw images, or even camera sensors, but found myself wondering this - with my monochrome camera conversion which has the Bayer filter removed, is it possible to increase the effective resolution of the camera? Not sure whether this makes sense, so I tried to make a diagram to explain what I mean;

With a red, 2 greens and blue going to make up each pixel, can these now be considered individually, and a 50Mp camera be made in to a 200Mp monochrome camera? I'd welcome any thoughts, and advice, even if it's 'stay off the egg nog' (although I'd prefer not to unless its absolutely necessary).

[bobfriedman]

yes it makes sense but with different arithmetic... if you have a 36mpx camera, 9Mpx Red, 9Mpx Blue and 18 Mpx for Green... so for some colors you get 18Mpx with others you get 9Mpx .. with a monochrome camera you get the full 36Mpx for all.

 

An initial comparison (next time with equal ISO).. processed identically in DarkTable, some level/contrast enhancement, demosiac turned off, matrix exposure, identical lighting D55

 

Nikon D800M ,Zeiss Apo Sonnar T* f/2 135mm

1s f/5.6 at 135.0mm iso100

http://www.pbase.com/bobfriedman/image/166750830/original.jpg

 

Nikon D800 Stock ,Zeiss Apo Sonnar T* f/2 135mm

0.62s f/5.6 at 135.0mm iso160

http://www.pbase.com/bobfriedman/image/166750834/original.jpg

 

100% crop center

 

Nikon D800M ,Zeiss Apo Sonnar T* f/2 135mm

1s f/5.6 at 135.0mm iso100

http://www.pbase.com/bobfriedman/image/166750845/original.jpg

 

Nikon D800 Stock ,Zeiss Apo Sonnar T* f/2 135mm

0.62s f/5.6 at 135.0mm iso160

http://www.pbase.com/bobfriedman/image/166750843/original.jpg

 

 

300% crop Monochrome Conversion

 

http://www.pbase.com/bobfriedman/image/166759206/original.jpg

[Alaun]

I think it is more complicated, e.g.:

 

If you have a surface with different colours, but all colours with the same grey level, you will get just

a grey picture without the Bayer-lenses, but you get the structures with the Bayer lenses.

[JMC]

Thanks Bob, that makes perfect sense. Dan at MaxMax mentioned about DarkTable and I think it's time for me to start using it - up until now I've not been doing anything special with my monochrome images in terms of processing, so I wont getting the best out of them.

 

Alaun, yes I get that, that's like monochrome film pictures. With the monochrome camera I use different filters on the lens to emphasize or tone down different coloured parts of the image as I would with B&W film.

[Cadmium]

Why does the black 'frame' not show any pattern? I notices also that the darker upper right square doesn't show the pattern as clearly as the other lighter squares.

I looked very close and I see no pattern on the darker frame.

[bobfriedman]

Why does the black 'frame' not show any pattern? I notices also that the darker upper right square doesn't show the pattern as clearly as the other lighter squares.

I looked very close and I see no pattern on the darker frame.

 

well black is the absence of color.. my guess for the upper square is that it has more contributing colors (RGB) than the others, but the pattern is there.

[bobfriedman]

attached the stock D800 color image for comparison

[bobfriedman]

Thanks Bob, that makes perfect sense. Dan at MaxMax mentioned about DarkTable and I think it's time for me to start using it

 

you need a processor that will turn of the demosaic operation.. DarkTable and AccuRaw Monochrome both do this.. DarkTable is open source and free to download.

 

By the way, Dan at maxmax.com does all my conversions.. highly recommend him.

[Andrea B.]

Assume a 36 MP camera.

 

Jonathan, the color demosaicing is more complex than what you have shown in your illustration. For any single photosite several overlapping 2x2 grids are evaluated to assign a colour to one photosite ("pixel"). So you really are getting "fairly close" to (my lame phrasing there) 36 MP worth of data from a 36 MP Bayered sensor. [Must also say that demosaicing algorithms abound. So which converter being used plays a big role here.]

 

Yes, you can make use of each pixel (if no demosaicing) in the monochrome sensor file, but you still only have 36 of them. However, without the demosaic blending which causes little artifacts and "blending" errors you get a slightly wider tonal range. And you have less noise because of faster exposures after the Bayer filter is removed. So you are usually going to see a more detailed (resolved) image - nothing dramatic like "4 times better", but noticeable.

 

 

^{Edit [corrected a tortured sentence structure so hoping for more clarity]}

[Andy Perrin]

We discussed this issue once before, Andrea, (you and I) with reference to whether one can get better resolution from >850nm IR images where the Bayer is fully transparent. I should go hunt down that thread.

 

ETA: here it is!

http://www.ultravioletphotography.com/content/index.php/topic/2034-sharpening-images/page__view__findpost__p__14330

 

[Andrea B.]

yep. I suppose the thing is is that the IR light would still be going through an extra dye layer even if the Bayer dyes are not being photonically affected so you would get potentially more noise due to less light gathered? Therefore I'm thinking that a true monochrome sensor would probably still come out a bit ahead?

 

I've been wondering all evening if a monochrome image file shouldn't probably be demosaiced anyway -- for better tonal gradients? If you just made a 1-1 pixel map wouldn't it be rather "jumpy" or pixel-ly looking? (Do I have any idea at all how to phrase what I'm trying to say? Nope.)

[OlDoinyo]

To gain access to the value returned by each subpixel, you would need non-stock camera firmware; the stock firmware combines the values and returns a matrix of composite values.

[Andy Perrin]

OlDoinyo- Andrea’s Raw Digger seems to return the values before demosaicing. If you click the link I put above there is an example.

See: http://www.ultravioletphotography.com/content/index.php/topic/2034-sharpening-images/page__view__findpost__p__14360

 

Andrea- yeah, the microlenses are still on there in the IR case even though the dyes are transparent to IR so I agree, it’s not exactly the same and there may be some losses. But I bet you can still extract the extra resolution.