Jason,
1. Your picture tool is never correct, you shouldn't stretch.
2. Your 3D's are generally as bad as they could be !
3. Local max/min added
4. I have missed Richard sheet, is it in that thread ?
I'm sure you are trying to go somewhere, but one way or another you will not remove noise like in 1D plots. You will only forge an image, because there is no function behind the captured pixel values. You can apply as much as can be invented about forging images or extracting some characteristics, but enhancing like removing the unknown, that you will never.
What you could do is invent a program that would scan each line of the image, apply a bias type of of subtract, same to each line and repeat the process up to the level the resulting spectral does not change anymore. You could then presume having removed noise by preserving the resulting spectral ratio. But that would be valid only with a SNR reasonably good.
What I'm saying here is not what they call globally "Thresholding":
threshold(M, thresh):
"These functions allow you to discard all of an image's intensity levels below or above a threshold, or to invert them. Thresholding can be used to remove background features (or noise) in an image or to shift its intensity spectrum. Inversion can be used to enhance details".
"This function can be used in two ways. When thresh is positive, elements of M less than thresh are set to zero; otherwise they retain their original value. When thresh is negative, elements of M greater than the absolute value of thresh are set to zero; otherwise they retain their original value. When thresh is zero, all elements are set to zero".
Still interesting.
jmG