Skip to main content
. Author manuscript; available in PMC: 2016 Jan 31.
Published in final edited form as: Magn Reson Med. 2014 Feb 18;73(2):828–842. doi: 10.1002/mrm.25176

Fig. 7.

Fig. 7

Velocity magnitude of DFW denoising on reconstructed undersampled data. From left to right: Original flow data, ESPIRiT reconstructed flow data, DFW with SureShrink, DFW with manually chosen thresholds, ESPIRiT with l1 spatial wavelet regularization and the 5.4-fold Poisson-disk sampling mask used in simulation. The flow phantom k-space was retrospectively subsampled by 5.4 using the same sampling mask for each VENC. ESPIRiT with and without l1 regularization were used afterward. Applying DFW with SureShrink reduces some incoherent artifacts in the reconstructed data, but is overly conservative. DFW with more aggressively chosen thresholds improves the performance significantly. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]