Is the Cosmic Microwave Background Radiation "noise" or "signal" ? In any one image it might be "noise" for one purpose and it might be the signal for a different purpose.
For any one image, if you model the noise as being some particular type, and model it as being a second particular type, and you find that the second model gives you "a better fit" than the first one, you cannot be sure that the second model is the correct one. Random processes are random and the true cause of the noise might have been the first model and the better fit of the second model might have been due to "overfitting" over a limited set of data.
We do Magnetic Resonance Spectroscopy (MRS). Each time we do an experiment, we get an enormously powerful peak. Nothing else comes within 1% of the power of that peak. So we should study the peak carefully and discard the rest as being noise, right? Wrong! For our purposes, that peak is noise and the interesting information is in that 1%. The big peak is a reflection of the amount of water in the system (and tissue mostly is water), whereas we care about parts per million of obscure chemicals, where 2 parts per million might be required for normal health and 4 parts per million might mean advanced cancer (or vice versus.)
