Fourier analysis of object boundaries from two dimensional digitized images

Contours of objects obtained from two dimensional image segmentation can be used to identify an object shape or define a projection or slice of a three dimensional object. When a contour is compared to reference contours or to adjacent slices or projections, it is necessary to extract the important features of the curves, reduce the effects of noise, and normalize the contours. The method of Fourier descriptors is compared with the method of B-splines for achieving these goals for three dimensional sampled boundary curves. The effective filters in both space and frequency domains are examined, and the results are demonstrated with simulations.