An Improved Homomorphic Filtering Method for Inhomogeneity Correction of Magnetic Resonance Images
LI Yin, WANG Ze, ZHU Yi-sheng (Dept. of Biomedical Eng. , Shanghai Jiaotong Univ. , Shanghai 200030, China)
In order to improve the uniformity and diminish the distributions overlapping of different tissue intensities in the magnetic resonance (MR) images, an improved homomorphic filtering correcting method was proposed, based on the smooth, multiplicative and slowly varying features of the bias field. This method combines I he homomorphic filtering and B spline smoothing. Firstly, a coarse bias estimation is got by Gaussian homomorphic filtering on the corrupted image. Secondly, B spline smoothing is performed on some interested area of the coarse estimation. Finally, a refined bias estimation is got by B spline extrapolation. This method can quickly remove the bias field from the corrupted MR images. Simultaneously, the spline smoothing step can eliminate the boundary artifacts caused by low-pass filtering. The simulations on both head phantom image and simulated MRI image show good results.