MRI Gibbs' ringing filtering with edge-protection

摘要

A magnetic resonance image (MRI) data array representing an image is filtered in k-space (Fourier space) domain to produce a low-pass filtered data array, a band-pass filtered data array and a high-pass filtered data array. These filtered k-space arrays are two-dimensionally Fourier-Transformed into the image domain where the magnitude of the band-pass filtered data array is thresholded and feathered to produce a fuzzy continuous valued (“gray-scale”) edge mask data array, and the real part of the high-pass filtered data array may, if desired, be soft-thresholded to produce a soft thresholded sharpening mask data array. The edge mask data array is multiplied with the sharpening mask data array and the result is added to the magnitude of the low-pass filtered data array in the image domain to produce a Gibbs' ringing and noise-filtered image to better represent the underlying anatomy.

主权项

1. A magnetic resonance imaging (MRI) processor system comprising:
memory configured to contain data arrays representing an image of patient anatomy in a k-space domain and in a spatial image domain, a data array in one of said domains being transformable into the other of said domains by Fourier Transformation; and at least one data processor configured to low-pass filter the data array, in at least one of (a) the k-space domain or (b) the spatial image domain of said image of patient anatomy, to produce a low-pass filtered data array in the k-space domain or the spatial image domain; wherein, if the low-pass filtered data array is produced in the k-space domain, it is transformed into the spatial image domain; band-pass filter the data array, in at least one of (a) the k-space domain or (b) the spatial image domain of said image of patient anatomy, to produce a band-pass filtered data array in the k-space domain or the spatial image domain; wherein, if the band-pass filtered data array is produced in the k-space domain, it is transformed into the spatial image domain; high-pass filter a data array, in at least one of (a) the k-space domain or (b) the spatial image domain of said image of patient anatomy, to produce a high-pass filtered data array in the k-space domain or the spatial image domain; wherein, if the high-pass filtered data array is produced in the k-space domain, it is transformed into the spatial image domain; threshold and feather said band-pass filtered data array in the spatial image domain to produce an edge mask data array; multiply said edge mask data array with said high-pass filtered data array in the spatial image domain; and add the result of the multiplying to said low-pass filtered data array in the spatial image domain to produce a Gibbs' noise filtered image data array in the spatial image domain.