INTRODUCTION - Fundamentals of Three-Dimensional Digital Image Processing - page 14

perform on an actual human body, such as observations, examinations (for
example, virtual endoscopy [Toriwaki00, Toriwaki04]), and simulated surgeries
[Taylor96]).
1.3.3 Characteristics of 3D image processing
The following is a simple summary of some of the characteristics of 3D image
processing, as compared to 2D image processing (in particular, transforma-
tions, recognition, and understanding).
(a) As compared to human recognition and understanding of
images
Not everything is understood about human pattern recognition capabilities,
but they are extremely powerful, at least in the case of 2D images. Further-
more, humans have a high level of intellectual ability with regard to under-
standing 3D conditions (the form and relative location of objects) based on
visual input from a 2D source. Reproducing such capabilities in a computer
is a central problem in the field of computer vision, yet as of today in most
cases the abilities of computers lag far behind that of humans. The 3D images
treated in this topic, on the other hand, while perhaps limited by the capa-
bilities of imaging technology, retain information about all points in the 3D
space, including the interiors of subjects. Humans, it would seem, go through
a process of combining 2D images, cross sections, to mentally build a 3D
structure. Strictly speaking, we cannot yet guarantee that such a process is
actually occurring, but when viewed from the outside it seems as if that is
indeed the case.
For computers, on the other hand, moving from two dimensions to three
is simply a routine extension to the processing of data held in an array, and
is not fundamentally different. Computers can easily access any given image
element in either two or three dimensions, but humans are not able to directly
view the internals of a 3D image. Furthermore, because true 3D images also
include all information about the internal structure of the subject, complex
processing related to the recovery of a 3D image from a 2D source, such as
occlusion or stereopsis, is unnecessary. In that regard, computer processing is
even preferable to human vision.
Moreover, while most true 3D images are of completely new things that
until very recently humans have had no contact with, most 2D images are
of things that humans have been in constant contact with since the days of
the wall paintings of Altamira and Lascaux. As such, it probably would be
extremely dicult for a human viewer to have a direct comprehension of a
3D image in the same manner as with a 2D image.