Robust estimator in diffusion tensor estimation

LS estimator is the classical diffusion tensor estimate method. It could get optimal result with highest estimate efficiency on the condition that the noise distribution is Gaussian strictly. But it is quite not robust, for even one outlier could make the error result. M estimator is more robust than LS estimator, but its breakdown point is very low, thus it could not resist the outliers effetely in fact. According the problems exit above, MM estimator is proposed to estimate the diffusionin this paper which has higher breakdown point and higher efficiency. MM estimator with high breakdown point and high efficiency is proposed in this paper. And then the MM estimator is applied in the synthetic and real data. Through comparing MM estimator with various other estimating methods, a more effect and more robust result can be drawn.