Luminous Landscape Forum
Equipment & Techniques => Beginner's Questions => Topic started by: MarkJohnson on March 22, 2016, 01:57:55 pm
-
A naive, ignorant question: does the Bayer array over-represent green? The two green to one red and one blue sensors seems perhaps a crude ratio, unlikely to reflect precisely the differential retinal RGB sensitivities, albeit green sensitivity being greatest. (Maybe these differential sensitivities don't have to be that precisely reflected in sensor anatomy.) With my D800E and now D810 I have often found greens a little too saturated, too bright and/or too yellow, more or less fixable with adjustments under HSL of green and maybe, if overall image not overly affected, yellow. I mostly use LR, sometimes C1 & occasionally DxO Optics Pro. Display is calibrated and printer profiled. Can my experience be related to Bayer sensor structure?
-
I've never seen a newbie question quite like this! :D
-
A naive, ignorant question: does the Bayer array over-represent green? The two green to one red and one blue sensors seems perhaps a crude ratio, unlikely to reflect precisely the differential retinal RGB sensitivities, albeit green sensitivity being greatest.
Since the sensitivities of the eye is weighted towards green (which represent as you say the greatest spectral sensitivity) The Bayer array does NOT "over-represent" green...it represents an appropriate amount of spectral response. There's a reason why there are 2 greens, one red and one blue–it works out well for sensors and demosiacing.
-
If you have too much green cast you can use the tint slider in both LR and C1 to compensate for that before having to rely on HSL adjustments. You might also need to make profiles for your camera (like with DCamProf) to get more natural colors out of the box.
I think it was demonstrated somewhere else in this forum that final image colors depend more on camera profiles and demosaicing than CFAs or sensor tech, so if you are worried you are not getting the most out of your camera you can stop and enjoy the pictures you get.
-
I should have added, I use a ColorChecker Passport camera profile.
The tint slider will affect the whole image. What would be nice in C1 color editor, would be if, when selecting a colour and then selecting display of just that colour in the whole image, we could make that, in PS terms, a selection, then make that selection into a new adjustment layer, then fiddle with tint etc just in that layer.
Jeff, it might be interesting to see results from sensors with, say, 1.5, 1.75, 2.0, 2.25, 2.5 times as many green receptors as red & blue.
-
Mark,
With my D800E and now D810 I have often found greens a little too saturated, too bright and/or too yellow, more or less fixable with adjustments under HSL of green and maybe, if overall image not overly affected, yellow...Can my experience be related to Bayer sensor structure?
You may find the initial post on this (http://www.fredmiranda.com/forum/topic/1234124/2&year=2013#11744473) page of interest.
-
Thank you Rob, that is interesting. The August 2013 post you reference says that Nikon is over sensitive to yellow but has extremely good hue resolution of green. I wonder if the then new Sony 36 MP sensor of the D800/E was being taken into account.
After my previous posts, I reflected that perhaps the way I often do 'white balancing', especially with landscapes, could be relevant. Sometimes I use a WB target such as the ones on the CCP but often in post I move LR Temp either way by +/- 10 increments then fine tune by +/- 1 increments until there is no overall blue or yellow cast, then the same with Tint until there is no overall green or magenta cast. The resulting Temp & Tint values are not those I would have had from the WB target, but satisfy me overall, mostly. Perhaps my issues with green to some extent come from doing this. I'll do the appropriate tests and comparisons when I can.
-
Mark,
I wonder if the then new Sony 36 MP sensor of the D800/E was being taken into account.
Probably - the same poster mentioned one day later in this (http://www.fredmiranda.com/forum/topic/1234124/2&year=2013#11745379) post "...the absolute DR of the latest Sony sensors..." and "...D800...".
-
Rob, the way I said I often WB landscapes (made in the middle of the day) can put the image a little further down towards the yellow end of Temp than if I'd used a WB target. If my Nikon sensor is indeed a little over sensitive to yellow, this could be why I sometimes find my greens a little on the bright and yellow side, with a slightly inaccurate green hue. I understand the issues a bit better now, thanks.
-
If you have too much green cast you can use the tint slider in both LR and C1 to compensate for that before having to rely on HSL adjustments.
Just to be clear, the number of R, G & B photo sites has ZERO to do with any kind of "tint" on the demosiaced rendered raw image. The photo sites record the spectral results prior raw processing to render into RGB images.
-
Further reading reminds me the relatively greater retinal sensitivity to green is in terms of luminance. So I assume that a 2G1R1B colour filter array gives more and adequate data for separating luminance levels, that a 3G1R1B CFA would give inadequate R/B data and that a 1G1R1B CFA would give inadequate luminance discrimination. Is that the whole story?
-
I would think that the RGBG array, besides reflecting luminance sensitivities of human eyes (with IR and UV filter attached to the sensor), is just easier to make and easier to compute, hence, arrived on the scene early on. There are other ways to arrange arrays, Fuji uses a more complicated alternative called X-Trans, less prone to moire, among other things. Here's some explanations, some of which may be slightly above my pay grade (I am not an engineer). And then there's the Foveon sensor....
http://spectrum.ieee.org/tech-talk/consumer-electronics/gadgets/ces-2012-fujifilms-new-cmos-sensor-uses-novel-color-filter-array
http://www.quadibloc.com/other/cfaint.htm
https://en.wikipedia.org/wiki/Color_filter_array
-
I should have added, I use a ColorChecker Passport camera profile.
The tint slider will affect the whole image. What would be nice in C1 color editor, would be if, when selecting a colour and then selecting display of just that colour in the whole image, we could make that, in PS terms, a selection, then make that selection into a new adjustment layer, then fiddle with tint etc just in that layer.
Jeff, it might be interesting to see results from sensors with, say, 1.5, 1.75, 2.0, 2.25, 2.5 times as many green receptors as red & blue.
You can use the color selection adjustment filter to do just that. You can narrow it in or spread it out from a color to affect the preferred range.
-
Yes NP, and the more complex Fuji array produces beautiful colours, at least in JPEG.
Yes too SZR, C1 Advanced Color Editor can change the colour range. The separate Color Balance Master tool does other interesting things, which if also doable in the Advanced Color Editor, would to a greater extent equate with a PS-style separate, totally editable adjustment layer (I think).
Thanks both for your thoughts on this.
-
A naive, ignorant question: does the Bayer array over-represent green?
In a word, no.
When RGB components are converted to relative luminance in colorimetric spaces, the following formula is used:
Y = 0.2126 * R + 0.7152 * G + 0.0722 * B
As you can see, human eye is more than twice as sensitive to green as to any of the other two components. So RGBG matrices are well-justified.
-
Interesting, thanks. What does Y represent in that equation (if you'll excuse my ongoing ignorance)? Do you know what physiological data forms the basis for the equation?
-
What does Y represent in that equation? Do you know what physiological data forms the basis for the equation?
It is perceived luminance.
Meaning that if, say, N photons hit blue-sensitive cone cell (S-cone) in human eye's retina, and then another N photons hit green-sensitive cell (M-cone), then the latter signal will be perceived by the eye 0.7152/0.0722 == 9.9 times brighter. Conversely, about 9.9 times more photons have to hit S-cone than M-cone for the eye to perceive both signals as equally bright.
All this have big implications for B&W (or, more exactly, grayscale) photo conversions. For the B&W image to look more natural, RGB components of the original color picture's pixels have to be converted into single "brightness" value for each pixel using formula close to the one for Y I mentioned above.
-
OK, thanks again. Googling, I see it comes from (very complex) colour space and gamma theory and the perceptual observations on which they are based, rather than retinal physiology as such.