Dominant Waveform Data

Question

I am working on some ideas that hopefully some of you guys might be able to help me with.  I don't know Photoshop well but I think it may be what I need to use for this project.
I need to be able to analyze a photo (or a portion thereof) and get the dominant waveform data.  I have been thinking that I need to work from RGB to Waveform but I am not sure that is the way to go.  I was wondering of there is a way to calculate this info from an image in Photoshop or even extract it from the code contained within an image.
As an example, I want to take a picture and have data that says essentially there are these kinds of blues at x frequencies, these yellows at x frequencies, etc. I am not looking for mapping info at this point (though that may be helpful, too.)
It is similar to doing an FFT analysis of an audio file. In the attachment you can see an analysis of a sound that shows all the pitch content as well as the time domain. The darker lines coreespond to dominant frequencies where the lighter ones refer to secondary ones. What I need for an image is obviously different but hopefully that helps you get the idea. I need to be able to read the data in the image and convert  the color data into the fequency spectrum. This is why I wonder if calculating the domiant waveform will be helpful.
I have played around with the Wide Histogram from Reindeer Graphics but that does not give me the dominant waveform (nor can I calculate it from the data it presents me).
If anyone has any thoughts on it this would be helpful. I posted this idea elsewhere but I have been rethinking it so thought it best ti repost it with the new ideas.
Thanks.
Mark Chambers

Answer 1

I neither know what a FFT-analysis of an audio-file is nor have I tested Wide Histogram from Reindeer Graphics, but I suspect what You want may not be achievable in Photoshop or at least only with great difficulty.
As I understand it, the histogram refers to the quantitative distribution of the pixels’ channels’ values and can not be referenced to wavelengths at all, because it does assess channel for channel and not pixel for pixel. What this means is that for example a peak in the histogram could refer to one area where all three channels are identical, therefore neutral or some areas where one/two channel/s is/are in dominance, therefore highly saturated. (see images)
It might be possible (utilizing the icc-profile’s connection to the l*a*b-space in some way that’s beyond me) to calculate wavelength and intensity pixel by pixel, but that would necessitate some pretty fancy Scripting and might be quite time-consuming.
Maybe someone else knows better, though.