A New Framework to Automatically Select Noise Model for Rician Noise Estimation in MR Images

In this paper, we study a set of histogram and higher-order statistical (HOS) features for automatically identifying the presence of large background in the magnitude MR images. The robustness and discriminative power of each individual feature and combining feature sets are investigated using different MR images including brain, cardiac, breast, spine, stomach and noisy images corrupted by Rician noise with different standard deviations, σ ={5,10,15,20,25,30,35}. The performances of the identification approaches are evaluated in terms of sensitivity, specificity, and accuracy. Experimental results obtained on 2544 MR images show that an approach based on the kurtosis and histogram peak ratio (HPR) features outperforms significantly as compared to that of other approaches reported in this work. The proposed approach can be used for selection of distribution model (Rayleigh or Gaussian) for accurate estimation of Rician noise level in MR images having large or little background regions.