Foreground subtraction of cosmic microwave background maps using WI-FIT (wavelet-based high-resolution fitting of internal templates)

We present a new approach to foreground removal for CMB maps. Rather than relying on prior knowledge about the foreground components, we first extract the necessary information about them directly from the microwave sky maps by taking differences of temperature maps at different frequencies. These difference maps, which we refer to as internal templates, consist only of linear combinations of Galactic foregrounds and noise, with no CMB component. We obtain the foreground-cleaned maps by fitting these internal templates to, and subsequently subtracting the appropriately scaled contributions of them from, the CMB-dominated channels. The fitting operation is performed in wavelet space, making the analysis feasible at high resolution with only a minor loss of precision. Applying this procedure to the WMAP data, we obtain a power spectrum that matches the spectrum obtained by the WMAP team at the signal-dominated scales. The fact that we obtain basically identical results without using any external templates has considerable relevance for future observations of the CMB polarization, where very little is known about the Galactic foregrounds. Finally, we have revisited previous claims about a north-south power asymmetry on large angular scales and confirm that these remain unchanged with this completely different approach to foreground separation. This also holds when fitting the foreground contribution independently to the northern and southern hemisphere, indicating that the asymmetry is unlikely to have its origin in different foreground properties of the hemispheres. This conclusion is further strengthened by the lack of any observed frequency dependence.