Wavelet-based Rician noise removal for magnetic resonance imaging
Nowak, Robert David
MRI; multiresolution; denoising
It is well-known that the noise in magnetic resonance magnitude images obeys a Rician distribution. Unlike additive Gaussian noise, Rician noise is signal-dependent and consequently spearating signal from noise is a difficult task. Rician noise is especially problematic in low signal-to-noise ration (SNR) regimes where it not only causes random fluctations, but also introduces a signal-dependent bias to the data that reduces image contrast. This paper studies wavelet-domain filtering methods for Rician noise removal. We derive a novel wavelet-domain filter that adapts to variations in both the signal and the noise. The new wavelet-domain filter reduces Rician noise contamination in both high and low SNR regimes.