Why IR shows up as purple on cheap cameras?

[fubar.gr]

If you point a cheap phone camera at an IR source, such as a TV remote control, you'll see a blu-ish or purple-ish light.



Actually any camera sensor can pick up IR, but more expensive cameras have lens with IR blocking filters.

But why purple? What is the process that goes on between the imaging sensor and the screen that basically changes a color to another?

[tec5c]

Essentially the spectral response for each red, green, blue pixel in a camera doesn't exactly match the spectral response of the receptors in your eyes. What you see through the camera is a mixture of mostly red and blue wavelengths with a slight touch of green.

[Gyro]

The optical filters on the individual pixels of the camera sensor don't work as expected when so far out of the visible band - there is no red, green or blue output from the IR led, but all of the sensor pixels are being activated in an 'unspecified' ratio (basically, somewhere near white). Presumably a better camera would have a better IR stop filter to reduce these out-of-band wavelengths getting to the sensor.

[tooki]

If you point a cheap phone camera at an IR source, such as a TV remote control, you'll see a blu-ish or purple-ish light.

Actually any camera sensor can pick up IR, but more expensive cameras have lens with IR blocking filters.

But why purple? What is the process that goes on between the imaging sensor and the screen that basically changes a color to another?

Semiconductor image sensors (CCD and CMOS) are intrinsically extremely sensitive to IR, and progressively less sensitive through to blue and then UV. (The opposite of silver halide film, by the way!) So we put IR filters over the sensors to block the IR, which would otherwise significantly compete with the sensitivity in the visible light spectrum. The fact that residual IR makes it through is simply a property of the imperfect filters we have available. That it looks purplish, as opposed to pure white, has to do with the additional filtering of the RGB filters of the Bayer pattern over each pixel (presumably, the green filter removes IR slightly more than the red and blue filters, giving the purple tinge).

It is untrue that only cheap cameras exhibit this effect. It affects all cameras with semiconductor sensors. If you didn't put in an IR filter over the sensor, the color of the image would be unusably wrong, and potentially the focus as well, since IR focuses at a slightly different distance than visible light. Any differences in IR sensitivity have to do with the characteristics of the components used.

The upper left quadrant of this picture is what a DSLR with its IR filter removed produces, in auto white balance. (The other 3 are with various filters and white balance settings.)



FYI, in security cameras and camcorders with an IR-illuminated night vision mode, you'll always hear a clunk when activating that mode, as the IR filter is physically removed from the optical path.

Essentially the spectral response for each red, green, blue pixel in a camera doesn't exactly match the spectral response of the receptors in your eyes. What you see through the camera is a mixture of mostly red and blue wavelengths with a slight touch of green.

Huh? Human vision is most sensitive to green, which is why the typical RGB Bayer pattern* devotes twice as many pixels in a cell to green as to red and blue:
RG
GB

This reduces noise in the green spectrum where we are most sensitive to it.

Semiconductor sensors are more sensitive to low wavelengths (red/IR) than longer ones (green/blue/UV), so the color filters and the software have to try and balance the relative sensitivity to produce a faithful (to human eyes) reproduction.

The optical filters on the individual pixels of the camera sensor don't work as expected when so far out of the visible band - there is no red, green or blue output from the IR led, but all of the sensor pixels are being activated in an 'unspecified' ratio (basically, somewhere near white). Presumably a better camera would have a better IR stop filter to reduce these out-of-band wavelengths getting to the sensor.

Exactly. The quality of the IR filter and the color filters, not their presence, which is mandatory, is what defines how an IR LED appears.


*Alternative Bayer patterns have been toyed with in the past, mostly by Sony, including CMYG (cyan/magenta/yellow/green) and CMYE (cyan/magenta/yellow/emerald). The idea was to have noise in the sensor be spread across the RGB output pixels.

[Cliff Matthews]

@Tooki - Thanks, great info! I always thought that "clunk" was some relay used to turn on the ring of IR LED's (although I didn't recall hearing the sound and some of those smaller "bullet" cams).

[SeanB]

Smaller cameras and the cheap ones simply omot the IR filter, and rely on the sensor response to visible light being enough to swamp the IR image. That is why on an overcast day those cameras produce reddish foliage, or in twilight, where the IR reflected from the leaves is higher in intensity than the green light they also reflect.

[fubar.gr]

Thanks guys! Really great info

[tooki]

Smaller cameras and the cheap ones simply omot the IR filter, and rely on the sensor response to visible light being enough to swamp the IR image. That is why on an overcast day those cameras produce reddish foliage, or in twilight, where the IR reflected from the leaves is higher in intensity than the green light they also reflect.

I find this to be highly unlikely. Semiconductor sensors are so sensitive to IR that I don't think visible light would ever "swamp" the IR image. Cheaper IR filter, maybe. None? No way.

If you can provide credible evidence that I am wrong, I will gladly concede.

[SeanB]

Cheap dashcams wth poor IR filtering so the built in LEDs are pretty useless as illumination, while they show up green foliage as red. These was also some footage by Reponut where he tried a cheap dashcam and it did the same, along with showing up clothing in odd colours.

[Nerull]

Good filters may filter more than IR as well - Canon DSLR filters block some red light as well as IR, to more closely mimic human vision which is less sensitive to red light. I'm not sure if they still do, but at times Canon has sold 'a' versions of their cameras specifically for astrophotography - these were fitted with filters that passed more red light, but still blocked IR and were still usable for daytime photography, even if the color wasn't quite as good. These cameras as much better at picking up faint red emission nebulas, but still don't see much IR, so the different-focus problem is avoided.

[timofonic]

Good filters may filter more than IR as well - Canon DSLR filters block some red light as well as IR, to more closely mimic human vision which is less sensitive to red light. I'm not sure if they still do, but at times Canon has sold 'a' versions of their cameras specifically for astrophotography - these were fitted with filters that passed more red light, but still blocked IR and were still usable for daytime photography, even if the color wasn't quite as good. These cameras as much better at picking up faint red emission nebulas, but still don't see much IR, so the different-focus problem is avoided.

Are those filters expensive? I wonder why that stuff can't be made at home...

[Wim_L]

It's tricky to make them at home, because the filters often are integrated into, or mounted directly on top of the sensor. Some are removable, but it can be tricky.

Anyway, professional astrophotography doesn't do it like that. They use a black and white CCD without filter (for better noise performance it's usually cooled), then manually add specific filters in front of the CCD. This offers two advantages: no need to limit resolution to avoid aliasing on the colour pattern, and you can pick some very specific filters, for example to measure or exclude some atomic emission lines.

As many astrophotography targets only change slowly, colour pictures can be made by taking multiple pictures with different filters and combining them afterwards. Simultaneous capture is possible too, by using dichroic filters and multiple black and white CCDs.

[tooki]

Cheap dashcams wth poor IR filtering so the built in LEDs are pretty useless as illumination, while they show up green foliage as red. These was also some footage by Reponut where he tried a cheap dashcam and it did the same, along with showing up clothing in odd colours.

Ah, that's plausible, since dashcams are a specialty camera that needs to be able to capture low light without mode switching.

Good filters may filter more than IR as well - Canon DSLR filters block some red light as well as IR, to more closely mimic human vision which is less sensitive to red light. I'm not sure if they still do, but at times Canon has sold 'a' versions of their cameras specifically for astrophotography ...

Isn't the "deep red" basically the near-IR? I don't think human eyes are particularly sensitive to it, so it doesn't much matter if it's filtered as well by a normal filter. Nikon eventually caught up and Now makes an astrophotography DSLR as well, by the way!

Are those filters expensive? I wonder why that stuff can't be made at home...

Well, do you have a vacuum deposition chamber at home? 

Good filters are actually very hard to make. It takes quite precision manufacturing to make filters that have no flaws, no unwanted color shifts, etc. And the narrower the frequencies the filter is to block/allow, the harder it gets. And then for photography use, you need antireflective coatings, which are literally impossible to create without the aforementioned vacuum deposition chamber.

But again, I'd hazard that for this purpose, even if you had all the equipment and chemistry, the biggest challenge would probably be dust. For a filter that goes in front of the lens (like a screw on filter), an imperfection wouldn't show up on the image because it's nowhere near focus. But the IR sensors are just above the actual sensor, and dust shows up as huge blemishes over many pixels.

[tooki]

Simultaneous capture is possible too, by using dichroic filters and multiple black and white CCDs.

I've wondered why the camera makers have never made a DSLR using 3CCD technology. (It used to be common in pro video.) I imagine the low light performance would be spectacular.

[Cliff Matthews]

I've wondered why the camera makers have never made a DSLR using 3CCD technology. (It used to be common in pro video.)

I seems large format sensors are a fixture now, so can you imagine the size of the prism block and size/weight of the camera? I think it would need a new name - The Schwarzenegger.

[tooki]

They'd be bigger, but not as big as you think, I suspect. (3CCD camcorders weren't a lot bigger than their single-CCD brethren.)

My hunch is more that it'd be impossible to maintain compatibility with existing camera systems (Canon EOS, Nikon F, etc.) because the distance between the lens mount and sensors would be much longer. But at the same time, I'd think that a new system with three sensors could have the potential for amazing image quality.

Of course, single-sensor image quality has made such strides that I guess the point is moot now...

[Nerull]

Cheap dashcams wth poor IR filtering so the built in LEDs are pretty useless as illumination, while they show up green foliage as red. These was also some footage by Reponut where he tried a cheap dashcam and it did the same, along with showing up clothing in odd colours.

Ah, that's plausible, since dashcams are a specialty camera that needs to be able to capture low light without mode switching.

Good filters may filter more than IR as well - Canon DSLR filters block some red light as well as IR, to more closely mimic human vision which is less sensitive to red light. I'm not sure if they still do, but at times Canon has sold 'a' versions of their cameras specifically for astrophotography ...

Isn't the "deep red" basically the near-IR? I don't think human eyes are particularly sensitive to it, so it doesn't much matter if it's filtered as well by a normal filter. Nikon eventually caught up and Now makes an astrophotography DSLR as well, by the way!

Are those filters expensive? I wonder why that stuff can't be made at home...

Well, do you have a vacuum deposition chamber at home? 

Good filters are actually very hard to make. It takes quite precision manufacturing to make filters that have no flaws, no unwanted color shifts, etc. And the narrower the frequencies the filter is to block/allow, the harder it gets. And then for photography use, you need antireflective coatings, which are literally impossible to create without the aforementioned vacuum deposition chamber.

But again, I'd hazard that for this purpose, even if you had all the equipment and chemistry, the biggest challenge would probably be dust. For a filter that goes in front of the lens (like a screw on filter), an imperfection wouldn't show up on the image because it's nowhere near focus. But the IR sensors are just above the actual sensor, and dust shows up as huge blemishes over many pixels.

The human eye is quite capable of seeing H-alpha, so its not into IR. H-alpha nebulas are easily visible through a telescope, but unmodded DSLR cameras have a difficult time capturing them. High end solar telescopes use a H-alpha bandpass filter and produce amazing views, much better than the standard attenuation filter.

And yes, external filters are pretty much universal on scientific cameras, and there are plenty of cameras available to the amateur astronomer that have a bare CCD and filter wheel, but using a DSLR is still fairly common.

There are a few companies that offer a service to replace the existing filter in many cameras, they were the only option for a while when Canon discontinued the 20Da, but they're still out there for other models. I'm not sure about the practicality of doing this at home.

Simply removing the filter completely, or replacing it with a spacer is also possible, but this almost requires the use of external filters. Its pretty much impossible to get a sharp focus with a camera sensitive to both visible light and IR at the same time, which is why the 'a' models still block IR.