Patent US 6,928,182

A de-convolution process is applied to an MR, CT or other image (25) of a scanned-object (23) to derive the point-spread function (22′) at an object-edge and to pin-point from the mid-point of its full-width-half-maximum FWHM, the location (30) of the true image-edge. With the object-image (25′) overlying the PSF function (22′) in the de-convolution space, sub-pixels which follow location (30) are transferred to before it to re-construct the image-edge (25′) for sharper conformity to the object-edge (23). Sharp definition of image-contour (37) facilitates accurate determination of area and volume of image profiles (35) and their segmentation. The accurate image-edge definition enables viable correction of geometrical distortion in stand-alone MR diagnosis and treatment planning.