Public Lab Research note


Red Filter Rising

by cfastie | November 12, 2013 02:00 | 92 views | 15 comments | #9769 | 92 views | 15 comments | #9769 12 Nov 02:00

Read more: stable.publiclab.org/n/9769


Image above: False color infrared version of a landscape scene taken with a modified Canon A2200 with a Wratten 25A filter. The red and blue channels have been swapped, so instead of displaying RGB in the R, G, and B channels, this image displays NNR in the R, G, and B channels.
.

It looks like Ned is really on to something with the switch to a red filter instead of the infrablue filter. I now have two Canon A2200 cameras, one with an internally installed glass BG3 infrablue filter and the other with an internal Wratten 25A gel filter instead of the original IR block filters. The results from the two cameras are quite distinct and seem promising for the red Wratten 25A. This has prompted me to try to figure out how these two filters affect the wavelengths being recorded by the cameras, and how false color infrared images can be made from the recorded data.
.

BG3_W25A_curve2k.jpg
Transmission curves for the Schott BG3 (infrablue) filter and the Wratten 25A (red) filter. The BG3 filter blocks most red and green light, allowing near infrared (NIR) to be captured in the camera's red and green channels. The Wratten 25A filter blocks almost all green and blue light allowing NIR to be captured in the green and blue channels.
.
In addition to the transmission graphs in the figure above, the sensitivity of the three color channels to different wavelengths determines what these cameras record. The spectral sensitivity depicted in the graph below is determined by the red, green, or blue Bayer filters over each pixel on the sensor.
.

NikonD200_SpectralResponseCCD.jpg
Spectral sensitivities of a digital camera with the IR block filter removed. Each of the three color channels (colored curves) captures some NIR light and also some visible light from outside the standard range of its color. For example, the red channel captures some blue, green, and NIR light in addition to red. Each camera model or family will have different response curves, but this example might be generally representative.
.

The filters I installed allow only one of the camera's three color channels to capture visible light more or less normally. With the BG3 filter, the blue channel captures visible blue light, although it also captures a considerable amount of NIR, and the green and red channels capture mostly NIR. With the Wratten 25A, the red channel captures visible red light and some NIR, the green channel capture mostly NIR, and the blue channel captures almost pure NIR, the only case where this is true.
.

Table below: Colors of light captured in the red, green, and blue channels of digital cameras when unmodified or when modified with a BG3 or Wratten 25A filter. Each channel captures one predominant color and also varying amounts of other colors. The blue channel in the camera with the Wratten 25A filter may capture the purest band, with mostly NIR light. ChannelCaptureTable.JPG
.

A standard way to highlight plants with near infrared information is to make a false color infrared image. Only three color bands can be displayed in a color image (e.g., jpeg), so NIR, red, and green are used and blue is left out. The NIR data is displayed as red, the red data is displayed as green, and the green data is displayed as blue. Living plants stand out as red in these unrealistic images because foliage reflects almost all NIR light and few other natural surfaces do.
.

Table below: The colors displayed in each color channel in false color infrared images. The traditional method for satellite images (NRG) uses NIR, red, and green data. Only one visible light channel is captured by either of my modified infrared cameras, so a different combination of channels is required to simulate a false color infrared image. Both of these combinations include two channels that contain mostly NIR light. The pluses indicate that additional wavelengths are displayed in that channel.
NNNtable.JPG
.

After swapping the channels so the bands listed above are displayed in those channels, the color balance can be adjusted so the result resembles traditional false color infrared images (NRG). I have been doing this with infrablue photos made with Rosco filters and with a BG3 filter and calling them NBG. I think they should be called NBN, and the related versions from Wratten 25A cameras should be called NNR. The first attempts with photos from a Wratten 25A camera are below. These appear to be much better simulations of traditional NRG images. These images are also superior because there is more detail in the distance due to reduced atmospheric scattering of red light compared to blue.
.

BigVal2panR_B.jpg
Photos from the two modified A2200 cameras. The camera with a Wratten 25A filter was custom white balanced on a piece of red origami paper under a partly cloudy sky. The camera with the BG3 filter was white balanced on a piece of blue origami paper under blue sky. Shutter speed was slow for these shots (1/20 or 1/25 second), so comparing focus or foreground crispness is not meaningful, but less distant haze in the upper photo results from using red light instead of blue.
.

BigVal2panNNN.jpg
False color infrared images from the photos above. These were made following the channel swapping protocol in the table above. Although motion blur could differ between the images due to slow shutter speed, the better color and contrast in the upper image is due to reduced atmospheric scattering and maybe also to a more pure NIR channel.
.
When a Wratten 25A filter or other red filter is used, the camera or photo or process should not be called infrablue. How about super-red?


15 Comments

Thanks Chris and Ned, this is an encouraging line of inquiry. Remind me on the source for the Wratten 25 filter material, or is it in the prior notes?

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Hi Pat, The Wratten 25A is an old standard photographic filter that is still widely available in glass and gel. I think Ned bought a genuine Kodak 3" square gel on eBay for ~$10. That product was discontinued in 2008, but it is still available for some reason. I guess it's main use was in black and white film photography, and there a lot of them searching for a new role in life. Other manufacturers make equivalents.

It is also possible to use a Rosco filter like this one. The spectral qualities will not be the same, and will be somewhat unknown. I don't know how reliable Rosco's spectral graphs are, but I have noticed that some of their graphs are used for more than one filter. If the pure NIR channel produced by the Wratten 25A is responsible for some of its advantages over the BG3, those advantages could be lost if a Rosco filter is used.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


looking at the sensor sensitivity graph for the Nikon D200, its interesting that green is the least sensitive channel in the infrared band, implying that that channel might actually get the cleanest visible light channel if calibrated correctly. MaxMax is showing green to be the least IR-passing channel in at least the models they studied.

I didn't do a big search here, but there are several green filters that pass infrared and block blue and red channels. rosco

Reply to this comment...


Mathew, That does seem to be a consistent pattern. I think you are right that a green filter might offer the purest visible light channel in a single camera system.

Reply to this comment...


It's weird because we think of green being reflective, but compared to IR, it's very absorbent.

Reply to this comment...


I think doing calculations based on red and green might be two ways to get to the same result that Chris was seeking in his Oaktober note-- the change in chlorophyll from living to dead. NDVI, since its an absorbance vs. reflectance index might not be the best with a green channel, because we're comparing reflectance of green to reflectance of IR, but that said, it does seem like there's something there.

Reply to this comment...


@Ned, do you already have a spectral sensitivity curve for the canon A2200, with its ir block filter removed (similar to the Nikon curve above)? If you do, is there any way you could post that here, or send it to me?

Also, is there a good method I can use to create that sensitivity curve for myself?

Thanks!

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Hi Roolark - I don't have a spectral sensitivity curve for the A2200. I've been using LEDs to get a rough idea of spectral sensitivity but that's a very crude approach. The typical approach for calibrating sensors is to use an integrating sphere with a monochromator or other light source that provides calibrated and very narrow band illumination. One of these days I'll (or someone else will) connect with someone who will let me use such a setup.

Reply to this comment...


How do you swap channels in an unmodified camera from RGB to NIR? What kind of camera are you using? I am really interested in this, I am photographing forest with different IR filters with a normal (unmodified) and an altered (de-filtered) camera, two cameras same model, at the same time. Trying to determine NDVI and NBR. Any input to determine which wavelength and color I have in each channel will be useful. Thanks!

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Tbtouaki,

It sounds like you have a full spectrum camera, one that has the IR blocking filter removed and nothing replacing it. By placing a filter in front of the lens, you can get various proportions of visible and NIR in each of the camera's channels. Which filters do you have? If you have the equivalent of a Wratten 87, then all three channels will be pure NIR because no visible light can pass. This is the best way to get both a pure NIR channel and get a pure visible channel (from the unmodified camera), but you have to rectify the two images (one from each camera). If you have the equivalent of a Wratten 25A, then no visible light will reach the blue channel, very little in the green channel, and lots in the red channel. With this filter, I use the blue channel for NIR and the red channel for red to compute NDVI. The blue channel will record a very pure NIR image, and the red channel will be mostly visible but contaminated with substantial NIR. If you have the equivalent of a Wratten 47B, Schott BG3, or Rosco #2007, then little visible light will reach the red channel, so it can be used for NIR and the blue channel can be used for blue visible light when computing NDVI. But neither channel will be very pure.

People have taken the IR block filter out of several different Canon Powershots (some here). I have done some testing with the Canon A2200 and the A810.

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Hi

I have several filters to put in front of the lenses and was considering which to put inside, to replace the IR filter.

Thank you so much for your explanation.

I sill try with a wratten 25 A first (if I can get one).

Thanks again Rita

Reply to this comment...


Hi I currently have a modified canon (IR block removed) with a wratten 25A Red filter. My pictures are really nice looking but I have big problems with NDVI. Here is a picture of mine with NIR on blue channel and Red on the red channel: NIR1_Red3_b.jpg I can, of course alter them, as mentioned on this post, and have the red on the blue channel and the NIR on the red channel, its like this: NIR3_Red1_b.jpg But when I calculate NDVI, if I only look at the image the values range from -0.83 to 0.14, but the pictures look nice, as soon as I change the range on the color ramp to -1; 1, because the values are extremely low, my NDVI turns like this: ndvi.jpg As if all vegetation were dead.

I have borrowed a MaxMax modified canon camera for NDVI II with great images, and made shure all parameters were the same in both cameras, and taken photos of the same object at the same time: canon260.jpg The NDVI on the MaxMax camera is great as you can see here: canon_ndvi.jpg

Does anyone have any suggestions as to how I can calibrate my camera or post process my images to calibrate them, using the good camera? I would really appreciate it. Thanks again, Rita

Is this a question? Click here to post it to the Questions page.

Reply to this comment...


Hi Rita, Your photo above (NIR1_Red3_b.jpg) will not produce very good NDVI. Most of the pixels of plants have the same value in each of the three color channels. That is why the photo looks so gray. Healthy plants reflect a lot more NIR than red, so the NIR channel should much brighter (higher values) than the red channel. When NIR is captured in the blue channel (as with Wratten 25A) then healthy vegetation should look very blue (turquoise). When the values in the blue channel (NIR) are a few times greater than the values in the red channel (red), than NDVI values will be in the appropriate range.

To get the NIR channel (blue) to be brighter, a custom white balance must be done on the camera. The procedure (for blue filtered cameras) is here: http://publiclab.org/wiki/infrablue-white-balance. For red filtered cameras, substitute red paper in direct sunlight for blue paper in the shade. The idea is to flood the sensor with red light while the custom white balance procedure is done.

To make NDVI images, the channels do not have to be swapped. That procedure is for making false color IR images.

Chris

Reply to this comment...


Hi I have tried several white balance options before, base on what I've read. On this pics the settings were exactly the same as the maxmax camera. As you mentioned I had determined I had a problem with my NIR band, the values are too low, I need to add some "gain". I was interested in knowing if anyone had any suggestions, if there was a way to measure the "gain" I need to add to my NIR band, maybe regression analysis, or a way to determine a function (to add or multiply with the band) anything to correct photos already taken. But I will definitely try your suggestions this weekend. Will let you know later how it worked out. Thank you so much. Best regards, Rita

Reply to this comment...


Hi I had wrong white balance for my camera. Thanks everyone for the help. Sorry for the trouble If anyone has any idea on how to save pictures taken with wrong white balance I have quite a few I'd like to save. However from now on, of course, I'll use correct white balance. Thanks Best regards, Rita

Reply to this comment...


Login to comment.