Here's my thoughts on shooting near-infrared (near-IR) with a color, digital camera.
So I picked up a Infrared filter compatible with several lenses within my Leica M system. I had shot some near-IR on my old Nikon/Nikkor system, but those filters were of a different thread size; thus the need for a new filter.
Figure 1. Near-IR filter screwed into a lens on my Leica system. R72 is the filter type, representing the cuttoff wavelength, blocking light with wavelengths below 720 nm. The "46" is just the filter thread size, meaning it will screw into lenses with a 46 mm thread diameter; this obviously may be different for different lenses.
Thought #1: If you get one, make sure it matches your thread diameter. Also, this might
not be a good application for square or magnetic filters, unless light leakage is properly handled: any light leakage around the sides of the near-IR filter (or even through the bayonet between your lens and camera) will wash everything out. A screw-on filter is a good choice here. (If you have a light-leakage problem in your bayonet, a simple hair scrunchy might fix the problem).
Thought #2: 720 nm (R72) seems to me to be a good cuttoff wavelength/filter type for a normal, digital, color camera if you're unsure on that. There are other filters on the market with different cuttoff wavelengths. The more cuttoff wavelength delves into the infrared, the more it has to fight the rolloff of the IR-cutoff filter that's built into your camera's sensor (more on that later). To me, 720 nm seems to be a fair compromise.
Figure 2. First light for the new, near-IR filter. I did some post-processing on this image in Lightroom to convert it to black-and-white (B&W), and play with contrast and whatnot. So, no, this isn't directly out of the camera. But I find it really cool. I took the photo at a bus stop not far from where I live, so I've looked at this scenery countless times. Seeing this same location in infrared is pretty trippy. Aside from looking trippy, the filter seems to introduce a little vignetting; I did not adjust the vignetting in Lightroom, that's all from the filter.
Thought #3: The Bayer filter in your color camera will reduce the effective resolution of your near-IR images to approximately 1/4 of what they are normally. This is because in order to optimize signal to noise ratio (SNR) you should heavily favor the red channel as the input to your images.
[Edit: the 1/4 reduction in resolution figure might be a bit exaggerated, but only because even without any near-IR shenanigans the Bayer filter reduces the effective resolution somewhat. The point is that with the near-IR filter attached, the effective resolution will be reduced significantly more than what the Bayer filter normally does.]
It stands to reason that the red channel should dominate your signal input when shooting near-infrared. I mean ideally, ALL the signal should be in the red channel. In practice, the Bayer filter isn't ideal and some signal is present in the blue and green channels, but they will just be a noisy mess.
Do
not try to use "auto white balance," "auto-color," or "auto levels" unless you're conformable with really, really, unnecessarily noisy pictures. These auto-settings will
boost the B and G channels to match the R, but you don't want that. The real signal is almost completely in the R channel, and thus the R channel should be the primary source before B&W conversion.
Figure 3. Only about 1/4 of the pixels correspond to the R channel in your color sensor, and the R channel is where your near-IR data is going to be.
When shooting near-IR, you should expect that some of the fine detail you may be used to won't be there. You won't be able to crop in on your photos as much as usual.
For example, if you have a 24 megapixel (MP) camera, when shooting with a near-IR filter, expect the results to be similar to shooting with a 6 MP camera without the filter, and then upsampling to 24 MP. Detail will be lost.
To get the best signal-to-noise ratio (SNR), heavily favor the R channel before B&W conversion. If your photo processing software allows, take all (or at least most) of your data from the R channel. (This advice may be overly simplistic, even it it's not bad advice. See below in a couple paragraphs for an explanation.)
Here's what I do in Lightroom when adjusting the white balance. I just adjust the sliders all the way to the right to avoid blues and greens (See Fig. 4.)
Figure 4. In Lightroom, I slide the Temp and Tint sliders to the right, to lessen the influence of B and G data.
You can make the argument that the above is wrong and the SNR is optimized by making
no color balance adjustments at all, and simply add the pixel values of each corresponding raw color pixel (of which the R channel pixel will naturally will have the higher value) to create the corresponding B&W pixel. And yes, that may be true, but its correct implementation depends on the B&W profile is used to create the B&W conversion. So, just to play it safe, I bias my data to the R-channel.
In the future, if I produce an image that I want to frame or enter it into a contest, I might be more careful about pre-B&W conversion color balance. but unti then, biasing to the R-channel doesn't hurt the SNR much, if it even hurts it at all (to any significant amount).
Though #4: You're also fighting the IR-cut filter that's part of the color sensor. This is in addition to the Bayer filter.
Maybe I should have started with this one. Plenty of uniformed photographers will propagate the false myth that, "You can't do near-IR photography with a digital camera unless you modify the sensor to remove the IR-cut filter. The sensor's IR-cut filter blocks all the near-IR light making near-IR photography impossible with that camera." That claim is false.
Well, it's partially false. It's not impossible. As a matter of fact doing near-IR photography with a camera's IR-cut filter intact is
very possible. The IR-cut filter is not ideal, and cutoff has a "rolloff" to it. As the wavlenths become greater, the attenuation from the IR-cut filter becomes greater. It doesn't cut the IR out completely.
But iit is true that your near-IR filter screwed on to the front of your lens is fighting IR-cut filter in your sensor. The only photons that get through are those lucky enough to make it through both filters (not to mention the Bayer filter too).
To remedy this, be prepared to crank up the ISO a lot. Consider doing long exposures, even on a bright and sunny day.
In my case, the effective attenuation of adding near-IR filter is roughly about 10 1/2 stops (a.k.a., steps). That's huge. But it's not impossible.
Thought #5: Let's do a comparison between "with and without" the near-IR filter.
Figure 5. Normal color photo of the Lily Pond taken at the wrong time of year. No near-IR filter present. Handheld. ISO 64, 1/250 sec.
The image shown in Fig. 5 was taken after the autumn equinox, making the the wrong time of year for this subject. As I've mentioned before, this subject should be photographed closer to the summer solstice around sunset such that the sun is north of west, shining some sunlight onto the north-facing wall on the other side of the Lily Pond.
The way it is here (Fig. 5), the north facing wall is in the shade and it's just all wrong. This is not a good photograph compared to what it could be.
Figure 6. Same as Fig. 5, but converted to B&W, and some additional retouching done in a vain attempt to rescue the crappy image as much as I could. (No near-IR filter present.) No retouching can save this photo. It is bland and lifeless.
Converting it to B&W didn''t help. Without the proper lighting, the image is doomed. The architecture remains bland and 2-dimentional. It's like eating a stack of pancakes with no syrup. Just look at that photograph. It's lifeless. It's soulless. What a dump.
But what if...
What if instead of having the architecture be the main subject, let's de-emphasize that and instead emphasize the palm trees and flora and maybe even the people. How do we do that? That's where the near-IR filter comes in. Let's re-take the photo with near-IR filter screwed onto the lens.
Figure 7. Lilly Pond photograph with near-IR filter attached. Handheld. ISO 50000, 1/125 sec. Minimal edits in Lightroom, just enough to export it to a JPEG format.
The near-IR photo is shown in Fig. 7. Note that it looks very red. Of course it looks red. Nearly all the data is in the R-channel because it's near infrared. that's how wavelengths work. Also note the jump in ISO from 64 to 50000, and a 1 step decrease in shutter speed. Jaysus.
So now to convert to B&W and do some normal post-processing.
Figure 8. Same as Fig. 7 but after conversion to B&W and some additional post-processing. (Near-IR filter attached.) Lookin' pretty good.
The final image is way better with the near-IR filter attached (see Fig. 8). We changed the emphasis from the blandly lit architecture (giving up on that) to the trees and foliage. The people look nice and ghostly too. Definitely a better photograph.
While it's not obvious unless you zoom in, the near-IR version of the photograph is way noisier than the photo without the filter. But I think this just gives it some added character. 'Like an old-timey thing.
Thought #6: In conclusion, if you want to de-emphasize water & most manmade structures, and instead emphasize trees and plants, particularly in the sunlight, and humans too, even though they look a little weird, then near-IR might be a good option.
Thought #7: On the other hand, if you
want the water or manmade structures to be the central focus of the image, and you want to minimize distractions from plants, near-IR can backfire and make your image worse.
Figure 9. No near-IR filter.
The image in Fig. 9 is not a good photo, but it shows a location that has some potential. If I were to get off my bum and walk around to the other side of the shrubbery, maybe I could get some leading lines directing the eyes to the interesting looking bridge in the distance.
Let's see how it looks in near-IR.
Figure 10. Near-IR filter attached.
Ahh Jasus, make it stop. Make it stop (see Fig. 10). My eyes. They're burning. The filter made it worse. so much worse. What happened to the potential of leading lines? And the bridge: it's just gone. All gone. All of it gone. I only see a cacophony of plants and distractions. Nothing good has come of this. Nothing. Jaysus.
Figure 11. Another near-IR image that did not turn out so well.
The Bea Evenson Fountain in Balboa Park does not look so great in near-IR either. At least not from the perspective I chose (see Fig. 11). Don't get me wrong, the trees in the background look great, but they're not the intended focus.
Thought #8: Near-IR photographs can be kind of hit-or-miss. But I'll be carrying around the filter in my camera bag for sure.
