Biomedical Engineering Reference
In-Depth Information
Figure 6.21:
(a) Wavelet modulus in first level of a PET brain image as shown
in Figs. 6.20 (a) and (b). (b) Thresholding of the wavelet modulus from (a) using
a wavelet shrinkage operator. (c) Thresholding of the wavelet modulus from (a)
with cross-scale regularization.
Spatial-frequency representations of a signal after wavelet expansion of-
fer the possibility to adaptively process an image data in different sub-bands.
Such adaptive scheme can for example combine enhancement of wavelet coef-
ficients in the coarse levels, and resetting of the most detailed levels for noise
suppression. We show in Fig. 6.22 how such adaptive processing can remarkably
FBP Reconstruction (Hann windows)
Adaptive Multiscale Denoising and Enhancement
Figure 6.22:
Denoising of PET brain data and comparison between unpro-
cessed and multiscale processed images.
