Shape-based interpolation (CT and MRI)

The need for shape-based interpolation is discussed, and some quantitative experiments are reported which were done to evaluate the relative performance of a number of interpolation methods for tomographic imaging of the human body. The datasets used for comparing the methods were three different sets of computer tomography (CT) slices of human heads and a set of magnetic resonance imaging (MRI) slices of a brain specimen. The initial experiments show that the interpolation method using an (approximately optimal) chamfer distance based on a 3*3 template in which pixels with an edge on the boundary are considered to be a unit distance from the boundary appears to be the method of choice.<
>