I have been looking at this tutorial from Astostuff on taking off the hot mirror on a D70, but cannot figure out from the page where the color mosaic filter is located.

Is it possible to remove the color mosaic filter off the sensor? Does converting the camera to IR involve removing the filter? Will I achieve any significant gains by removing the color mosaic filter? Does the color filter not absorb IR photons?

  • I don't believe you can take that off. There's an anti-aliasing filter and the color array that are made as part of the sensor. (The article doesn't remove the color array bit, just the IR filter) Jun 27, 2011 at 18:39
  • The anti-aliasing filter is sometimes removable, and it not actually deposited onto the CCD/CMOS sensor like the bayer filters are. Generally, the aliasing filter is just part of the filter-stack just like the hot-mirrror.
    – Fake Name
    Jun 28, 2011 at 3:56
  • People who do astrophotography know a lot about camera modding. Try and lurk (or ask) on an astrophoto forum. Cloudy Nights DSLR & Digital Camera Astro Imaging & Processing forum Jun 29, 2011 at 16:29
  • I have a broken camera here (7700 IIRC) which has a duff CCD. As it clearly had water in the sensor, I don't mind if it breaks it. Will try various techniques on it including the "boil in acetone" method, and let you guys know how it gets on. -A mailto mandoline at cwgsy dot net
    – user11795
    Sep 28, 2012 at 9:58

6 Answers 6


It's possible to remove the CFA (colour filter array) but extremely difficult. There is at least one company that will do it (Maxmax as asalamon74 states). Doing this is entirely unneccesary for IR conversion, in fact one of the things I like about IR photography is playing with the faint colours that result, which requires the CFA.

As to why you'd want to remove it... sensitivity. Each of the colour filters filters out approximately two thirds of incoming light. Removing them would give you a greyscale camera that's 1.6 stops more sensitive, so where you'd have to use ISO1600 normally with this camera you could shoot at ISO500

Also you wouldn't have to demosaic the resultant images so you'd get an increase in sharpness similar to what you get with a Sigma Foveon camera.

If you do a deep IR conversion (880nm or shorter) then you lose colour information anyway so you have nothing to lose by removing the CFA.

  • +1 But, if I understand correctly, each sensel is filtered for exactly one color, so all you get from removing the filter is +2/3 stops, changing ISO 1600 to ISO 1000, not ISO 500.
    – whuber
    Jun 27, 2011 at 19:54
  • @whuber by removing the filter you [approximately] triple the amount of light, if before you only got red light, you now get red, plus green, plus blue! A threefold increase would be 1.6 stops (doubling is one stop, quadrupling is two stops...) however the response of each filter overlaps to a degree so it's not exactly a factor of three, probably more like 2.5, or 1.3 stops.
    – Matt Grum
    Jun 27, 2011 at 20:05
  • 1
    @whuber it's actually like this: suppose each sensel gives a unit response, so the total response is 1+1+1+1 = 4. Now lift the green filter: the two formerly green sensels give 3 response each. Lift the red filter: the former red sensel gives 3. Similarly, the former blue filter gives 3. The total response, with all filters gone, equals 3+3+3+3 = 12. The ratio 12:4 = 3. White light is made up of red+green+blue if you have a green filter, you are filtering out two thirds of the light, giving you one third the response you'd have without the filter. Take the filter off and you get 3x the light
    – Matt Grum
    Jun 27, 2011 at 20:33
  • 2
    @Matt Your calculation is correct under the assumption each filter cuts out 2/3 of the light. In retrospect, that makes a huge amount of sense, because each filter needs to suppress the other 2/3 of the visible spectrum. To put it more simply: regardless of what color is placed over each sensel, it is originally restricted to about 1/3 of the spectrum. Removing the filters exposes every sensel to the full spectrum. Ergo, the light sensitivity triples. Thank you for your patience!
    – whuber
    Jun 27, 2011 at 22:17
  • 1
    @jr The maxmax site offers to remove the filters even from a Canon T2i, which is a recent camera (1 year old). However, the cost is twice that of a conversion for a 5D MKII (much older). @Matt FWIW, the same site states that after filter removal, "Visible light B&W cameras may gain about 1 stop of exposure." That agrees with your estimate; in my terms, it is equivalent to each color filter removing half the visible spectrum on average.
    – whuber
    Jun 28, 2011 at 13:15

If you remove the color filter array then (theoretically) you'll have a B&W DSLR. Removing the filter is quite complicated, there are a few companies offering DSLR B&W conversion like maxmax. Check their webpage they have quite good sample photos.

  • +1 That web page has great explanations and illustrations that provide answers to the original question. The answers are: beneath the microlens array and AAF; yes (it's possible but difficult); yes (for full IR); yes (depending on what you mean by "significant"); no (the color filter does absorb IR photons).
    – whuber
    Jun 27, 2011 at 20:34

I recently looked into the response curve for the bayer patterns and I only came across one camera that gave the curve through the NIR region.

Note that this curve is the sensitivity of the particular sensor with bayer attached (q(cmos)*q(bayer)). The NIR region is pretty much the same as the monochrome version of this camera, which means the bayer pattern doesn't do anything in the NIR region. Assuming that the bayer filter material is the same across the line, you don't need to remove it. Just remove the NIR cut filter.

If your NIR light is centered around 850nm, then all the pixels will have equal sensitivity and you can make a nice NIR image by horizontal and vertical binning, without removing the bayer pattern.

Bayer response NIR

  • *binning: in a machine vision camera you can set "binning" in hardware, which is to group the pixels 2x2, halfing the resolution in both direction, but gaining better signal noise ratio (which you need in the NIR area). Sep 28, 2012 at 16:38

At least from anything approaching a practical viewpoint, the color filter is not removable (you can probably get it removed, but removing it yourself is mostly out of the question). It's part of the sensor itself, in front of the sensor wells, but behind the micro-lenses. To remove it, you'd have to remove the microlenses, remove the color filter, and then put new microlenses back on.

While it's fairly easy to fabricate a sensor without a color filter to start with (and that has been done), removing one from an existing sensor would be more difficult -- to the point that it's mostly impractical.


The color filter array is typically right on top of the sensor, under an infrared filter panel. After removing the IR filter, the CFA can be mechanically scraped off of the sensor:

Scraping CFA from camera sensor

Effects of this destructive process are detailed here.

  • 1
    Woah! for people who hate their gear.
    – ths
    Mar 1, 2017 at 15:17

It's not necessary to remove the Bayer filter as part of an IR conversion. In my testing, the Bayer filter appears to be transparent to IR.

Compare the curves of the spectral sensitivity and QE of various wavelengths of light on a CMOS sensor. You can easily tell that the curves match almost perfectly.

Removing the Bayer filter usually removes the micro-lenses layered on the filter as well. So, Bayer removal will tend to reduce low-light performance, and create pixel vignetting, especially with shorter focal length lenses.

This answer should not suggest that Bayer removal is a bad idea for full-spectrum and B&W conversions, however.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.