AUTOSTEREOSCOPIC VISUALIZATION AND MEASUREMENT:
PRINCIPLES AND EVALUATION
Jie Shan, Chiung-Shiuan Fu, Bin Li, James Bethel, Jeffrey Kretsch', Edward Mikhail
Geomatics Engineering, School of Civil Engineering, Purdue University
550 Stadium Mall Drive, West Lafayette, IN 47907-2051, USA, jshan@ecn.purdue.edu
Commission V, WG V/6
KEY WORDS: Visualization, Stereo, Vision, Measurement, Computer graphics, Photogrammetry
ABSTRACT:
Autostereoscope is a technology that allows for a viewer to obtain stereoscopic effect without wearing glasses. This may potentially
be used as a replacement or alternative to the current goggle-based stereoscopic technologies used in visualization and mapping
practice. This paper studies the principles of popular autostereoscopic technologies, fundamental issues in using this technology for
visualization, photogrammetry, and the performance of autostercoscopic photogrammetric measurement. In particular, the
autostereoscopic effect is studied in terms of viewing zone and perceived depth. As a fundamental step towards autostereoscopic
photogrammetry, the imaging geometry of such displays is analytically presented. To evaluate the properties and performance of the
autostereoscopic measurement, we conduct a series of experiments using a backlight autostereoscopic display. A stereo pair at a
pixel size of 25 and 50 microns are used in the study. Three dozens of well defined and easily identified feature points are measured
by seven operators using the developed autostereoscopic measurement toolkit Auto3D. The consistency of these measurement
results is analyzed. In addition, they are also compared with the ones obtained from regular stereoscopic display. The work is a
primary effort towards lighter and mobile image interpretation and measurement environment.
1. INTRODUCTION
Accurate and realistic 3D data collection and interpretation
require stereoscopic observation. Although photogrammetry
has been using stereo instruments for over a century,
continuous development in stereo display industry provides
many alternatives. In particular, the recent autostereoscopic
technology has been brought into the attention of
photogrammetrists and photogrammetry instrument vendors
(Petrie, 2001). In contrast to the traditional photogrammetric
technologies, autostereoscopic measurement is goggle-free
(Okoshi 1980; Motoki et al 1995) or aid-free (Petrie, 2001), and
can be used in mobile and field environment. This advantage
contributes possible technical alternatives in photogrammetric
practice and attracts emerging research on autostereoscopic
mapping and interpretation system (Petrie, 2001).
However, the performance of this new technology needs to be
thoroughly evaluated in terms of interpretation and mapping
capabilities. In this paper, we study the measurement properties
of the autostereoscopic display and conduct several
photogrammetric tests to evaluate the performance of
autostereoscopic measurement as. a possible technical
alternative for photogrammetric practice. A brief introduction
to the principles of the autostereoscopic technology is first
presented. We then quantitatively show‘ the 3D
autostereoscopic geometry, including the exact geometric shape
of viewing zone, the movement boundary of operators for
autostereoscopic measurement, and the perceived depth. The
movement boundary is important for measurement because
operators’ 3D perceived depth varies according to the position
of their eyes. To carry out autostereoscopic measurement and
evaluate its performance, a photogrammetric toolkit Auto3D is
developed based on the DTI autostereoscopic monitor. Design
* with the National Geospatial-Intelligence Agency
considerations in the Auto3D development are discussed.
Finally, we compare autostereoscopic measurement results
with the ones obtained from common monoscopic and
stereoscopic tools. Multiple operators are involved in the tests
by measuring a number of carefully selected feature points.
The results and discussion of the photogrammetric experiments
are presented in this paper.
2. AUTOSTEREOCSOPIC PRINCIPLE
Here the term “autostereoscope” is to indicate that a viewer can
perceive 3D information without viewing aids, such as goggles
and spectaculars. The most popular autostereoscopic
technologies are parallax barrier (Okano et al, 1999; Sexton,
1992) and lenticular plate (Sexton 1992, Pastoor and Wópking
1997). However, the general principle of an autostereoscopic
system can be described by using the parallax barrier geometry
(Okoshi, 1976; 1980). As shown in Figure 1, the parallax
stereogram, an image that comprises interleaving stripes from
the left and right images of a stereo pair, is placed in front of a
barrier made of opaque material with periodic transparent
vertical slits. Each fine transparent slit acts as a window to the
corresponding image stripes. The stereogram and barrier are so
arranged that the left eye and right eye of a viewer only
perceive the corresponding left image and right image,
respectively. Therefore, the barrier creates several 3D viewing
zones to provide the binocular parallax according to the
viewer's position. Similar to this general principle, the
lenticular plate consists of an array of cylindrical lenticular
lenses instead of parallax barrier (Hattori, 1991). Both
lenticular and parallax barrier techniques support multiple
viewing zones while the viewer moves the position.
