extracting spectra using models or filters

I had an interesting conversation with Matt Nixon (Cambridge) today, about new ways one might extract spectra of stars if the true goal is exoplanet atmosphere transmission spectra. Right now most methods bin down the data until there is enough signal-to-noise in every bin/super-pixel to extract a difference spectrum in every pixel. But there are other ways to bin down, which could be guided by the theoretical spectral models. In the most extreme case, you would do the atmospheric retrieval (as they call it in the business) in the un-extracted 2D image data from the spectrograph! We discussed what pieces of technology would be needed to make this happen. I love this idea, because it is at the intersection of sophisticated hardware (these spectrographs are amazing) and sophisticated theory (the exoplanet atmospheres codes are non-trivial).