How Can Digital Cameras Possibly Capture Violet Light?

[peter.ell]

First, I am curious about why CCDs need color filters in the first place. It is said that they are necessary because digital camera sensors can only sense intensity of light, but not wavelength (color), which makes sense initially... except that CCD sensors operate by utilizing the photoelectric effect, right? And isn't the photoelectric effect wavelength dependent? If so, than why would digital camera sensors require the use of color filters over the pixels of the sensors in order to detect and differentiate between various colors?

With that out of the way, here's the real question:
I know that digital camera sensors capture only three colors per pixel, red, green, and blue just like the corresponding pixels on any computer monitor. Purple is not a spectral color, and so it can be easily captured by a camera sensor and displayed on a screen by using red and blue pixels. But violet is a spectral color, having its own wavelength beyond that of blue. If a digital camera takes a picture of something that is violet, theoretically two things should occur due to the lack of the sensor's ability to pick up and distinguish violet from blue: violet objects would look black because the sensor is not able to respond to violet light, or violet objects would look blue because the sensor is picking up the violet light with the blue-sensitive pixels. Yet this does not seem to be the case.

So what's going on to allow digital cameras to capture violet light in a way that allows displays to represent it?

Thank you so much!

 

[davenn]

amused with your comments on purple and violet...
as they are commonly used as one in the same ... one school teaches violet, one teaches purple

to your ending comments ... you will find that most CCD chips have a VERY wide bandwidth, from IR through visible to UV, so in my understanding they would have no problem registering violet (purple)
in practical use from one extreme to another ... security cameras make great use of the CCD's IR sensitivity abilities for nighttime use. On the other hand I have fitted on my hi priced digital DSLR camera an ultra violet filter.

Dave

 

[sophiecentaur]

Your intial statement about digital imaging devices having Red Green and Blue sensors is actually not correct. Yes - with red green and blue phosphors you can reproduce visible colours on a TV screen but the three analysis curves in your eyes and in good colour camera systems are, in fact, very broad band. One has a peak at long wavelengths (reddISH), one at medium wavelengths (greenISH) and one at the shorter wavelengths (blueISH). To 'see' a colour accurately, you need an output from all three sensors so they each need to be sensitive over virtually the whole visible range. If the analyses were just narrow band, you would only 'see' red, green or blue - plus some crude complementary colours.
There have been threads about this recently where it's been discussed at length but I need to make the point that accurate Analysis and Synthesis of this subjective quantity are very different and should not be confused.

A UV filter is very handy for photos at high altitude, in particular, where the camera will 'see' UV and reproduce this as blue when displayed. It gives a serious blue caste on some pictures which is virtually impossible to eliminate with software because the damage has already been done in the sensors, which don't distinguish between the invisible and the visible - the 'green' sensor having 'given up' by that time.

 

[0xDEADBEEF]

Two parts to your question:
1) A CCD pixel has a response curve for different wavelengths. Typically broader than what the eye can detect. This means a certain amount of blue light will lead to a certain voltage (charge actually) the same amount of red light will lead to another voltage. So if all the light would have the same intensity you could tell colours by the voltage, but since the real world has high and low intensity light the colours cannot be distinguished.

2) The Eye has three types of sensor cells for colour vision. Red,Green and Blue ones. The red type of sensor cell has another small response in the small wavelength region, this is why the smallest wavelengths appear reddish (purple) again. In order to make a picture appear right on a screen you just need to send the right intensity of the three colours to stimulate the sensor cells, it doesn't really matter what the exact wavelength combination is in the real world. So ideally your ccd has three types of pixels that have exactly the same response as the cells in your eye, so you can tell your screen to set the three colours accordingly. Unfortunately this never quite works out, which is the reason why there is something called white balance that needs to be adjusted for digital images.

 

[sophiecentaur]

It's very complicated. In order to get a large gamut of displayable colours on a display, you need to have bright phosphors which are as near to a spectral colour as possible. The display can give any colour 'inside' the triangle of the three phosphors when plotted on the so-called Chromaticity diagram. Each camera and each display may have different analyses and phosphors and the information (the RGB values) needs to be corrected so that, in principle, any camera can get good colour reproduction on any display. And remember - it's all totally subjective. The 'non-spectral' colours that we see are combinations of reds and blues with less greens. We give them names but they can't be shown on a spectrometer as single wavelengths.
Go into any TV store and see the variety of colours on a row of different makes of TV.

But at least Digital cameras have a better chance of good colour fidelity than you could ever get with colour film. (Far more knobs to twiddle to get the right answer.)

For good colour printing, they use six or more dyes / inks and, very often, special 'spot' colours which your average inkjet printer just doesn't have. I am waiting for a display to come out with some special extra phosphors in it to help with the 'flesh tones'. But perhaps it isn't necessary.

 

[olivermsun]

...ideally your ccd has three types of pixels that have exactly the same response as the cells in your eye, so you can tell your screen to set the three colours accordingly. Unfortunately this never quite works out, which is the reason why there is something called white balance that needs to be adjusted for digital images.

Not just that, but when you view objects under different light sources (which aren't always "white"), human vision attempts to compensate for it. You still see an orange as an orange under poor lighting -- even if it looks more like a rotten tomato. Your camera can only guess using based on what it "sees" through the lens (or using an external sensor/setting).

Has the OP ever considered why black and white film photographers used to use various colored filters for routine shots?

 

[sophiecentaur]

...and that Panchromatic Film was such a big deal when it arrived.

 

[phyzguy]

By the way, digital cameras typically don't use CCD's because they are too expensive. They typically use digital image sensors, which are simply silicon diodes optimized to be sensitive to light.

 

[russ_watters]

cmos

 

[Skaperen]

I know that digital camera sensors capture only three colors per pixel, red, green, and blue just like the corresponding pixels on any computer monitor. Purple is not a spectral color, and so it can be easily captured by a camera sensor and displayed on a screen by using red and blue pixels. But violet is a spectral color, having its own wavelength beyond that of blue. If a digital camera takes a picture of something that is violet, theoretically two things should occur due to the lack of the sensor's ability to pick up and distinguish violet from blue: violet objects would look black because the sensor is not able to respond to violet light, or violet objects would look blue because the sensor is picking up the violet light with the blue-sensitive pixels. Yet this does not seem to be the case.

So what's going on to allow digital cameras to capture violet light in a way that allows displays to represent it?

Well, just how is it actually being represented? Are you seeing something that is a mix of red wavelength and blue wavelength (what you call purple)? Or are you seeing a violet wavelength that is shorter than the blue wavelength?

It will depend on the camera and/or sensor. Most of the image formats represent/store just red, green, and blue, or some mathematical mix of them (typically a monochrome mix and 2 difference channels). Displays generally have just these three primary colors. I can't see any point in having infrared and/or ultraviolet in an imaging display. So you are going to be limited to seeing it in these three colors

I'm sure every consumer camera has no separate channel for ultraviolet (or infrared, even though I wish they did). What probably happens, which could vary by camera manufacturer/model, is that the color filters on the sensor surface react to UV in some way that might make it look purple, blue, or black. If purple, then the red filter is passing some near UV through (along with the blue filter), and giving enough red at UV objects to look purple. If blue, then the the blue filter is passing near UV while red is not. And if black, then all filters, or some other general filter not specific to the separation filters, is/are blocking the near UV.

Also, things that directly look purple even though they have a wavelength on the far short edge of blue, near UV, might be doing so not because your eyes have a special means to detect that wavelength, but instead because there is some minimal sensitivity to that very short wavelength by the red in the cones (along with the blue in the rods). So it would look purplish. Some cameras could do just the same. Or not.

 

[sophiecentaur]

You will only see a colour on a display that is within the traingle of the three primary phosphors on the display (whatever analysis you care to use). Violet is outside this triangle so it just cannot be displayed.

When you see 'purple' on a TV display, you are being shown blue (phosphor) with some red as well. This just cannot be made to match a spectral violet. The colour described as Magenta is produced with equal values of R and B signal and no G. Adding G will desaturate this deep magenta and the result may be referred to as violet - by people who can't actually recall what violet actually looks like.
The fact is that, apart from when we see a rainbow against a dark cloud background, there are very few instances when we actually see a true (spectral) violet / indigo. Some gas discharges couldalso produce a spectral violet but I don't know which; they could, maybe look really impressive in nightclub 'neon'-type displays.
It is possible to make a violet pigment (subtractive mixing) but it would tend to be very low luminance.

@skaperen
I don't know of any electronic colour system that uses an extra colour channel. Until someone decides to produce a four phosphor display which could show your violets then there is no point in detecting / measuring them. As it is, far-blue sensitivity is more of a nuisance than anything in outdoor scenes because of haze effects. I really wonder, however, why the colour sensors aren't deliberately filtered to eliminate the UV problem. I have looked for a suitable Cokin filter but haven't come across one yet.

 

[sophiecentaur]

By the way, digital cameras typically don't use CCD's because they are too expensive. They typically use digital image sensors, which are simply silicon diodes optimized to be sensitive to light.

CCD or CMOS, depending on the camera manufacturer and cost / size of the sensor array. Look it up- and read some Camera Mags.

 

[jsgruszynski]

> isn't the photoelectric effect wavelength dependent

Actually this is what Einstein won his Nobel prize for (not relativity) and is a key aspect of quantum mechanics.

http://en.wikipedia.org/wiki/Photoelectric_effect

The simple answer: no, not once the photon energy is above the work function energy or in the case of semiconductor photodetection, the band gap energy. The difference is converted into electron kinetic energy (which become lattice phonons in a semiconductor) but the current (number of electrons per photon) doesn't change.

 

[carlgrace]

CCD or CMOS, depending on the camera manufacturer and cost / size of the sensor array. Look it up- and read some Camera Mags.

The vast majority of camera manufacturers are using CMOS sensors these days, even for SLR cameras.

There are probably a few super high-end cameras using CCDs but the sun is setting on that technology for consumer devices.

 

[sophiecentaur]

But what has this to do with the analysis and syntheses of Violet and Indigo?

 

[Bob S]

The colors in a rainbow vary uniformly from about 400 nm (violet) to 650 nm (red). Purple is not a rainbow color. See visible spectrum with names at

http://eosweb.larc.nasa.gov/EDDOCS/Wavelengths_for_Colors.html#violet

Go photograph a real rainbow.

 

[sophiecentaur]

It is true that there is a continuous range of wavelengths, corresponding to the spectral colours.
It is also true that, if you photograph a rainbow, you will not 'see' violet and indigo on your TV display where those bands should be. The screen just cannot produce them with the normal phosphors. What you do see is a purplish colour that is on the magenta side of blue - cos that's the best the TV can do. And we have 'learned' that is the colour violet.

 

[Bob S]

Purple, although not a natural single wavelength color, is easily generated by combining red and blue phosphors. For this reason, purple is sometimes called "minus green".

Bob S

 

[sophiecentaur]

Magenta is what is called 'minus green', actually - because it is white without the green channel.. But none of the red/blue phosphor combinations is violet or indigo.