“…From these n images, the tensor is estimated to be the one which minimizes a notion of error to the set of acquired data (Johansen-Berg and Behrens, 2009). There are several methods for the estimation and regularization of second order tensor fields, including: weighted least squares (Basser et al, 1994), variational methods for the estimation of the image volume with positivity and regularity constraints (Chefd'hotel et al, 2002(Chefd'hotel et al, , 2004Neji et al, 2007;Tschumperlé and Deriche, 2003a,b), estimation in a Riemannian space (Arsigny, 2006;Fillard et al, 2007;Lenglet, 2006) and the use of sparse representations (Bao et al, 2009;Luo et al, 2009). …”