Desktop version

Home arrow Computer Science arrow Computational Diffusion MRI: MICCAI Workshop, Athens, Greece, October 2016


Patch Matching

The similarity of a reference patch Pi;k with another patch Pj;l(d) associated with the d-th subject is characterized by weight

where Zi;k is a normalization constant to ensure that the weights sum to one. Here йм(г, к) is a parameter that controls the attenuation of the exponential function. As

in [8], we set /гм(г, к) = ^2pdfkM(Vi_k), where ft is a constant [8] and dfk is the estimated noise standard deviation, which can be computed globally as shown in [9] or spatial-adaptively as shown in [8]. The former is used in this paper. Parameter /гх = flctx controls the attenuation of the second exponential function, where ctx is a scale parameter. |M(P i;k)| denotes the length of the vector M(Pi;k).

Given D subjects, a “mean” signal can be computed based on the weights resulting from patch matching:

where S(xi, qk; d) is the measured signal associated with the d-th subject at location xi e R3 with wavevector qk e R3. Vi;k is a local x-q space neighborhood associated with (xi, qk), defined by a radius rs in x-space and an angle as in q-space. Note the bias associated with the Rician noise distribution is removed in this process [9]. a is the Gaussian noise standard deviation that can be estimated from the image background [9]. Without patch matching, a “simple averaging” version of (6) is given as

Found a mistake? Please highlight the word and press Shift + Enter  
< Prev   CONTENTS   Next >

Related topics