INTRODUCTION - Fundamentals of Three-Dimensional Digital Image Processing - page 17

Fig. 1.4. A 3D image constructed from a sequence of cross sections.

the arboroid structure. Such a procedure (2D processing with 3D image re-
construction) would be worthwhile if the distance between cross sections is
large as compared to the resolution within cross sections. In order to make
such a method work, however, the structure must extend in a direction near
perpendicular to the plane group of cross sections. If the direction that the
figure predominantly extends in is not known beforehand, then features that
run parallel to the cross sections may be lost.
When extracting the ellipsoids from the cross sections (the 2D images),
implementing a method by which processing of the n th cross section will only
extract those ellipsoidal forms that form a vertical overlay with those in the
immediately preceding ( n 1)th cross section also might be an appropriate

technique by which to make smooth vertical connections in the perpendicular
direction of the subject. Such a technique would be a natural method when
there is some degree of spacing between cross sections.
As an alternative method, standard 3D processing technique might, for
example, use a 3D difference filter and threshold processing to extract a 3D
edge surface and then perform 3D axis thinning to find diagram cores. If the
distance between cross sections is su ciently small (for example, within about
twice the size of the picture elements used in the cross sections), then pro-
cessing might consist of the use of appropriate methods (linear interpolation,
etc.) to fill in additional cross sections, or conversely elements from within

−