Note: Descriptions are shown in the official language in which they were submitted.
CA 02310956 2000-06-27
1
COMPACT HIGH RESOLUTION PANORAMIC
VIEWING SYSTEM
This is a division of co-pending Canadian Patent Application Serial Number
2,277,002, filed on July 8, 1999.
Background of the Invention
Field of the Invention
The present invention relates to a viewing system; more particularly, a
panoramic viewing system.
Description of the Related Art
In an effort to operate more efficiently, it is desirable to perform some
tasks
using telepresence. Telepresence refers to providing visual or other types of
sensory
information from a device at a remote site to a user that makes the user feel
as if he/she
is present at the remote site. For example, many businesses now hold meetings
using
telepresence. Telepresence is also useful for distance learning and remote
viewing of
CA 02310956 2000-06-27
V. S. Nalwa 13
2
events such as concerts and sporting events. A more realistic telepresence is
provided to a
user by providing the user with the capability to switch between views, and
thereby
mimic, for example, looking around a meeting room.
In the past, when several views were made available to a user, several cameras
with different optical centers were used. Such a situation is illustrated in
FIG. 1. FIG. 1
illustrates cameras 2, 4, 6, and 8 with optical centers 10, 12, 14, and 16,
respectively.
When the user decided to change views, he or she simply switched between
cameras. In
more sophisticated systems, when a user decided to change views, he or she was
able to
obtain a view from optical centers 10, 12, 14, or 16 as well as from
additional optical
to centers 18, 20, 22, 24, or 26. Views associated with optical centers such
as 18, 20, 22,
24, and 26 were obtained by using views from the two cameras nearest to the
selected
optical center. For example, a view from optical center 18 was obtained by
using the
views from cameras 2 and 4 and interpolating between the two views so as to
simulate a
view from optical center 18. Such procedures introduced irregularities into
views. In
addition, forming these interpolated views required a large amount of
computational
power and time, and thereby made this technique expensive and slow to respond
to a
user's commands. This computational overhead also limited the number of users
that
could simultaneously use the system.
2o Summary of the Invention
One embodiment of the present invention provides an omnidirectional or
panoramic viewer in which multiple cameras effectively have a common optical
center at
least one of these cameras having its field of view redirected by a planar
mirror. The
field of view of each of the cameras is arranged to foam a continuous 360
degree view of
an area when taken as a whole. The user can sweep through 360 degrees of
viewing,
where each view has the same or nearly the same optical center, by simply
using the
output of one camera, more than one or the combination of cameras without
requiring the
computational overhead of interpolation used in the prior art. Such an
arrangement may
be used to enhance use of virtual meeting rooms by allowing a viewer to see
the meeting
CA 02310956 2000-06-27
V. S. Nalwa 13
room in a more natural format. This format corresponds closely to a person
sitting in the
actual meeting who simply turns his or her head to change the view at a
particular time.
In another embodiment of the invention, the cameras are positioned so that
they
each view a different reflective surface of a solid or hollow polyhedron such
as a solid or
hollow pyramid. This results in each camera having a virtual optical center
positioned
within the pyramid. The cameras are positioned so that their virtual optical
centers are
offset from each other. The offsets produce narrow blind regions that remove
image
distortions received from the edges of the pyramid's reflective surfaces.
In still another embodiment of the invention, a stereo panoramic view is
provided
to through the use of multiple virtual optical centers. A reflective
polyhedral element, such
as a pyramid, redirects the field of view of each camera in a first set of
cameras to form a
group of substantially co-located virtual optical centers at a first location
within the
pyramid. The pyramid also redirects the field of view of each camera in a
second set of
cameras to form a group of substantially co-located virtual optical centers at
a second
15 location within the pyramid. Panoramic images from the first and second
virtual optical
centers provide a stereo panoramic view when one panoramic image is provided
to a
user's left eye and the other panoramic image is provided to the user's right
eye.
In yet another embodiment of the present invention, polyhedrons such as
pyramids having reflective surfaces are stacked base to base or nested within
each other
2o to produce a compact panoramic viewer. Using multiple reflective
polyhedrons in such a
manner permits using many cameras with the same or nearly the same virtual
optical
center. Using many cameras divides a large viewing area into many smaller
areas where
an individual camera views each smaller area. Since each camera views a
smaller area,
increased resolution is provided to the user.
25 In another embodiment of the present invention, the reflective polyhedron
such as
a pyramid is supported by a post that passes through the vertex of the
pyramid. Cameras
are then mounted to the post to provide a panoramic viewer with a mounting
structure
and a structure for supporting individual cameras.
CA 02310956 2000-06-27
4
In still another embodiment of the present invention, a nearly spherical view
is
provided to a user by placing a camera at the common virtual optical center of
the
viewer. In order to enhance the spherical view, the camera at the common
virtual optical
center may use a wide angle lens.
In yet another embodiment of the present invention, the viewing device may
include any type of image processing device. If the image processing device is
a camera
or other type of image capture device, a panoramic image is captured for the
user, and if
the image processing device is a projector or other type of image producing
device, a
panoramic image is produced for the user.
In accordance with one aspect of the present invention there is provided a
panoramic viewing apparatus, comprising: a plurality of image processing
devices, each
having an optical center and a field of view; a reflective element being at
least partially
polyhedral having a plurality of reflective facets facing in different
directions, each of at
least two of the plurality of reflective facets redirecting a field of view of
one of the
plurality of image processing devices to create a plurality of virtual optical
centers; and
a support member intersecting an inner volume of the reflective element, the
reflective
element being secured to the support member and the plurality of image
processing
devices being secured to the support member.
Brief Description of the Drawings
The present invention taken in conjunction with the invention disclosed in co-
pending Canadian Patent Application Serial Number 2,277,002, filed on July 8,
1999,
will be described in detail hereinbelow with the aid of the accompanying
drawings, in
which:
FIG. 1 illustrates a prior art multiple camera viewing system;
FIG. 2 illustrates a four camera omnidirectional or panoramic viewing system
using a four-sided pyramid with reflective surfaces;
FIG. 3 illustrates how a reflective surface of the pyramid is used to provide
each
camera with a common optical center;
FIG. 4 illustrates the top view of the pyramid illustrating the camera
positions;
FIG. 5 illustrates an eight-sided pyramid with reflective side surfaces;
- CA 02310956 2000-06-27
r
4a
FIG. 6 is a top view of the pyramid of FIG. 5;
FIG. 7 is a block diagram of a system to control data produced by the cameras;
FIG. 8 illustrates the association between the data received from the cameras
and the view presented to a user;
FIG. 9 illustrates an addressing scheme for the memory of FIG. 7;
FIG. 10 is a block diagram of the controller of FIG. 7;
FIG. 11 illustrates the viewing system of FIG. 2 with a fifth camera;
FIG. 12 illustrates a top view of the pyramid of FIG. 2 with displaced virtual
optical centers;
FIG. 13 illustrates the pyramid of FIG. 12 with shades positioned in blind
regions;
FIG. 14 illustrates a panoramic viewer using pyramids stacked base to base;
' CA 02310956 2000-06-27
V. S. Nalwa 13
FIG. 15 illustrates a panoramic viewer using nested pyramids;
FIG. 16 illustrates a spherical viewer using nested pyramids;
FIG. 17 illustrates a stand used to support a panoramic viewer;
FIG. 18 illustrate two types of distortion;
5 FIG. 19 illustrates a calibration process;
FIG. 20 illustrates the association between data received from the cameras and
the
view presented to the user with distortion;
FIG. 21 illustrates how distorted image data is stored;
FIG. 22 illustrates how mapped image data is stored; and
1o FIG. 23 is a block diagram of a panoramic camera system where image mapping
is used.
Description of the Preferred Embodiment
FIG. 2 illustrates a four camera system for providing a 360 degree view to a
user,
where the cameras each have a common or nearly common virtual optical center
within
the pyramid. Pyramid 40 has reflective sides 42, 44, 46, and 48 and may be a
hollow,
solid or truncated structure. In a preferred embodiment, each of the
reflective sides
forms a 45 degree angle with a plane parallel to base 50 and passing through
the vertex of
pyramid 40. Cameras 52, 54, 56, and 58 are associated with pyramid reflective
surfaces
48, 42, 44, and 46, respectively. The cameras may be image gathering devices
such as an
optical scanner. As a result, camera 52 views a reflection from surface 48 to
enable it to
view objects in the direction of arrow 60. Camera 54 views a reflection from
surface 42
to view objects in the direction of arrow 62. Camera 56 views a reflection
from surface
44 to view objects in the direction of arrow 64, and camera 58 views a
reflection from
surface 46 to view objects in the direction of arrow 66. Each camera has a 90
degree field
of view; however, larger fields of view may be used and the overlapping
portion of the
images may be removed by deleting or combining the pixels associated with the
overlapping views. The combination of the four cameras viewing reflections
from their
associated reflective surfaces on pyramid 40, produce a 360 degree view of the
area
CA 02310956 2000-06-27
V. S. Nalwa 13
surrounding pyramid 40. When the mirrors are at 45 degrees with respect to the
pyramid
base, it is desirable to locate the optical center of each camera on a plane
that is parallel to
base 50 and intersects vertex 70 of pyramid 40. Each camera's optical center
should also
be located on a line that passes through vertex 70 and is perpendicular to the
base line of
the camera's associated reflective surface. For example, the optical center of
camera 54
is located on line 72. Line 72 is perpendicular to base line 74 of reflective
surface 42.
Line 72 is in a plane that passes through vertex 70 and is parallel to base
S0. Likewise,
the optical center of camera 56 is positioned on line 76 which is
perpendicular to baseline
78, the optical center of camera 58 is positioned on line 80 which is
perpendicular to base
to line 82, and the optical center of camera 52 is positioned on base line S4
which is
perpendicular to base line 86.
Each camera optical center is positioned on one of the above described
lines at a distance X from vertex 70 and each camera has its optical axes or
direction of
view pointing perpendicular to base 50. (The distance X should be such that
the
reflective surface reflects as much of the camera's field of view as desired;
however, the
defects in the reflective surface become more visible when the camera is moved
closer to
the reflective surface.) This positioning of optical centers results in the
cameras sharing a
virtual optical center located at, or substantially at, position 90. Virtual
optical center 90
is located a distance X from the vertex 70 on a line that passes through
vertex 70 and is
2o perpendicular to base 50.
Although a pyramid configuration has been discussed in this example, different
planar mirror geometries may be used to redirect fields of view so that the
cameras have
virtual optical centers that are substantially co-located. For example, solid,
hollow or
partial polyhedrons may be used. Additionally, in the case of a pyramid
configuration the
base and vertex do not have to be physically present and can be thought of as
conceptual
aids such as a base plane or end and vertex point or end.
FIG. 3 illustrates another view of pyramid 40 where only camera 54 is shown
for
the sake of simplicity. Camera 54 is positioned on line 72 so as to have a
virtual optical
center at, or nearly at, position 90 within pyramid 40. If camera 54 has a 90
degree field
CA 02310956 2000-06-27
V. S. Nalwa 13
of view in the direction perpendicular to base 50, and if the optical center
of camera 54 is
at a distance of X from vertex 70 along line 72, camera 54 has a 90 degree
view in the
direction of arrow 62. In similar fashion, cameras 56, 58, and 52 have 90
degree views in
the direction of arrows 64, 66, and 60, respectively. This arrangement
inexpensively
produces a 360 degree field of view of an area because cameras with a 90
degree field of
view have relatively inexpensive optics.
FIG. 4 is a top view of pyramid 40. FIG. 4 illustrates the placement of the
optical ,
center of camera 54 along line 72. Line 72 should be in a plane passing
through vertex
70 and is parallel to base 50. The line should also be perpendicular to base
line 74 of
to pyramid 40. The camera's optical center should be positioned a distance X,
or a distance
substantially equal to X, from vertex 70 along line 72. Point 100 is located
on base 50 at a
position where a line from vertex 70 perpendicularly intersects base 50. In a
similar
fashion, the optical centers of cameras 56, 58 and 52 are positioned the
distance X, or a
distance substantially equal to X, along lines 76, 80 and 84, respectively.
I S FIG. 5 illustrates an eight-sided pyramid 120. Pyramid 120 has reflective
surfaces
122 where each of surfaces 122 forms a 45 degree angle with a plane that
passes through
vertex 130 and is parallel to base 124. As with the four-sided pyramid of FIG.
2, each
reflective surface of FIG. 5 may have a camera associated with it. Each
camera's optical
center is positioned on a line that is in a plane that passes through vertex
130 and is
2o parallel to base 124. The line is perpendicular to base line 132 of the
reflective surface
associated with the camera to be positioned. Using an eight-sided pyramid
offers the
advantage of using cameras with only a 45 degree horizontal field of view to
obtain a 360
degree view. Cameras with only a 45 degree field of view have inexpensive
optics and
enable a 360 degree view to be constructed using relatively inexpensive
components.
25 FIG. 6 is a top view of pyramid 120. As discussed with regard to FIG. 5,
each
camera's optical center is positioned along a line 134 which is in a plane
that passes
through vertex 130 and is parallel to base 124. The optical centers are
positioned a
distance X, or a distance substantially equal to X, along line 134 which is
perpendicular
to the appropriate base line 132. Point 140 is on base 124 at the point of
intersection
CA 02310956 2000-06-27
V. S. Nalwa 13
between base 124 and a line that passes through vertex 130 and is
perpendicular to base
124.
Polyhedrons or pyramids having more or less reflective sides may be used. The
advantage of using pyramids having a large number of sides is that cameras
with
moderate to small fields of view may be used. Cameras with moderate fields of
view
have relatively inexpensive optics. The number of sides used in a pyramid is
somewhat
limited by the cost of providing a large number of cameras. A 360 degree view
of a scene
may be provided using a pyramid having three reflective sides. it may be
expensive to
use only a three-sided pyramid in order to provide a 360 degree field of view.
This
to embodiment of the invention uses three cameras each with a 120 degree field
of view,
and cameras with such a wide field of view use relatively expensive optical
components.
In applications where a full 360 degree view is not desired, it is possible to
build a
viewer that does not have a camera associated with each reflective surface of
the pyramid.
In addition to eliminating an unnecessary camera, it is also possible to
eliminate an
t 5 unnecessary pyramid polyhedron surface by using reflective elements that
are partial
pyramids or partial polyhedrons.
Although a pyramid configuration has been discussed in this example, different
planar mirror geometries may be used to redirect fields of view so that the
cameras have
virtual optical centers that are substantially co-located. For example, solid,
hollow or
2o partial polyhedrons may be used. Additionally, in the case of a pyramid
configuration the
base and vertex do not have to be physically present and can be thought of as
conceptual
aids such as a base plane or end and vertex point or end.
FIG. 7 illustrates a block diagram of a system for controlling data produced
by the
cameras of a viewing device such as the viewing device described in FIGS. 2
through 4.
25 Cameras 52, 54, 56 and 58 obtain a 360 degree view of an area via their
associated
reflective surfaces of pyramid 40. The image signal or output signal of
cameras 52, 54,
56 and 58 are passed through analog to digital converters (AID) 160, 162, 164,
and 166,
respectively. The output of the cameras can be thought of as a stream of
pixels and the
output of the A/Ds can be thought of as data representative of the pixels from
the
' CA 02310956 2000-06-27
V. S. Nalwa l3
- 9
cameras. The output of the A/Ds are passed through mux 170. Mux 170 allows the
pixel
data from each of the A/Ds to reach memory 172. Controller 174 cycles the
select lines
of mux 170 so that the outputs of all of the A/Ds are stored in memory 172.
Mux 170 is
switched at a rate that is four times the pixel rate of the cameras. If more
or less cameras
are used, the rate at which mux 170 is switched will be increased or slowed
accordingly.
It is also possible to eliminate mux 170 and to store the output of each AID
in a separate
memory. Controller 174 is implemented using a microprocessor which provides
control
signals to counters that control the switching of mux 170 and counters used to
provide
addressing to memory 172. The control signals to the counters include reset,
enable and
1 o a starting offset.
As a result of the pixel information being passed to memory 172, memory 172
contains a 360 degree view of a scene. Pixel information stored in memory 172
is passed
through digital to analog converter (D/A) 176 and to video display 178. The
actual
portion of memory 172 that is passed to video display 178 via D/A 176 is
controlled via
user input device 180. User input device 180 may be a common device such as a
mouse,
joystick, or keyboard. The user may simply lean a joystick to the right to
shift his view to
the right, lean the joystick to the left to shift the view to the left, or
leave the joystick in
the center to keep the view unchanged. Based on the input from user device
180,
controller 174 varies offsets and starting addresses that are used to provide
addressing to
2o memory 172.
FIG. 8 illustrates the relationship between the data provided by the cameras
and
the view available to the user. Since the cameras share a virtual optical
center, the view
can be thought of as a cylindrical view. Sector 200 can be thought of as
representing the
information provided by camera 52, sector 202 can be thought of as
representing the
information provided by camera 54, sector 204 can be thought of as
representing the
information provided by camera 56, and sector 206 can be thought of as
representing the
information provided by camera 58. The surface of the cylinder in each sector
can be
thought of as a collection of columns, where each column is composed of
pixels. For
example, sector 200 can be thought of as a collection of columns including
columns 210,
CA 02310956 2000-06-27
V. S. Nalwa 13
212, 214, and 216. Likewise, the output produced by camera 54 can be thought
of as a
collection of columns which include column 218 in sector 202 and the output of
camera
58 can include columns such as column 220 in sector 206.
FIG. 9 illustrates how memory 172 is divided to provide easy access to
different
5 views based on signals from user input device 180. Sections 230, 232, 234,
and 236
correspond to sectors 206, 200, 202, and 204, respectively. Each of sections
230, 232,
234, and 236 can be thought of as a block within memory 172. The blocks in
memory
172 are broken into columns of sequential addresses. The first column of
memory
segment 230 corresponds to the first column of pixels of sector 206. The
number of
1o memory positions associated with a column should be at least sufficient to
have one
location for each pixel in a particular column. For example, if a column of
pixels from
FIG. 8 includes 1000 pixels, each column associated with the memory segments
of FIG. 9
should have at least 1000 locations. The number of columns associated with a
particular
memory segment should be at least equal to the number of columns associated
with a
particular section of the cylinder of FIG. 8.
If a camera scans in a horizontal direction, sequential pixels are written in
adjacent columns, but possibly different rows, of a particular memory segment
by simply
changing an offset to a counter generated address. The overall write address
is generated
by adding the offset to the counter's output. This offset is changed at the
rate in which
2o the horizontally scanned pixels are received. After a horizontal scan is
completed, the
counter is incremented and once again the offsets are changed at the
horizontal scan rate.
As a result, when addressing a particular segment of memory during a write
cycle, the
columns are addressed by changing the offset at the horizontal pixel scan
rate, and
incrementing the counter at the vertical scan rate. This type of addressing
scheme is used
for accessing columns within each memory segment. When addressing different
memory
segments during a write cycle, a write segment offset is added to the sum of
the counter
output and the column offset. The write segment offset permits addressing
memory
segments 230, 232, 234, and 236. The segment offset is changed at the same
rate as mux
170 is switched.
V. J. l~a~wa '.r
CA 02310956 2000-06-27
11
Pixel data is read from memory 172 in a similar fashion. The sum of a counter
output and two sets of offsets are used to generate a read address. Once an
initial starting
column has been picked, the read address is generated by switching a read
column offset
at a rate that is equal to the horizontal scan rate of a video display. After
reading one
horizontal scans worth of data, the read counter is incremented and the read
column
offsets are changed at a rate equal to the horizontal scan rate of the
display. As a result,
the offset addresses are changing at the display's horizontal display rate and
the counter is
incremented at a rate equal to the vertical scan rate of a display. It is
possible to read data
out at a rate faster or slower than required by the video display; however, if
read out
to faster, a buffer memory should be used, if read out slower the video
display may appear
choppy to the viewer.
It should be noted that the cylindrical arrangement of pixels of FIG. 8 is
typically
displayed on a flat or nearly flat display. As a result, the image may be
displayed by
compensating for converting between a cylindrical surface and a flat surface.
This may
~ 5 be carned out using a simple conversion algorithm within a common digital
signal
processing integrated circuit. Methods for these types of conversions are well
known in
the art and can be found in "A Guided Tour of Computer Vision, Vishvjit S.
Nalwa,
Addison-Wesley Publishing Co., Reading, Massachusetts, 1993". It is also
possible to
carry out the conversion using a very high resolution display.
20 ~ It should be noted that if the view selected by a user corresponds
exactly to the
view of a particular camera, such as camera 52, columns 240-248 are read from
memory
170. Column 240 is the first column in segment 232 and column 248 is the last
column
in segment 232. If the user decides to move the view in a counter-clockwise
direction,
the start column will shift to the right so that the read operation begins at
column 246 and
25 ends at column 250. It should be noted that column 246 is the second column
associated
with memory segment 232 which has the pixel data from camera 52, and that
column 250
is the first column of pixel data associated with camera 56. As the user
shifts the view,
the starting column shifts in relationship to the user's commands. For
example, if the
user indicates that the view should shift in a counter-clockwise direction,
the start column
CA 02310956 2000-06-27
V. S. Nalwa 13
12
of FIG. 9 moves to the right, similarly, if the viewer indicates that the view
should shift in
a clockwise direction, the start column shifts to the left. As before, columns
are
addressed by using offsets, if the offsets involve moving between memory
segments, a
read segment offset is added to the sum of the column offset and counter
output.
FIG. 10 illustrates a block diagram of controller 174. Controller 174 includes
microprocessor 270 and memory 272. Memory 272 includes RAM and ROM. Processor
270 receives commands on line 274 from user input device 180. Microprocessor
270
controls start, stop and reset of counter 276. Counter 276 controls the select
lines of mux
170. Counter 276 counts at a rate that is four times the horizontal scan rate
of the
to cameras. Write address generator 278 provides write addressing for memory
172. Write
address generator 278 includes a counter, register for storing offsets and
adder for adding
the offsets and counter output. Microprocessor 270 controls the offset
selection and the
counters used by write address generator 278. The write addresses are formed
as
described with regard to FIG. 9. Read address generator 280 provides read
addresses to
15 memory 172. Read address generator 280 includes a counter, register for
storing offsets
and adder for adding the offsets and counter output. As with write address
generator 278,
microprocessor 270 controls the offset selection and the counters of read
address
generator 280. Microprocessor 270 also controls the starting column used by
the
counters based on inputs provided on line 274 from user input 180.
2o a The write and read addresses are provided to memory 172 separately if
memory
172 is implemented using a two port memory. If memory 172 is implemented with
a
single port memory, the write and read addresses are multiplexed to memory
172.
FIG. 11 illustrates the viewing system of FIG. 2 with a fifth camera. Camera
or
image gathering device 400 is located in pyramid 40 with the optical center of
camera
25 400 located at, or nearly at, virtual optical center 90. Camera 400 views
objects in the
direction of arrow 410. The resulting view coupled with the views of the
remaining four
cameras, provides a nearly spherical view. If the cameras of FIG. 11 are
replaced with
image producing devices, the nearly spherical viewing system becomes a nearly
spherical
projection system. It should be noted, that a camera or projection device, may
be placed
CA 02310956 2000-06-27
V. S. Nalwa 13
13
at the virtual optical center of viewing/projection devices having pyramids
with three,
four or more sides. It should also be noted that base edges 420 of the
reflective surfaces
should be beveled to avoid undesirable obstruction of camera 400's field of
view. It is
also possible to avoid undesirable image artifacts from base edges 420 by
moving camera
or image processing device 400. Device 400 should be moved so that device
400's
optical center is positioned away from virtual optical center 90 in the
direction of arrow
410. Device 400's optical center should be positioned so that the device's
used field of
view does not include edges 420.
FIG. 12 illustrates a top view of the pyramid of FIG. 2. In reference to FIG.
2,
to camera 52, 54, 56, and 58 have been moved upward in the direction of base
50. As a
result, virhial optical centers 500, 502, 504 and 506, which correspond to
cameras 52, 54,
56 and 58, respectively, are moved away from virtual optical center 90. It is
desirable to
move the virtual optical centers so that camera 52 captures an image between
lines 508
and 510 that are unaffected by an edge of the pyramid, camera 54 captures an
image
between lines 512 and 514 that are unaffected by an edge of the pyramid,
camera 56
captures an image between lines 516 and 518 that are unaffected by an edge of
the
pyramid, and camera 58 captures an image between lines 520 and 522 that are
unaffected
by an edge of the pyramid. This results in the cameras not capturing images
distorted by
edges of the pyramid from narrow planar shaped regions. In particular, planar
regions
524, 526, 528, and 530 are not used and form blind regions. This offers the
advantage of
removing image regions that are distorted by the edges of the reflective
pyramid.
Eliminating these portions of the fields of view alleviates the need to
provide image
processing that compensates for image artifacts at the edges. It is desirable
to keep
virtual optical centers 500, 502, 504, and 506 closely clustered so that
planes 524, 526,
528, and 530 are only as thin as necessary to avoid edge artifacts. By
maintaining such
thin planes, the need to process the images at their common boundaries is
removed while
minimizing the noticeable effect seen by a user.
FIG. 13 illustrates the pyramid of FIG 12 with shades 560, 562, 564, and 566
positioned in planar regions 524, 526, 528, and 530, respectively. The shades
reduce the
CA 02310956 2000-06-27
V. S. Nalwa 13
14
amount of unwanted light that enters the cameras. Similar shades may be placed
in blind
regions between device 400 and one or more of the other image processing
devices. It is
also possible to place a shade on base 50 with the edges of the shade
extending beyond
the edges of the base to reduce the amount of unwanted light that enters
cameras 52, 54,
56, and 58 from sources behind base 50.
FIG. 14 illustrates reflective pyramids 602 and 604 arranged in a base-to-base
configuration. The bases may be in contact with each other or spaced apart.
Reflected
pyramids 602 and 604 each have four reflective side facets. Pyramid 602 has
reflective
side facets 608, 610, 612, and 614. Reflective pyramid 604 has reflective
sides 616, 618,
620, and 622. Pyramid 602 includes vertex 624 and pyramid 604 includes vertex
626.
Vertices 624 and 626 are on a line 628 that is perpendicular to the base of
each pyramid.
Each pyramid has four image processing devices such as cameras with a field of
view
being redirected by a reflective surface. With regard to pyramid 602, a camera
with an
optical center positioned at point 630 has a field of view in the direction of
arrow 632
where that field of view is redirected by reflective surface 608. A second
camera with an
optical center at point 634 has a field of view in the direction of arrow 636
which is
redirected by reflective surface 610. A third camera with an optical center at
point 638
has a field of view in the direction of arrow 640 which is redirected by
reflective surface
612. A fourth camera with an optical center at point 642 has a field of view
in the
2o direction of arrow 644 which is redirected by reflective surface 614.
Regarding reflective
pyramid 604, a first camera with an optical center at point 646 has a field of
view in the
direction of arrow 648 which is redirected by reflective surface 616. A second
camera
with an optical center at point 650 has a field of view in the direction of
arrow 652 which
is redirected by surface 618. A third camera with an optical center at point
654 has a field
of view in the direction of arrow 656 which is redirected by reflective
surface 620. A
fourth camera with an optical center at point 658 has a field of view in the
direction of
arrow 660 which is redirected by reflective surface 622. The cameras
associated with
each of the pyramids are positioned in a way similar to how the cameras were
positioned
with regard to FIGS. 2, 3, 4, 11, and 12 so that each set of four cameras
shares a common
_ Tn-_.
CA 02310956 2000-06-27
V. S. Nalwa 13
l5
virtual optical center or have closely clustered virtual optical centers
within their
associated pyramid. Each set of cameras may also have offset virtual optical
centers
within their associated pyramid. The cameras may be positioned so that the
cameras
associated with each pyramid share a common virtual optical center along line
628 where
the bases of the two pyramids meet. It is also possible to position the
cameras so that
their offset virtual optical center are clustered about a point on line 628
where the bases
of the two pyranuds meet.
The structure of FIG. 14 increases the vertical field of view as compared to
the
viewers discussed with regard to FIGS. 2, 3, and 4. The viewer of FIG. 14
increases the
l0 vertical field of view by using two cameras rather than one camera for the
same or nearly
the same vertical dimension. It should be noted that a projector may be
constructed by
replacing the cameras with image producing devices. It should also be noted
that
reflective pyramids 602 and 604 may be rotationally misaligned with respect to
each
other. This misaligned relationship is obtained by rotating one or both of the
pyramids
t 5 about an axis that passes through the vertices of both pyramids. For
example, the axis
may be co-linear with line 628. As a result of this rotation, the side edges
of the
reflective side facets of pyramid 602 will not align with the side edges of
the reflective
side facets of pyramid 604.
Although a pyramid configuration has been discussed in this example, different
2o planar mirror geometries may be used to redirect fields of view so that the
cameras have
virtual optical centers that are substantially co-located. For example, solid,
hollow or
partial polyhedrons may be used. Additionally, in the case of a pyramid
configuration the
base and vertex do not have to be physically present and can be thought of as
conceptual
aids such as a base plane or end and vertex point or end.
25 FIG. 15 illustrates two reflective pyramids. Reflective pyramid 702 is
nested
within reflective pyramid 704. It should be noted that more than two
reflective pyramids
may be nested. For example, another reflective pyramid may be nested within
reflective
pyramid 702 and yet another reflective pyramid may be nested within the
pyramid that is
nested within pyramid 702. Vertex 706 of pyramid 702 and vertex 708 of pyramid
704
~~ ___
CA 02310956 2000-06-27
V. S. Nalwa 13
16
are on a line 710 which is perpendicular to the bases of both pyramids. Once
again, each
pyramid includes four image processing devices such as cameras each with a
field of
view that is redirected by a reflective surface of their associated pyramid.
Pyramid 702
includes reflective side facets 712, 714, 716, and 718. Reflective pyramid 704
includes
reflective side facets 720, 722, 724, and 726. Four cameras are positioned so
that their
field of view is redirected by the reflective surfaces of pyramid 702. A first
camera with
an optical center at point 730 and a field of view in direction of arrow 732
has its field of
view redirected by reflective surface 712. A second camera with an optical
center at
point 734 and a field of view in the direction of arrow 736 has its field of
view redirected
to by reflective surface 714. A third camera with an optical center at point
738 and a field
of view in the direction of arrow 740 has its field of view redirected by
reflective surface
716. A fourth camera with an optical center at point 742 and a field of view
in the
direction of arrow 744 has its field of view redirected by reflective surface
718. It should
be noted that pyramid 702 and its associated cameras are positioned so that
the field of
view of the cameras is not obstructed by pyramid 704. This is accomplished by
allowing
pyramid 702 to extend beyond the base of pyramid 704. Regarding pyramid 704, a
first
camera with an optical center at point 750 and a field of view in the
direction of arrow
752 has its field of view redirected by reflective surface 720. A second
camera with an
optical center at point 754 and a field of view in the direction of arrow 756
has its field of
2o view redirected by reflective surface 722. A third camera with an optical
center at point
758 and a field of view in the direction of arrow 760 has its field of view
redirected by
reflective surface 724. A fourth camera with an optical center at point 762
and a field of
view in the direction of arrow 764 has its field of view redirected by
reflective surface
726. The cameras associated with each of the pyramids are positioned in
accordance with
the positioning illustrated with FIGS. 2, 3, 4, 11, and 12 so that the eight
cameras share a
virtual optical center at position 770 or have closely clustered virtual
optical centers
within pyramid 702. Each set of cameras may also have offset virtual optical
centers
within pyramid 702.
- CA 02310956 2000-06-27
V. S. Nalwa 13
17
The panoramic viewer of FIG. 1 S can be provided with a ninth camera having an
optical center at point 770 and a field of view in the direction of arrow 772
to provide a
viewer with a partially-spherical view. The camera having an optical center at
position
770 may use a wide-angle lens to provide a broader view.
FIG. 16 illustrates the partially-spherical viewer of FIG. 15 with an
additional
camera having an optical center at point 780 and a field of view in the
direction of arrow
782 where that field of view is redirected by planar mirror 784. It should be
noted that
optical center 780 is on line 710 which passes through the vertices of pyramid
702 and
704 as well as virtual optical center 770. It should also be, noted that point
780 is placed a
1o distance away from planar mirror 784 that is equal or nearly equal to the
distance between
planar mirror 784 and virtual optical center 770. By placing a camera with an
optical
center at point 780 and having the field of view redirected by planar mirror
784, the
partially-spherical viewer of FIG. 15 becomes a spherical viewer. In order to
increase the
field of view of the camera positioned with an optical center at point 780,
the camera may
15 be provided with a wide-angle lens. It should also be noted that planar
minor 784 may be
replaced with a curved mirror to provide a wider field of view for the camera
positioned
at point 780 and minimize the need for a wide-angle lens.
Although a pyramid configuration has been discussed in this example, different
planar mirror geometries may be used to redirect fields of view so that the
cameras have
2o virtual optical centers that are substantially co-located. For example,
solid, hollow or
partial polyhedrons may be used. Additionally, in the case of a pyramid
configuration the
base and vertex do not have to be physically present and can be thought of as
conceptual
aids such as a base plane or end and vertex point or end.
Regarding FIGS. 15 and 16, it should be noted that a projector may be
25 constructed by replacing the cameras with image producing devices. It
should also be
noted that reflective pyramids 702 and 704 may be rotationally misaligned with
respect to
each other. This misaligned relationship is obtained by rotating one or both
of the
pyramids about an axis that passes through the vertices of both pyramids. For
example,
the axis may be co-linear with line 710. As a result of this rotation, the
side edges of the
CA 02310956 2000-06-27
V. S. Nalwa 13
18
reflective side facets of pyramid 702 will not align with the side edges of
the reflective
side facets of pyramid 704.
FIG. 17 illustrates a stand used to support a panoramic viewer. Reflective
pyramid 800 is mounted to stand or post 802 using a support member such as
hollow tube
804. The pyramid is secured to hollow tube 804 at vertex end 806. The hollow
tube is
secured to stand 802 by angle brackets 808. Hollow tube 804 extends beyond
vertex end
806 so that cameras 810 may be supported by tube 804. The cameras are mounted
to tube
804 by strap or belt 812 which presses cameras 810 against spacer 814. The
pressure
provided by clamp or strap 812 provides friction between camera 810, spacer
814, and
1o the outer surface of tube 804 and thereby mounts cameras 810 to tube 804 in
a secure
fashion. It is also possible to provide a second strap and associated spacers
at end-section
816 of cameras 810. Video and power connections to cameras 810 are provided by
cables 818 which are fed through hollow tube 804 and out through space 820
which is
between post 802 and the base of pyramid 800. It should be noted that hollow
tube 804
may be replaced with a solid support member; however, a hollow support member
provides a convenient path for routing cables. Feeding the cables through tube
804
prevents the cables from entering the field of view of cameras 810. Rubber
stands or feet
824 are provided at the base end of pyramid 800. These stands may be used in
place of
post 802 to provide flexibility in application where the user does not want to
use post
802.
It is also possible to invert the viewer of FIG. 17 so that the viewer is
supported
by end 830 of tube 804. In this configuration cables 818 will simply be passed
out
through an opening at end 830 of tube 804. In this configuration tube 804 is
mounted to
post 802 at end 830 using angle brackets similar to angle brackets 808. It is
also possible
to mount end 830 to any convenient structure to support the panoramic viewer.
The stand of FIG. 17 is applicable to the viewer of FIGS. 14, 15, and 16. As
discussed with regard to FIG. 17, the viewer is mounted to a hollow tube
passing through
the vertices or vertex ends of both pyramids.
_' CA 02310956 2000-06-27
V. S. Nalwa 13
19
Calibration
A higher quality image may be produced by calibrating the camera system.
Calibration may be used to determine image mapping functions (which may be
implemented as look up tables) that compensate for different types of image
distortion.
For example, the mapping functions may be used to correct barrel distortion
which is a
distortion introduced by a wide-angle lens. Mapping functions may also be used
to
correct other types of distortions such as a rotational distortion resulting
from misaligned
charged coupled devices within the cameras. FIG. 18 illustrates a combination
of barrel
distortion and rotational distortion, where the distortion results in
rectangular object 900
appearing as distorted image 902. Distorted image 902 is rotated with respect
to
undistorted object 900 and a barrel distortion is seen where edges 904 and 906
of
rectangular object 900 appear as edges 908 and 910 of image 902, and where
edges 912
and 914 appear as edges 916 and 918, respectively. This distortion may be
corrected
using mapping functions that are determined by calibrating the camera system.
FIG. 19 illustrates a process for calibrating the camera system. A vertical
column
of equally spaced elliptical dots 930 is placed in a fixed position. The dots
are white on a
black background where the major axis of the elliptical dot is in the vertical
direction.
Panoramic camera 940 is then rotated in small discrete steps about an axis 942
passing
through the virtual optical center of the panoramic camera. At each step, the
distorted
2o image of the column of elliptical dots is viewed and a mapping function for
that data is
determined to remove the distortion. This function maps the image such that
the distorted
image of each vertical column of equally spaced dots is a vertical column of
equally
spaced dots in the mapped image. Note that although the images of the white
dots have
black gaps between them, the mapping function is computed to apply to every
image
pixel (including the pixels between the white dots) through interpolation. An
image
mapping function is determined at each of the discrete steps as the camera is
rotated, the
union of these mapping functions is combinable into a 2-D to 2-D mapping that
ensures
not only that each vertical column of equally spaced dots appears as a
vertical column of
equally spaced dots in the image, but also that these columns are spaced
horizontally in
CA 02310956 2000-06-27
V. S. Nalwa 13
proportion to the angular rotation between their image acquisitions, the
latter providing a
cylindrical (rather than flat) image of the scene.
FIGS. 20 and 21 illustrate how the data representative of the vertical column
of
elliptical dots is represented as a result of distortion. FIG. 20 is similar
to FIG. 8 in that it
illustrates the relationship between the data provided by the cameras and the
view
available to the user. It should be noted that the vertical column of dots is
not in a single
column of FIG. 20 as a result of the distortion. The distortion has caused the
dots to
occupy columns 960, 962, 964 and 966, rather than just a single column. FIG.
21 is
similar to FIG. 9 in that it illustrates how image data is stored. When the
distorted image
to data is stored in memory as represented in FIG. 21, the data representative
of the dots also
occupies several columns where columns 980, 982, 984 and 986 correspond to the
columns 960, 962, 964 and 966 of FIG. 20, respectively. The image mapping
function
determined during the calibration phase is used to correct for this distortion
when the data
is read from the memory represented in FIG. 21. The corrected or undistorted
image data
15 may then be displayed to the user or written to a memory used to store data
representative
of the undistorted image. FIG. 22 illustrates the relationship between the
data read from
the memory of FIG. 21 and the undistorted view made available to a user. For
example,
the mapping function associated with column 1000 specifies that when reading
data for
use in the uppermost portion of column 1000, data is read from column 980 and
when
20 reading data for use in the portion of column 1000 just below, data is read
from column
982. The mapping function also specifies that when reading data for use in the
middle
portion of column 1000, data is read from column 984. Moving further down
column
1000, data is then read from column 982, then column 980, and eventually from
column
986 when data for use at the bottom of column 1000 is retrieved. As a result
of reading
data, as specified by the mapping function, the column of data will appear
vertical to a
user viewing a display. FIG. 22 illustrates that the data retrieved from the
memory of
FIG. 21 now appears as a vertical column where the distortion is no longer
evident. A
similar mapping function, as determined during calibration, is used for each
column of
FIG. 22 to produce an undistorted image for display. It should be noted that
multiple
CA 02310956 2000-06-27
V. S. Nalwa 13
21
discrete rotational steps used to calibrate the panoramic camera could be
substituted by a
group of several columns illustrated in FIG. 22.
Color and intensity calibration may also be carried out using a procedure
similar
to the procedure illustrated in FIG. 19. In this case, column 930 of
elliptical dots is
replaced by a known color pattern. The panoramic camera is then rotated so
that each
camera captures an image of the color pattern. Several color patterns (such as
various
shades of red, green, blue, and gray) could be used one by one. Then on a
pixel-by-pixel
basis, the data from each camera is adjusted to correct any red, green, or
blue distortion so
that the produced image has a color pattern that closely matches the
calibration color
1o pattern. Additionally, the intensity of each pixel from each camera is
adjusted so that
there is relatively uniform intensity and color within a single camera's image
and
between the images of the multiple cameras when viewing a scene with constant
color
and brightness. As discussed with regard to the mapping function, the pixel-by-
pixel
adjustment may be stored in a table. A less precise, but simpler method of
color and
15 intensity calculation may be used. This method simply involves manually
adjusting the
color and intensity controls of each camera to get correct color and intensity
when
viewing a scene with a particular color and intensity. It should be noted that
by using this
method, all of the pixels of a particular camera receive the same adjustments.
FIG. 23' illustrates a panoramic camera system where calibration based image
2o mapping correction is used. FIG. 23 is similar to FIG. 7; however, it
should be noted that
a frame buffer memory and an additional microprocessor have been included.
Cameras
52, 54, 56, and 58 gather image data and then pass the data to analog-to-
digital converters
160, 162, 164, and 166, respectively. The output of the analog-to-digital
converters are
then passed through red, green, blue, and intensity adjustment units 1010,
1012, 1014,
25 and 1018. It is possible to place these units before the analog/digital
converters, if the
adjustment units are analog units. Additionally, it is also possible to use
cameras that
have the adjustment units built into each camera. In any case, the adjustment
units are
programmed or set to adjust the color and intensity as determined by the
calculation
procedures. Each of these units adjust the red, green, and blue levels and the
overall
CA 02310956 2000-06-27
V. S. Nalwa 13
22
levels of the signals from the analog-to-digital converter. It should be noted
that if
cameras 52 through 58 are color cameras, analog-to-digital converters 160 to
166
typically receive three signals and output three signals, where each pair of
input and
output signals corresponds to one of the colors red, green, and blue. Units
1010 through
1016 simply adjust the relative amplitudes of the red, green, and blue signals
in
accordance with the settings determined during the calibration procedure. Each
of units
1010 through 1018 also adjust the overall amplitude of the red, green, and
blue signals in
accordance with the overall intensity calibration settings. The outputs of the
red, green,
and blue intensity adjustments are then passed through a multiplexer as
discussed in FIG.
l0 7, and are passed to frame buffer memory 1030. It is also possible to
replace frame buffer
1030 with an individual frame buffer for each of red, green, blue and
intensity units 1010,
1012, 1014, and 1018. The outputs of each of the individual frame buffer may
then be
passed to microprocessor 1030 via multiplexer 170.
Frame buffer memory 1030 is operated in a fashion similar to memory 172 of
FIG. 7 and stores the data representing the distorted images as was discussed
in reference
to FIG. 21. Microprocessor 1040 then reads the data from frame buffer memory
1030
using the mapping functions determined during the calibration procedure and
then writes
the data into display memory 1050. Recalling the discussion associated with
FIG. 22, the
data representing undistorted images is then stored in memory 1050 for
retrieval by the
users. The users can retrieve the data as is discussed in reference to FIG. 7
where the data
read out is determined based on a user's input. It is also possible for the
entire contents
of display memory to be made available to each user. The data may be
communicated to
each user through a communication network such as a telephone network or a
data
network, or it may be directly communicated to the user via a dedicated wired
or wireless
communication path. The user then may use a digital-to-analog converter to
convert the
data into an analog format that may be displayed for the user or the user may
use the
digital data directly and forego the use of a digital-to-analog converter.
