Filter out visible light from an IR image?

I have an IR camera that capture both visible light & Near IR wavelength. I'm wondering if it's possible to just retain the IR and block all the visible light.

I'm new to Image processing and Matlab so please be as specific as possible. Thank you!

Answer 1

[Star Strider]

My impression is that you would have to have a visible-light blocking filter on your camera lens if you want only IR images.

[John D'Errico]

As I point out, this can be done with color management techniques, and knowledge of the camera sensitivies.

[Dao Lam]

I don't have that filter. Is it possible to remove the visible light after the picture is taken?

[Star Strider]

I doubt it, but then I don’t have any experience with Raspberry Pi.

According to the Raspberry Pi NoIR documentation, it seems that you would have to have a physical filter to eliminate visible wavelengths. I would surf the Raspberry Pi forums to see what they suggest.

[Dao Lam]

Thank you! I really appreciate your help.

[Star Strider]

My pleasure!

Answer 2

[John D'Errico]

Add a filter that kills off any light in the visible spectrum, but not IR. Whats the problem?

If your goal is to do that AFTER image capture, that will require knowledge of the sensor sensitivities, and the RGB definition of what it returns. Then you can form a simple mapping to give an approximation. So a shaper, a matrix multiply and another shaper on the output.

Image Analyst would be a better source here, since it has been many years since I stopped doing these things, and I just wrote the fundamental codes for such operations, rather than the modeling itself. But the basic idea is to back up from the code values returned from the camera, through a set of shapers, then if you know the spectral sensitivies of the sensors, you can infer what each sensor might have "seen" to produce those code values. Then you multiply by an effective low pass filter, to retain only the part of the spectrum you care about. All of this reduces to a simple matrix multiply and shaper mapping.

[Dao Lam]

Thank you. You're right. I want to do it AFTER image capture. The camera I'm using is the NoIR camera module of Raspberry Pi. Do you have any resources that could help? What is a shaper?

[Image Analyst]

That camera http://www.raspberrypi.org/products/pi-noir-camera/ just has the IR filter, which is normally in front of the sensor, removed. Now a normal CCD has a spectral sensitivity that covers a huge range of the visible and IR parts of the spectrum. For a given image, you don't know how much came from what part of the spectrum. A given image, say it's uniform with a gray level of 100 gray levels, could be produced on a monochrome (but still broadband) CCD sensor with all green light (550 nm) or red light (630 nm) or all infrared light (greater than 700 nm). You just don't know unless you make assumptions or do things. Like you could illuminate your scene with only infrared light, or you could put an IR passband filter in there to block all IR light. But to just take some gray scale image and split it into two images, a visible only image and an infrared only image, I don't think is possible, unless, like I said, things were known or reasonable assumptions were made.

By the way, an infrared camera is not a thermal camera so don't think it is. IR and thermal operate in different wavelength ranges. A CCD will respond out to about 1050 nm while the thermal range is way, way out in the long infrared like 8,000 to 12,000 nm (8-12 microns). Your simple camera will not see "hot" things.

[John D'Errico]

Good. I'm happy to see Image Analyst chime in, as his knowledge is far superior to mine. His point about near IR and true thermal imaging is an important one.

Answer 3

[Spandan Samiran]

I am currently working on the NDVI image and crop analysis. I have removed the IR filter of my camera and using it to take near IR band images. But I am not sure if the images I take has a near IR band in it or not. Can you guide me how to detect near IR band in an image if present?