what is this “Initialize noise model from images” for? What is the difference between turning it on and off? Any suggestions
This option uses a more informed initialization for the noise model. Without it, we initialize the noise model with a constant value. With it enabled, we average the images power spectra and then use that to initialize the noise model. It helps when marginalization is enabled, since marginalization is sensitive to the noise model used, but usually the noise model converges after one or two iterations anyway.