Image restoration process for magnetic resonance imaging resonance imaging

Claim 2

An image restoration process for correcting the warped reconstructed image produced by a magnetic resonance imaging (MRI) method where by determining the inhomogeneities of the magnetic field used in the MRI method corrected pixel locations for a corrected image are calculated, each such pixel location having a known offset, 1 V , 1 H , from a group of four pixels (x 1 , x 2 , x 3 , x 4 ) of the reconstructed warped image each of said four pixels having a known intensity value including a signal portion and a noise portion, said restoration process interpolating said four values to determine an intensity value for said corrected pixel location comprising the following steps: for each corrected pixel location, y, computing the corresponding intensity value using ##EQU4## where α i is the relative weight to be given each x i in computing said intensity value of y, by determining α 1 , α 2 , α 3 , α 4 within the following constraints and taking into account said offset: ##EQU5## where β 2 is a constant and whereby noise variance of said noise portion is uniform for each corrected pixel location and is independent of location of each pixel within said group of four pixels; and displaying or storing said computed intensity values for each corrected pixel location to provide a said restored image.

Claim 3

A process as in claim 1 where β 2 is greater than 1/4 and equal to or less than 1/3.

Claim 4

A process as in claim 1 where β 2 is substantially equal to 1/3.

Claim 5

A process as in claim 1 where α 1 , α 2 , α 3 , α 4 for a particular y are determined by: ##EQU6## and d H and d v are the dimensions of a said group of four pixels.

Claim 6

A process as in claim 1 where said noise portion of said signal is identical but independently distributed noise.

Claim 7

An image restoration process for correcting the warped reconstructed image produced by a magnetic resonance imaging (MRI) method where by determining the inhomogeneities of the magnetic field used in the MRI method corrected pixel locations for a corrected image are calculated, each such pixel location having a known offset, 1 V , 1 H , from a group of "n" pixels (x 1 , x 2 , x 3 , x n ) of the reconstructed warped image, each of said pixels having a known intensity value including a signal portion and a noise portion, said restoration process interpolating said pixel values to determine an intensity value for said corrected pixel location comprising the following steps: for each corrected pixel location, y, computing the corresponding intensity value using ##EQU7## where α i is the relative weight to be given each x i in computing said intensity value of y, by determining α 1 , α 2 , α 3 , . . . αn within the following constraints and taking into account said offset: ##EQU8## where β 2 is substantially a constant and whereby noise variance of said noise portion is uniform for each corrected pixel location and is independent of location of each pixel within said group of pixels; and displaying or storing said computed intensity values for each corrected pixel location to provide a said restored image.

Claim 8

An image restoration process for correcting the warped reconstructed image produced by a magnetic resonance imaging (MRI) method where by determining the inhomogeneities of the magnetic field used in the MRI method corrected pixel locations for a corrected image are calculated, each such pixel location having a known offset, 1 V , 1 H , from a group of "n" pixels (x 1 , x 2 , x 3 , x n ) of the reconstructed warped image, each of said pixels having a known intensity value including a signal portion and a noise portion, said restoration process interpolating said pixel values to determine an intensity value for said corrected pixel location comprising the following steps: for each corrected pixel location, y, computing the corresponding intensity value using ##EQU7## where α i is the relative weight to be given each x i in computing said intensity value of y, by determining α 1 , α 2 , α 3 , . . . αn within the following constraints and taking into account said offset: ##EQU8## where β 2 is substantially a constant and whereby noise variance of said noise portion is uniform for each corrected pixel location and is independent of location of each pixel within said group of pixels; and displaying or storing said computed intensity values for each corrected pixel location to provide a said restored image.

Claim 9

A process as in claims 1 or 6 including the following steps for sharpening said restored image: forming the complex Fourier transform of said restored image; forming a product by multiplying the above said transform by an inverse SINC function in the Fourier domain to filter said image; inverse transforming said product to form a sharpened image in the spatial domain.

Claim 10

An image restoration process for correcting a warped reconstructed image made up of image points produced by a magnetic resonance imaging (MRI) method; interpolating said image points to provide a restored image; forming the complex Fourier transform of said restored image; forming a product by multiplying the above said transform by an inverse SINC function in the Fourier domain to filter said image; inverse transforming said product to form a sharpened image in the spatial domain.

Claim 11

An image restoration process for correcting a warped reconstructed image made up of image points produced by a magnetic resonance imaging (MRI) method; interpolating said image points to provide a restored image; forming the complex Fourier transform of said restored image; forming a product by multiplying the above said transform by an inverse SINC function in the Fourier domain to filter said image; inverse transforming said product to form a sharpened image in the spatial domain.

Claim 12

A process as in claim 8 including the steps of: in said multiplying step utilizing a fast convolution technique of the circular or periodic type; saving the border regions of said restored image; utilizing said inverse SINC function in a spatial domain to filter said border regions; and forming a composite image in the spatial so main using said inverse transformed sharpened image for the central portion of said composite image and said filtered border regions for the borders of said composite image.

Claim 13

A process as in claim 8 where said inverse SINC function is mollified.

Claim 14

A process as in claim 10 where said mollified inverse SINC function is of the form ##EQU9## where B is a fixed constant depending on pixel to pixel spacing of said image and α determines the degree of mollification.

Claim 15

A process as in claim 11 including the step of adjusting "α" in accordance with ambient noise.

Claim 16

A process as in claim 8 where said inverse SINC function is truncated.

Claim 17

A process in claim 9 including the steps of rescaling said composite image and replacing any negative values with zero.

Claim 18

A process as in claim 9 including the following steps for forming said inverse SINC function: forming said function in mollified form as ##EQU10## where B is a fixed constant depending on pixel to pixel spacing of said image and α determines the degree of mollification; taking a discrete inverse Fourier transform of said function to form a function in the spatial domain; truncating said spatial domain function; Fourier transforming said truncated spatial domain function; and storing both said Fourier and spatial domain functions.

Claim 19

A process as in claim 6 where n is equal to 16 and where a "cubic B spline" type of interpolation is made and where β of Equation III is substantially constant between 0.46 and 0.50.