Denoising of complex MRI data by wavelet-domain filtering: Application to high b-value diffusion weighted imaging
Wirestam R, Bibic A, Sthlberg F, Ltt J, Brockstedt S
Magnitude MR images show a Rician noise distribution and a non-zero minimum signal, often referred to as the rectified noise floor. Quantification is severely hampered in low-SNR measurements, for example, high b-value diffusion MRI. To reduce this problem, real and imaginary MRI data in the image domain were filtered, before construction of the magnitude image. The noise-reduction filtering (or denoising) was accomplished by Wiener-like filtering in the wavelet domain. The advantage of denoising the complex MRI data in the image domain, compared with filtering the complex k-space data, is that image artefacts caused by filtering-induced phase errors are avoided.