An Ecological Approach to Partial Binocular-Overlap
James E. Melzer and Kirk Moffitt
2701 Orchard Parkway
San Jose, California 95134
A partial binocular-overlap helmet-mounted display (HMD) allows the presentation of wide-field-of-view imagery with no
loss of resolution and a reduction in size and weight. One trade-off with these attributes is binocular rivalry created by the
edge of the imagery seen by one eye overlaying conlinuous imagery seen by the other eye. Three distinct methods are
considered that reduce this rivalry-and there are trade-offs with each approach. These three methods are the use of optical
stops or filters that provide a luminance gradation that softens the overlapping binocular edges, left/right eye assignment for
the flanking monocular regions, and contour lines superimposed on the imagery that correspond to the binocular/monocular
borders. These approaches to improving the quality of partial binocular-overlap HMD imagery are considered within an
ecological framework, where departures from ecological validity may impact visual perception and system performance.
A partial binocular-overlap HMD presents the user with a central binocular image flanked by monocular images. This
design offers several advantages over fully- or lOO%-overlapped HMD imagery.
1 Compared with full overlap, a 50% partial
overlap provides a 50% increase in horizontal field-of-view (FOV). This FOV increase, illustrated in Figure 1, is achieved
with no gain in weight or size, or loss of resolution. A wide horizontal FOV is critical for tasks involving balance,
orientation, night pilotage, and situational awareness. Trade-offs to achieving a wide FOV with partial overlap include
problems with distortion, roll compensation, rcquircd overlap, and binocular rivalry.l
Monocular 1 Binocular I Monocular
Figure 1. Partial binocular-overlap results in a gain in the horizontal FOV. The
dashed lines indicalc the boundaries of the binocular region. The center of
the display FOV is the ccntcr of the binocular region.
An ecological approach provides a framework for Lhc design and evaluation of partial binocular-overlap HMDs.
Ecological vision, as used in this paper, considers the human and visual environment to be reciprocally coupled and not
iudependent.2p3 An ecological analysis of a man-machine system begins with an evaluation of the environmental constraints
relevant to the human operator. One criterion of this analysis is the concept of ecological validity, which is based on the
assumption that ecological evaluations or nalural sccncs and Ihe way scenes are viewed by the observer are relevant to
vision.4v5 The quality of display systen~s n~ay bc improved by increasing their ecological validity. The degree of ecological
validity in binocular HMDs will bc: shown to bc linked lo image quality and freedom from display artifacts.
2. RIVALRY AND LUNING
Luning refers to the crescent-shaped dark images that lie on the monocular sides of the monocular/binocular boundaries of
a partial-binocular overlap video display.
The circular aspect of luning is due to the fact that miniature CRTs used with
HMDs have circular images. The dark images are characterized by a tendency to alternate over time with the display imagery,
and to promote oculomotor instability. Image alternation indicates that luning is probably due to binocular rivalry between
the eyes. Luning is exacerbated by the unnatural black edges which surround the HMD imagery.
We have preliminary evidence that luning is primarily a cosmetic problem, and does not affect performance. This
evidence comes from studies at Kaiser Electronics on the presence of luning crescents, and the detection of small targets in and
around tithe luning area of the display. Field reports from simulator and flight tests have revealed that the luning crescents are
frequently not noticed after pilotage activities begin imposing workload. Nevertheless, performance for other types of tasks
could be affected, and distraction and fatigue effects from the rivalry remain unknown.
The results from several Kaiser Electronics’ studies will be described. The reader is cautioned that these are preliminary
laboratory evaluations that lack the richness of the cockpit environment. For example, several of the experiments rcq!?e?
fixation on a target, the tasks were one-dimensional and workload was minimal, and the imagery was static. On Ihc ot!lcr
hand, these experiments did provide a simulation of critical aspecw: of binocular HMDs.
Three methods of attenuating luning with partial binocular-overlap imagery have been identified.
These methods are:
1. Luminance roll-off of the binocular imagery, 2. Left/right eye assignment of the monocular regions, and 3. Drawing
contour lines on the imagery that correspond to the binocular edges. These three methods will be discussed in more detail in
the following section. They will then be considered within an ecological framework.
3. DISPLAY AND IMAGERY CONFIGURATION
3.1 Luminance Roll-Off
The first method of attenuating luning is by softening the high-contrast boundary edge.
Observations at Williams Air
,Force Base6 and at the Air Force Armstrong Aerospace Medical Research Laboratory, and formal studies at Kaiser Electronics,
have indicated that rolling-off the luminance at the edge of the binocular region blends the binocular imagery into the black
region and reduces the perception of luning. For example, one Kaiser Electronics’ study found that a cosine taper of display
luminance was optimal for reducing luning. In terms of utility for a binocular HMD, luminance roll-off can be implemented
with optical filters or electronically at the CRT. Filters involve introducing optical elements into the HMD system, while
rolling-off the CRT image requires image processing.
3.2 Eye Assignment
The second method of attenuating luning involves the arrangement of the monocular regions.
displays may be configured to have either convergent or divergent overlap.
The right side of Figure 2 shows the configuration
of divergent imagery, and the left side of this figure shows a corresponding geometric layout of real-world surfaces.
can observe the effects of a divergent HMD design by looking at a distant object through two short tubes that are canted
Your right eye will see more of the right visual field, your left eye will see more of the left visual field,
and both eyes will view a central binocular region.
Note that the image placement is divergent, and that the eyes are
making divergent eye movements.
The left side of Figure 2 shows a real-world geometric arrangement that results in divergent imagery.
A solid object that
is viewed with both eyes is located in front of background imagery and a black surround.
An example would be a book held
at arm’s length. The depth discrepancies between this geometric arrangement and binocular HMD imagery will be discussed
in the next section.
and into a black
images are continuous
and are both coplanar
Figure 2. The geometric and imagery layout of a divergent arrangement. LE=Left eye, RE=Right eye.
We previously stated that ‘we had not observed any left/right eye assignment effects on luning with simulated partial
1 These observations were based on a very small FOV (-10’). Subsequent work with larger FOVs
revealed that eye assignment does affect luning.
We also equated a divergent arrangement with placing a septum between the
eyes, but now think that a near object viewed against a more distant background is a better conceptualization.
A convergent overlap image configuration is shown in the right side of Figure 3. Similar to the divergent example, the
reader can observe the effects of this design by looking at a distant object through two short tubes that are canted slightly
inwards. Your right eye will see more of the left visual field, your left eye will see more of the right visual field, and both
eyes will view a central binocular region. This configuration is commonly referred to as the
and is used in
non-pupil-forming head-up displays (typically with a binocular overlap of 5-109, where the exit aperture of the optics forms
the -knothole and the observer views the collimated display imagery. Real-world geometry that provides convergent imagery
is shown in the left side of Figure 3. As an example, a convergent view is present whenever we peer through an opening
such as a Gndow.
T$e cCce of divelgent versus convergent overlap may depend on the packaging difficulties with binocular HMDs.
Factors such as display FOV, eye relief? and exit-pupil diameter determine the size of the combining optics. Large combining
limit or preclude the necessary interpupillary distance (IPD) adjustment and fitting to the head and face with a
convergent HMD design.
\ dno6ar /Black’
Figure 3. The geometric and imagery layout of a convergent arrangement. LE=Left eye, RE=Right eye.
3.3 Contour Lines
The third method of attenuating luning requires that explicit contour lines be drawn on top of the HMD imagery, such
that they outline the binocular region.
That is, the binocular imagery is distinguished from the monocular imagery by thin
black lines drawn on the imagery. These contour lines segment the image similar to the close binocular object in Figure 2.
This method can be implemented with either divergent or convergent imagery, and has no impact on HMD optics, mechanics,
or electronics. The use of contour lines to attenuate luning came from an ecological evaluation comparing divergent and
convergent configurations. This evaluation will be discussed further in Section 4.3.
4. AN ECOLOGICAL EVALUATION
An ecological evaluation examines the relationship between the organism and its environment,
The basic approach taken
in our ecological evaluation is to compare partial binocular-overlap HMD imagery with analogous real-world scenes. This
evaluation can provide remedies to lessen the difference between the display and the real-world and to improve image quality.
The three luning-attenuation methods discussed in the previous section will now be evaluated using this ecological approach.
4.1 Luminance Roll-Off
The black surround shown in Figures 2 and 3 is unnatural and is a major contributor to luning. Since a bright surround
is not feasible, tapering the edge of the image provides a means of softening the high-contrast boundary edge.
The effect is
similar to the low to moderate contrasts that predominate in the natural world. We have experimentally shown that a variety
of luminance roll-off functions serve to lessen the perception of luning.
While other researchers have reported success with putting an optical stop into the HMD optical train6, we have observed
a “fuzzy” image with the use of a cosine luminance roll-off in a simulated partial binocular-overlap HMD. The rolled-off
image from one eye either combines or rivals the unaltered image from the other eye. The fuzziness problem is not severe,
and is generally preferable to the dark luning crescents that are otherwise present.
4.2 Eye Assignment
With a natural scene, the geometric arrangement of the environment and the separation of the left and right eyes determine
the eye that receives the visual information. With a partial binocular-overlap HMD, this assignment of scenery to eye is no
longer deterministic. Monocular regions of the scene can be assigned to either the left or right eye.
A central point of this
paper is that one of the two eye assignments is more ecologically valid, and results in superior image quality.
Researchers at The Smith-Kettlewell Eye Research Institute in San Francisco have studied the effects of ecological
validity on binocular rivahyp They demonstrated that there are natural opto-geometrical restraints that determine the
combinations of depth and left/right eye views of a display. Violating these constraints with ecologically invalid displays
was shown to result in a significant increase in binocular rivalry at the unpaired monocular regions of the display.
The left half of Figure 2 shows a three-dimensiona scene of a close object occluding the background.
Note that there are
unique (unpaired or monocular) regions of the background seen by each eye (shown as dotted regions). The corresponding
stereogram is shown to the right of the 3D drawing. The binocular object is seen as nearer than the surrounding surface due
to it having more binocular disparity. Reversing the eye that views each monocular region, while maintaining the
object/background depth relationship, results in an ecologically invalid display. The resulting stereogram has no counterpart
in the physical world, and was found to create visual suppression and rivalry.4
Similarly, Figure 3 shows a view through an aperture to a background surface. The unpaired monocular area is now
nasal relative to the binocular area for each eye. Reversing the eye that views each monocular region, while maintaining the
aperture/background depth relationship, resulted in an ecologically invalid display and increased rivalryP
Experiments conducted at Kaiser Electronics over the last two years have revealed several interesting eye-assignment
effects. In an informal study, 24 out of 2.5 observers rated a convergent display as having less luning than a divergent display.
This comparison was formalized in a study involving convergent and divergent overlap, random-dot and realistic scenery, 45”
and 60” FOVs, and placement of the superimposed edge on the left or right side of the display. Luning was found to be least
evident with a convergent display and a real-world scene.
We propose that the increased luning found with divergent versus convergent partial binocular-overlap displays results
from the greater ecological invalidity of the divergent configuration. Consider the viewing conditions for divergent imagery.
Binocular imagery in the natural world, such as a book held at arm’s length, is immediately surrounded by unpaired monocular
regions--sections of the background that are occluded by the book for one eye. The right eye sees part of the right-hemifield
where the left eye sees part of the object. Luning is typically not noticed in such situations when attention is focused on the
book and the eyes are appropriately converged. There are two distinct features of this imagery. The first is that the object is
closer than the background. That is, the object and the background have binocular disparity that is seen as depth. There is
also evidence that the eye assignment of the unpaired monocular regions provides depth information4J
Second, the object
has an explicit monocular/binocular border, and is separate from and occludes the background. Neither of these features exist
with HMD imagery, where the central binocular region is at the same “depth” as the monocular regions and these regions are
Similarly, when viewing a scene through an opening or aperture, the flanking monocular imagery affords the observer
depth information. Convergent HMD imagery is similar to the real world in providing seamless monocular/binocular regions
that are equally distant. In this sense, the convergent HMD is valid with respect to the ecological world. Invalidity may
come from the (black) surround, which is close in the real world but coplanar with the HMD imagery.
If much of the luning with a divergent display is due to the continuous monocular/binocular regions, then luning should
be reduced by making these regions discontinuous. This discontinuity may ameliorate the problem of these coplanar image
4.3 Contour Lines
Our ecological analysis of divergent HMD imagery led to the prediction that explicitly defining the binocular region
would increase ecological validity and improve image quality.
The near object in the left side of Figure 2 creates borders that
separate the binocular object and the. monocular background. Even though the HMD imagery remains continuous and without
depth, explicit borders should improve the display.
Using a binocular HMD simulation, we experimented with explicitly defining the binocular imagery by drawing black
contour-lines on the continuous imagery that corresponded to the edge seen by the other eye. The binocular imagery was
distinguished from the monocular imagery by a superimposed thin black line, as illustrated in Figure 4. In this
demonstration, the contour line had the predicted effect of reducing luning. In fact, the reduction was frequently total-no!
unlike many of the reports with convergent imagery.
A formal experiment conducted at Kaiser Electronics on the efficacy of contour lines replicated the finding that convergent
overlap produced less luning than divergent overlap, and found that displays with contour lines produced less luning than
displays without contour lines. Naive observers viewed a simulated partial binocular-overlap HMD image for 20 seconds on
each trial, and pushed a button whenever they perceived that luning crescents were dominating the imagery. The results from
this experiment, illustrated in Figure 5, show that
converging the imagery or inserting contour lines will reduce the
perception of luning.
The results of this contour-line experiment and previous investigations of luning conducted at Kaiser Electronics suggest
several alternatives for reducing the magnitude of luning with partial binocular-overlap displays. A convergent configuration
is effective at reducing luning compared to a divergent configuration. The use of contour lines is also an effective means of
reducing luning. The results shown in Figure 5 indicate that combining convergent overlap with contour lines would
optimize the display. The trade-off is the design and packaging of a converged optical system and the continuous presence of
Explicit contour lines superimposed on divergent and convergent HMD imagery. The left side
shows the left-eye (LE) and right-eye (RE) views. The right side shows the resulting binocular
view. The X’s designate the center of the binocular region.
Figure 5. Apparent luning as a function of overlap format and contour lines.
The data from this study support an ecological explanation of luning with partial binocular-overlap displays. Increasing
ecological validity lessens the perception of luning.
Departing from ecological imagery results in increasing amounts of
The first implementation of contour lines with an HMD was with Raiser’s WideEyeTM partial binocular-overlap HMD,
which has divergent imagery. Informal reports from trade-show attendees were that the contour lines virtually eliminate
hming. The utility of contour lines is that they appear to reduce luning with divergent HMDs, and that they are relatively
easy to generate. The major trade-off is that they are always present on the display.
The data from Figure 5 show that contour lines reduce residual luning with a convergent HMD.
problems with convergent configurations include the coplanar, black areas surrounding the imagery that create high-contrast
edges. The contour lines trace these high-contrast edges, and may reduce the negative effects.
5. VIEWING CONDITIONS
An ecological approach may be appropriate for other attributes of HMDs. Table 1 presents a list of attributes of natural
scenes and a list of attributes of partial binocular-overlap HMD imagery. Most of the differences between natural scenes and
HMD imagery are due to limitations in technology and biomechanics. For example, the natural instantaneous FOV for
humans exceeds 2OO”, while the HMD FOV ranges from about 60” for a lightweight flight-worthy system to about 120” for a
heavier and larger simulator system. With a helmet-presented FOV approaching 60”, the observer begins to react to the HMD
as his direct view of the world (as opposed to watching a display).
With a FOV of about 90”, the observer becomes immersed
in the imagery.7
Advances in optical materials and design, lightweight composite materials, and electronic displays will gradually allow
larger FOVs for flight-worthy HMD systems. Similarly, HMD attributes such as brightness, resolution, and color are driven
by the available technology+urrently miniature CRTs. These quantitative factors can be compared with the last five
attributes listed in Table 1, which represent qualitative differences between natural scenes and HMD imagery. For example, as
the head tilts, the natural world remains gravitationally upright. With an HMD and head tracker, the image of the world will
tilt with the head unless roll compensation is incorporated into the HMD system. There has been speculation that roll
compensation may only be required with larger FOVs, and roll compensation is not currently implemented with a fielded 40”
HMD. The question becomes: When does roll compensation significantly improve pilotage? Such qualitative design
decisions are difficult to answer, but offer the promise of immediate and significant improvements in HMD quality and
Comparison of Natural Scenes with
Partial Binocular-Overlap HMD Imagery
No temporal lag
Natural eye assignment
No noticeable rivalry
Partial Binccular-OverlaD HMD
Imagery lags head movements
Distortion compensation required
Imperfect image alignment
None or limited, w/ variable sensor separation
Roll-compensation absent or lagging
Arbitrary eye assignment
The ecological validity of HMD imagery has been considered as a criterion in the evaluation of three methods of
attenuating rivalry or luning with partial binocular-overlap HMDs. Luminance roll-off eliminates the unnatural high-
contrast edge, manipulation of left/right eye assignment increases the binocular cormspondence of the HMD with the natural
world, and contour lines compensate for the unnatural continuity of binocular/monocular imagery and the black surrounding
Manipulation of eye assignment and/or contour lines appear to offer the most promise for improving the image quality
of a partial binocular-overlap HMD with the least cost or artifact. In either case, luning can be attenuated by implementing
ecological remedies. This approach generalizes to other aspects HMDs, such as color, stereo, and roll compensation. The
challenge with HMDs is to identify those designs that are ecologically valid and provide safe, economical, and useable
1. J.E. Melzer & K. Moffitt. “Partial binocular-overlap in helmet-mounted displays,” Proc. SPIE, Vol. 1117,
System Optics II, 56-62,
2. J. Flach. “An ecological alternative to egg-sucking,”
Human Factors Society Bulletin, 32,4-6,
3. K.J. Vicente. “A few implications of an ecological approach to human factors,”
Human Factors Society Bulletin, 33,
4. S. Shimojo & K. Nakayama. “Real world occlusion constraints and binocular rivalry,”
Vision Research, 30,69-80,
5. K. Nakayama & S. Shimojo. “Da Vinci stereopsis: Depth and subjective occluding contours from unpaired image
6. CAE Electronics, “Wide-Field-Of-View, Helmet-Mounted Infinity Display System Development”,
Brooks AFB TX, 1984.
7. E.M. Howlett, “Wide angle orthostereo,” Proc. SPIE, Vol. 1256,
Stereoscopic Displays and Applications,