Multiresolution restoration of medical signals using the renormalization group and the super-coupling transforms

This paper presents a multiresolution approach to the restoration of Magnetoencephalographic (MEG) signals corrupted by colored Gaussian noise. We compare two methods, namely the renormalization group transform (RGT) and the super-coupling transform (ST). We conclude that although the RGT approach requires fewer site updates than the ST approach in order to converge, the ST approach is overall much faster. The multiresolution algorithm was tested with real and simulated data. In the case of simulated data, where the original signal's peak-to-peak value is known, the algorithm worked well with noise levels up to 80% of this value.