Wavelength in every rgb value?

Hello,

Although I'm not keen in advanced physics and math, I believe I have quite simple question, however as I googled for info, there were a lot of different and ambiguous answers.

Now, what I need to do is to retrieve approximate wavelength value from every pixel RGB values (corresponding wavelengths for red,green and blue) in a photo.

As I understand I'll have some problems with gray (black, white) and violet colors.

Another thing I don't get is whether I need to know pixel spacial coverage in the photo or is it enough to convert photo to HSV form and somehow process Hue values(which I have an idea how, but not completely sure). If so, must each band separated before or after HSV conversion(if it is the way at all).

What I need it for is spectral signature analysis, maybe there is some other way to determine wavelength values of the pixel?

I would really appreciate if someone with better knowledge in this field would help me and give me a start.

Answer 1

[Walter Roberson]

RGB to wavelength cannot be done, in general.

http://www.mathworks.com/matlabcentral/newsreader/view_thread/313712

http://www.mathworks.com/matlabcentral/newsreader/view_thread/105251

[Medardas]

hmm..dunno, how did I miss those questions, but anyway, the question when becomes, how it is possible to extract and identify spectral signatures from multispectral images?

Answer 2

[Image Analyst]

Many colors are not a single wavelength, for example brown or skin tones. There has been some work in the area:

 Heikkinen, V.[Ville], Lenz, R.[Reiner], Jetsu, T.[Tuija], Parkkinen, J.[Jussi], Hauta-Kasari, M.[Markku], Jääskeläinen, T.[Timo],
Evaluation and unification of some methods for estimating reflectance spectra from RGB images,
JOSA-A(25), No. 10, October 2008, pp. 2444-2458.

They probably take the spectral responsivity curves of the three sensors in the camera and add them together weighted by the RGB values. But I don't think you need to do that. I think you probably just need to look at my color segmentation demos in my File Exchange: http://www.mathworks.com/matlabcentral/fileexchange/?term=authorid%3A31862