Do all CCDs/LCDs have tough time duplicating color of monochromatic light?

[NeoPTLD]

I set the white balance to "daylight" then photographed a projection of monochromatic 510nm blue-green light. (Lumileds calls it cyan)

This is what it looks like in picture in two different cameras. Looks
http://img90.imageshack.us/img90/5191/canon510nmgreenpi5.jpg

http://img146.imageshack.us/img146/7288/panasonic510nmgreenkm8.jpg

No shade in this progressive chart quite matches the real color
http://habitat.igc.org/faces/cyan-2.html

Why is this?

 

[Mark R]

All colour devices (be they sensors or displays) have a limited colour gamut (or range). The eyes, too, have a limited gamut and this means that there are problems with overlap.

The result is that there are certain colours that cannot be accurately recorded by sensors, or accurately displayed/printed. Professional graphics designers will try to limit their use of colour to those tones that can be accurately displayed and printed. This is one of the reasons why there are so many different colour systems (Pantone, Lab, sRGB, Adobe RGB, etc.)

Monochromatic light is a particular problem, because the colour sensitiviyt curves of most image sensors don't match the colour sensitivity curves of the human eye. This can produce marked colour shifts with monochromatic light, even though under white light the results are excellent.

As it is, most LCDs have a poor colour gamut (even compared to CRT). However, there are some very specialist LCDs using RRGGBB displays (using 2 different wavelengths of each colour) which have a very wide gamut - wide enough to easily cover all colours in the NTSC and Adobe RGB gamuts. (By comparison a typical LCD can only display about 70% of the NTSC or Adobe gamuts).

 

[Syppion]

It is also important to realize that since digital cameras rely on different pixels to detect different colors, it is not uncommon for color CCD manufacturers to double the number of green diodes (pixels) - resulting in a single "image pixel" may actually consisting of data read in from 1-blue, 1-red, and 2-green sensitive diodes. An example can be seen on page 10 of this spec-sheet. Of course each of these green diodes has a response curve associated with it which is typically calibrated (voltage to digital level) to attempt to recreate the response curve of the human-eye - not reality.

Why green? While there are four photopigments in the human eye, the combination of rods and cones makes us most sensitive to green - and thus the most sensitive to errors in green-balance. (I'll replace the links with better ones when I get a chance.)

This also gave rise to a discrepency in how many pixels a camera has - depending on if you were talking image pixels or CCD pixels (an early marketing ploy).


Edit: This is a bit better of a link, but still not the one I remember.

 

[CycloWizard]

Originally posted by: Syppion
It is also important to realize that since digital cameras rely on different pixels to detect different colors, it is not uncommon for color CCD manufacturers to double the number of green diodes (pixels) - resulting in a single "image pixel" may actually consisting of data read in from 1-blue, 1-red, and 2-green sensitive diodes. An example can be seen on page 10 of this spec-sheet. Of course each of these green diodes has a response curve associated with it which is typically calibrated (voltage to digital level) to attempt to recreate the response curve of the human-eye - not reality.

There is actually a new imager out that has detectors for each color on every pixel, though there is no commercial camera using it yet. A guy I work with here is building a camera around this imager and he's already received about 120 orders at $5k a piece, though the prototype isn't done yet. Of course, I can't recall the name of the imager, because that would be too helpful, but I'll see if I can remember it or ask him tomorrow.

 

[Calin]

Maybe the fact that the new LCD panels only have 6bits per pixel adds to the error

 

[xtknight]

They are on the chart in these marked places: http://xtknight.atothosting.com/images/cyan-r2.png

I just did a maximum Gaussian blur to obtain the RGB values for the Canon and the Panasonic since those images were a little "rough".

Of course, a monitor can't display the gamut that your eye can see. And, no type of digital device can have the accuracy either.

Here is a modern LCD's gamut on the CIE diagram: http://common.ziffdavisinternet.com/uti...t_image/12/0,1425,sz=1&i=123356,00.gif

8-bit/6-bit doesn't really affect the gamut that much. It's mainly a product of how white the backlight is. Color accuracy severely suffers on 6-bit displays.