A Long-range Vision System for Projection Mapping of Stereoscopic

Content in Outdoor Areas

Behnam Maneshgar

, Leila Sujir

, Sudhir P. Mudur

and Charalambos Poullis

Department of Computer Science and Software Engineering, Concordia University, Montreal, Canada

Department of Studio Arts, Concordia University, Montreal, Canada

Keywords:

Stereoscopic Projection, Outdoor Projection Mapping, Long-range Projection.

Abstract:

Spatial Augmented Reality, or its more commonly known name Projection Mapping (PM), is a projection tech-

nique which transforms a real-life object or scene into a surface for video projection (Raskar et al., 1998b).

Although this technique has been pioneered and used by Disney since the seventies, it is in recent years

that it has gained signiﬁcant popularity due to the availability of specialized software which simpliﬁes the

otherwise cumbersome calibration process (Raskar et al., 1998a). Currently, PM is being widely used in ad-

vertising, marketing, cultural events, live performances, theater, etc as a way of enhancing an object/scene by

superimposing visual content (Ridel et al., 2014). However, despite the wide availability of specialized soft-

ware, several restrictions are still imposed on the type of objects/scenes on which PM can be applied. Most

limitations are due to problems in handling objects/scenes with (a) complex reﬂectance properties and (b) low

intensity or distinct colors. In this work, we address these limitations and present solutions for mitigating these

problems. We present a complete framework for calibration, geometry acquisition and reconstruction, esti-

mation of reﬂectance properties, and ﬁnally color compensation; all within the context of outdoor long-range

PM of stereoscopic content. Using the proposed technique, the observed projections are as close as possible

[constrained by hardware limitations] to the actual content being projected; therefore ensuring the perception

of depth and immersion when viewed with stereo glasses. We have performed extensive experiments and the

results are reported.

1 INTRODUCTION

Many successful techniques have been developed for

capturing and modeling the shape and reﬂectance

properties of objects/scenes (Torrance and Sparrow,

1967), (Phong, 1975), (Oren and Nayar, 1995),

(Lafortune et al., 1997). These have been particu-

larly successful in cases where the acquisition is per-

formed under controlled lab conditions without the

presence of any dynamic elements. Although PM can

also be used indoors in a similar fashion, the majority

of its applications involve large-scale and/or outdoor

objects/scenes. Perhaps the only work reported in the

literature to address the capture of complex geometry

and the estimation of reﬂectance properties of outdoor

objects is by Debevec et al. (Debevec et al., 2004).

In an outdoor setting there are several challenges:

(a) there is no control over the lighting, (b) there is no

control over other dynamic elements which may be

present in the scene e.g. people walking, cars pass-

ing by, clouds, rain, etc, (c) most often limited time is

available to perform the capture because of the afore-

mentioned challenges. An example of one of our out-

door projection mapping experiments is shown in Fig-

ure 1(a). The viewers were wearing stereo glasses

in order to perceive depth since the projected content

was stereoscopic, as shown in Figure 1(b). Objects

with complex reﬂectance properties such as the win-

dows and columns are not speciﬁcally handled during

the projection and resulted in color distortions to the

red-cyan stereo content which further caused the loss

of depth perception and some noticeable visual arti-

facts.

In this paper, we address the problem of long

range projection mapping of stereoscopic content on

outdoor areas and propose a complete framework

which automates the following processes: (a) system

calibration, (b) structure and appearance information

acquisition, (c) approximating model of projection

surface’s reﬂectance properties, (d) color compensa-

tion to the extent possible with the given projection

surface. The result is compensated image/video con-

tent such that its projection onto the particular surface

will produce an image/video which, when viewed,

290

Maneshgar B., Sujir L., Mudur S. and Poullis C.

A Long-range Vision System for Projection Mapping of Stereoscopic Content in Outdoor Areas.

DOI: 10.5220/0006258902900297

In Proceedings of the 12th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications (VISIGRAPP 2017), pages 290-297

ISBN: 978-989-758-224-0

(a)

(b)

Figure 1: (a) The facade of the Roman Baths, UK. (b)

Stereoscopic content being projected onto the facade.

will seem as close to the original as possible. This

is, of course, limited by the projection surface prop-

erties, as it may not always be possible to completely

compensate for the surface reﬂectance behavior.

The paper is organized as follows: Section 2 gives

an overview of the state-of-the-art in the area and Sec-

tion 3 presents a technical overview of the proposed

framework. In Section 4 we discuss the calibration

of the cameras and the projector with respect to each

other, and compute both extrinsic and intrinsic param-

eters for them. Next, Section 5 describes the geometry

acquisition. The reﬂectance properties are modelled

to their best approximation as described in Section 6.

Lastly, Section 7 presents the color compensation and

illustrative experimental results are provided in Sec-

tion 8.

2 RELATED WORK

R. R. Garcia and A. Zakhor (Garcia and Zakhor,

2013) calibrated a multi-camera-projector system to

set up a multi-view structure light. By taking the ad-

vantage of binary codes they encode each projector

pixel. By projecting the patterns on a screen and cap-

turing the images sequence by the cameras and de-

coding the binary codes, they generate dense corre-

spondences between the cameras and projector. Bun-

dle adjustment is performed to calibrate the system.

Svoboda et al. (Svoboda et al., 2005) also proposed a

fully multi-camera self-calibration method for virtual

environments. They wave a detectable bright spot in

the working volume and capture the volume with at

least three synchronised cameras. These points are

validated through pairwise epipolar constraints. They

calibrate the system by using these correspondences.

3D reconstruction is a well-studied area in com-

puter graphics. Many techniques are proposed and

many commercial products are available on the mar-

ket. Microsoft Kinect is one of the most popular de-

vices and comes with game consoles. It is able to scan

the scene/object with high accuracy in short range. Y.

Furukawa and J. Ponce (Furukawa and Ponce, 2010)

proposed an algorithm to reconstruct an object/scene

by using calibrated multi-view stereo. The algorithm

detects feature points of each image and then ﬁnds the

matches between each pose and outputs a dense set of

patches covering the surface of the object/scene.

The diffuse component and specularity or reﬂec-

tion component of the scene is a very important issue

in computer graphics, computer vision and in general

vision systems. Typically, techniques try to ﬁrst de-

tect the specularity of the scene and then cancel or

reject them as an outlier. Several methods with differ-

ent viewpoints proposed to tackle this problem, Sin-

gle image techniques and multiple image techniques

have been proposed to extract diffuse map and specu-

lar map of the scene. Lin et al. (Lin et al., 2002) pro-

posed a color-based method to identify and separate

specular component from an input image sequence

by using a multi-baseline stereo system. Fries et al.

(Feris et al., 2004) proposed a multi-ﬂash method to

achieve this separation. They used a ﬁxed camera and

a varying ﬂashlight to capture images of the scene

with different light source positions. Seitz et al.(Seitz

et al., 2005) proposed a method for the cancellation

of n-bounced inter-reﬂection light. In this, they ﬁrst

proved the existence of a set of linear operations by

applying the inverse light transportation theory. Then,

by probing the scene using a very narrow beam of the

light, can compute the operators. Michael D. Gross-

berg et al. (Grossberg et al., 2004) introduced an im-

proved radiometric model to control the appearance

of an object in short-range. By using a camera and a

projector, they try to make the captured image similar

to the projected image.

The above techniques have been largely applied

A Long-range Vision System for Projection Mapping of Stereoscopic Content in Outdoor Areas

291

to short-range capture and monocular projection. The

work reported in this paper is mainly concerned with

long range and stereoscopic projection. We present

our investigations in extending the above techniques

and/or development of new techniques, as may be

needed, due to the new problems posed by long-range

capture, calibration and projection.

3 SYSTEM OVERVIEW

An overview of our system is shown in Figure 2. In

the ﬁrst stage, the system is calibrated. This includes

calibration of the individual cameras, calibration of

each camera with respect to other cameras, and the

calibration of the projector with respect to the cam-

eras. Next, the geometry of the projection surface is

captured using a structured-light scanning technique.

Using the images and geometry of the surface, the re-

ﬂectance properties at each point are estimated. Sur-

face points with complex reﬂectance properties i.e.

transparent or translucent are identiﬁed. Finally, the

original stereoscopic content is compensated to ac-

count for the reﬂectance properties of the projection

surface prior to projecting it on the surface.

4 SYSTEM CALIBRATION

Accurate calibration is of imperative importance

when dealing with long-range vision-projection sys-

tems, such as in this case. A small error in image

space i.e. pixels, can lead to vast displacements in

the projected space. In this section we describe our

system calibration process, which involves:

• the calibration of the cameras

• the pose recovery with respect to the cameras and

intrinsic parameter calibration of the projector

4.1 Camera Calibration

Perhaps the most popular technique for calibrating a

camera is the one proposed by Tsai et al. (Tsai, 1987)

and Zhang et al. (Zhang, 2000). Given a set of points

in world space and their corresponding image points,

one can recover both the intrinsic and extrinsic pa-

rameters of the camera. The pinhole camera model is

used to describe these parameters which are speciﬁed

by the camera matrix C in equation 1,

C =





α −αcot(θ) u

sin(θ)

0 0 1





{z }

intrinsic









| {z }

extrinsic

(1)

where α = k f

, β = k f

, ( f

, f

) is the focal length

on the x and y axis respectively, θ is the skew angle,

is the principal point on the x and y axis respec-

tively and r

1−3

x−z

determine the camera’s rotation

and translation relative to the world. Lens distortion

is also taken into account and is modeled by:

D =





(2)

which Bell et al. (Bell et al., 2016) explained in

their work.An inherent assumption during camera cal-

ibration using the above method is that the captured

images of 3D objects with known calibration, say, a

checkerboard, are in-focus. This is indeed the case

for many applications. However, when dealing with

projection mapping in outdoor areas this is not the

case. The cameras and projector are focused on the

projection surface, which is far away. Calibrating the

system using a traditional technique means that the

checkerboard must be also positioned at the same far

distance. Although this does not pose a physical lim-

itation, it most often leads to inaccurate estimation

of the parameters. The reason being that at greater

distances the checkerboard will only occupy a very

small area of the captured image, therefore making

the distribution of world-image correspondences de-

generate. This leads to erroneous calculations. In par-

ticular, distortion parameters cannot be accurately re-

covered in the case where the captured images of the

checkerboard do not provide good coverage across the

entire area covered by the camera. In order to over-

come this problem, we follow an approach similar to

Tyler Bell et al. (Bell et al., 2016) which, instead of

a calibration object, uses projected patterns which are

by-design robust to out-of-focus cameras.

This method encodes feature points into phase

shifted patterns being displayed on a monitor visible

to the cameras. The feature points can then be accu-

rately decoded even when blurred because this does

not affect the phase of the pattern sequence. One

vertical and one horizontal phase map are required

where each vertical/horizontal line has a unique phase

value. Thus, each pixel appearing on the monitor has

a unique pair of (Φ

,Φ

) to identify the feature. These

phase maps are carried by the phase shifting patterns.

Equation 3 is used to generate N equally phase-shifted

vertical and horizontal fringe patterns and is given by,

(u,v) = 0.5



1 + cos (Φ

+ 2iπ/N)



(u,v) = 0.5



1 + cos (Φ

+ 2iπ/N)



(3)

GRAPP 2017 - International Conference on Computer Graphics Theory and Applications

292

Figure 2: Pipeline

where i is the index of fringe pattern, Φ

and Φ

represent the vertical and horizontal phase maps, re-

spectively. Each phase-map is then extracted as fol-

lows,

φ(x,y) = tan

−1



∑

i=1

sin(2iπ/N)

∑

i=1

cos(2iπ/N)



(4)

where I

is the intensity of a speciﬁc pixel in the

captured image. This equation generates a non-

continuous wrapped phase map with values in the

range of [−π,π]. Next, the phase-maps are un-

wrapped to produce a unique phase value for each

of their columns/rows in the pattern. Adding an off-

set to each section of the phase-map generates an un-

wrapped phase-map which has a unique phase value

at each horizontal/vertical pixel line as shown in the

following equation,

Φ(x,y) = φ(x,y) + k × 2π (5)

The process can be summarized as shown in Algo-

rithm 1. Figure 3 shows the encoding of the features

using the vertical and horizontal phase maps.

After decoding of the correspondences, the cam-

era can be calibrated using the traditional technique.

The result is the intrinsic and extrinsic parameters for

each camera. The extrinsic parameters are given with

respect to the ﬁrst (top-left) encoded feature point in

the monitor.

Algorithm 1: Camera calibration and pose estimation

using phase shifting.

1 generate vertical and horizontal fringe patterns

2 display and capture images from monitor at

different poses

3 compute wrapped phase-map

4 calculate k in Equation 5

5 calculate unwrapped phase map

6 match the encoded feature points with the

decoded phase map

7 calibrate camera using the traditional technique

Figure 3: This ﬁgure shows the encoding process of the

feature points.

4.2 Projector Calibration

The projector is treated as an inverted camera and is

calibrated with the traditional method using a set of

2D to 3D correspondences resulting from the geome-

try acquisition. The explanation on the extraction of

the 2D-3D correspondences is deferred to Section 5.

Given a set of 3D world points and their correspond-

ing 2D image locations, the projector’s intrinsic and

extrinsic parameters are recovered.

5 GEOMETRY ACQUISITION

The geometry of the projection surface directly im-

pacts the appearance of the surface as well as the pro-

jected content. Identifying surface points with com-

plex reﬂectance properties requires that the geometry

(surface points and normals), and reﬂectance proper-

ties (response to light) of the projection surface are

known.

Structured-light scanning (Herakleous and

Poullis, 2014) is used to capture the geometry of the

projection surface. Encoded patterns are projected

onto the surface in sequence. The cameras capture

one image per pattern. These images are decoded

to produce a dense correspondence between the

projector’s pixels and the camera’s pixels. A render

A Long-range Vision System for Projection Mapping of Stereoscopic Content in Outdoor Areas

293

of the reconstructed geometry of the Roman Baths

(Figure 1) is shown in Figure 4. This was generated

from 44 images captured by three cameras.

Figure 4: A render of the reconstructed geometry of the

Roman Bath. 44 images captured by three cameras.

6 ESTIMATION OF SURFACE

REFLECTANCE PROPERTIES

As part of the geometry acquisition images are cap-

tured from each camera from different viewpoints.

Each image records the brightness at every visible

surface point from that particular viewpoint. Given

three such measurements per surface point, we ap-

proximate the local reﬂectance properties.

We use the Phong illumination model, which is a

local illumination model that is easy to compute and

use, to describe the local interactions between the ma-

terial and the light and is given by:

I = Lκ

cosθ + Lκ

cosφ

(6)

where I is the brightness of the reﬂected light as

recorded by the camera, L is the incident radiance

emitted from the projector, K

is the 3-vector diffuse

reﬂection coefﬁcient, K

is the 3-vector specular re-

ﬂection coefﬁcient and α is the shininess coefﬁcient

of the material.

A non-linear optimization (Levenberg, 1944) is

used to compute the optimal values for the material

parameters such that the energy function E( f ) is min-

imized,

E( f ) = E

data

( f ) + E

smooth

( f ) (7)

where E

data

( f ), the energy data term and E

smooth

( f ),

the energy smoothness term are as deﬁned below.

Energy data term E

data

( f ): This term is a measure

of how appropriate the optimized material parameters

are given the observed data and is deﬁned as,

data

( f ) =

∑

i=0

− I

(8)

where n is the number of cameras, I

is the ren-

dered image as viewed from camera i using the ac-

quired geometry and the material parameters being

optimized, and I

is the observed image captured by

camera i.

Energy smoothness term E

smooth

( f ): This term is a

measure of the smoothness between brightness values

in neighbouring pixels of the rendered image and is

given by,

smooth

( f ) =

∑

i=0



w×h

∑

j=0

∑

− B



(9)

where n is the number of cameras, j is the num-

ber of pixels within the rendered image from cam-

era i, m is the 8-neighbourhood around the pixel j,

is the brightness at pixel j, and B

is the bright-

ness at pixel m. This term ensures that the opti-

mal values will provide smooth results. This is illus-

trated in Figure 5 which shows a comparison between

the smoothed/non-smoothed procedures for comput-

ing material coefﬁcients. For the statue in Figure 5a

the specular map without smoothing can be seen in

Figure 5b and with smoothing in Figure 5c. As ev-

idenced, without the smoothness term there is noise

between neighbouring pixels which is removed when

the smoothness term is introduced.

Figure 6 shows the results of the energy minimiza-

tions for three synthetic test/validation cases shown in

Figure 6a: a perfectly specular, a diffuse/specular and

a perfectly diffuse sphere, respectively. Figures 6b,

6c, 6d show the progress of error minimization corre-

sponding to the shininess, diffuse and specular coefﬁ-

cient respectively.

7 COLOR COMPENSATION

Changes in the appearance of the projected content

caused by the reﬂectance properties of the projection

surface need to be compensated. We use the estimated

geometry and reﬂectance properties at each surface

point to compensate the content prior to its projec-

tion. For each frame of the stereoscopic content to-be-

projected we calculate the compensated projector’s

brightness L

for each pixel p as follows,

= L



cos(θ) + κ

cos(φ)





cos(θ) +κ

cos(φ)



(10)

where L

is the original image’s brightness value

at pixel p, L

is the compensated image which, when

projected on the surface, ideally yields L

GRAPP 2017 - International Conference on Computer Graphics Theory and Applications

294

(a) (b) (c)

Figure 5: (a) The object/projection surface. (b) The estimation of the specular parameters without the smoothness term and

Due to the limited range of brightness values the

projector can produce there is a restriction on the col-

ors for which compensation will work.

Further, in this work, the content is stereoscopic

which already uses a limited range of values because

of the anaglyph processing. We have found that this

works well for stereoscopic content being projected

on outdoor surfaces such as building facades where

the surface is primarily diffuse with strong specular

components in the presence of windows, light-ﬁxtures

and other such objects.

As previously mentioned, the reﬂected color from

the object depends on the material’s reﬂectance prop-

erties and the emitted light. Knowing the reﬂectance

properties of each surface point allows us to compen-

sate up to a factor. First, using additive color mix-

ing we calculate an image which when projected can-

cels out [if needed] the colors on the projection sur-

face and makes the surface appear grayish. Next, we

calculate the image which when projected on top of

the ’grayish’ surface will be as close to the original

as possible. The result is the original image with in-

creased brightness and reduced contrast depending on

how bright the grayish image has to be.

8 EXPERIMENTAL RESULTS

AND CONCLUDING REMARKS

The proposed technique has been tested and the re-

sults are presented. All reported results were gener-

ated on an Intel-i7 PC with commodity hardware. The

projector used was a Panasonic PT-VW435N projec-

tor with a native resolution of WXGA 1280 × 800.

The statue of Alexander shown in the experiments

has dimensions 25.5cm × 19cm × 14.5cm and was

used for comparison purposes with (Herakleous and

Poullis, 2014).

A 3D print of the Roman Baths was used for ex-

perimentation due to access restrictions on the real

site. These experiments were conducted in relative

scale. Figure 1 shows a projection on the real Roman

Bath’s building. Figure 7a shows a stereoscopic pro-

jection [without compensation] being projected onto

the Roman Baths. Color distortions occurring on the

anaglyph image due to the reﬂectance properties of

the projection surface have a negative impact on the

depth perception of the viewer. Figure 7b shows the

stereoscopic projection after compensation using the

proposed technique. Color distortions are minimized

by taking into account the effect of the reﬂectance

properties of the projection surface. The expected

projection is shown in Figure 7c.

The expected projection cannot always be

achieved [as in the above case] because of the limi-

tations of additive color mixing and the hardware. In-

tuitively, a projection surface with a white-ish color

can reﬂect a larger percentage of the projected light

by the projector, therefore, more colors can be com-

pensated. For example, a bright red color projected

onto a white surface will appear as red. On the other

hand, a projection surface with a darker color will ab-

sorb the projected light; a bright red color projected

onto a dark surface will appear as dark red.

For the immediate future, we will investigate

perception-driven compensation, by attempting to

preserve major visual features. In addition, we will

explore the application of different reﬂectance mod-

els.

ACKNOWLEDGMENT

This research is based upon work supported by the

Social Sciences and Humanities Research Council of

Canada under Grant No. SO1936, Concordia Univer-

A Long-range Vision System for Projection Mapping of Stereoscopic Content in Outdoor Areas

295

(a)

(b)

(c)

(d)

Figure 6: (a) Three synthetic test cases: perfectly specu-

lar, perfectly diffuse, and diffuse/specular. (b) Energy mini-

mization for shininess. (c) Energy minimization for diffuse

coefﬁcients. (d) Energy minimization for specular coefﬁ-

cients. (e) Total energy minimization.

sity’s Faculty of Engineering and Computer Science

under Grant No. VH0003, and the Concordia Univer-

sity CASA Research Grant No. CS1136.

REFERENCES

Anaglyph image. https://goo.gl/9vSgkP. Accessed: 2016-

11-25.

Bell, T., Xu, J., and Zhang, S. (2016). Method for out-of-

focus camera calibration. Applied optics, 55(9):2346–

2352.

Debevec, P., Tchou, C., Gardner, A., Hawkins, T., Poullis,

C., Stumpfel, J., Jones, A., Yun, N., Einarsson, P.,

Lundgren, T., et al. (2004). Estimating surface re-

ﬂectance properties of a complex scene under cap-

tured natural illumination. Conditionally Accepted to

ACM Transactions on Graphics, 19.

Feris, R., Raskar, R., Tan, K.-H., and Turk, M. (2004).

Specular reﬂection reduction with multi-ﬂash imag-

ing. In Computer Graphics and Image Process-

ing, 2004. Proceedings. 17th Brazilian Symposium on,

pages 316–321. IEEE.

Furukawa, Y. and Ponce, J. (2010). Accurate, dense, and ro-

bust multiview stereopsis. IEEE transactions on pat-

tern analysis and machine intelligence, 32(8):1362–

1376.

Garcia, R. R. and Zakhor, A. (2013). Geometric calibra-

tion for a multi-camera-projector system. In Applica-

tions of Computer Vision (WACV), 2013 IEEE Work-

shop on, pages 467–474. IEEE.

Grossberg, M. D., Peri, H., Nayar, S. K., and Belhumeur,

P. N. (2004). Making one object look like another:

Controlling appearance using a projector-camera sys-

tem. In Computer Vision and Pattern Recognition,

2004. CVPR 2004. Proceedings of the 2004 IEEE

Computer Society Conference on, volume 1, pages I–

452. IEEE.

Herakleous, K. and Poullis, C. (2014). 3dunderworld-

sls: An open-source structured-light scanning sys-

tem for rapid geometry acquisition. arXiv preprint

arXiv:1406.6595.

Lafortune, E. P. F., Foo, S.-C., Torrance, K. E., and Green-

berg, D. P. (1997). Non-linear approximation of re-

ﬂectance functions. In Proceedings of the 24th An-

nual Conference on Computer Graphics and Inter-

active Techniques, SIGGRAPH ’97, pages 117–126,

New York, NY, USA. ACM Press/Addison-Wesley

Publishing Co.

Levenberg, K. (1944). A method for the solution of certain

non–linear problems in least squares.

Lin, S., Li, Y., Kang, S. B., Tong, X., and Shum, H.-Y.

(2002). Diffuse-specular separation and depth recov-

ery from image sequences. In European conference

on computer vision, pages 210–224. Springer.

Oren, M. and Nayar, S. K. (1995). Generalization of

the lambertian model and implications for machine

vision. International Journal of Computer Vision,

14(3):227–251.

GRAPP 2017 - International Conference on Computer Graphics Theory and Applications

296

(a) (b)

Figure 7: (a) A stereoscopic projection without compensation. Color distortions due to the reﬂectance properties of the

projection surface negatively affect the depth perception of the viewer. (b) The stereoscopic projection after compensation

using the proposed technique. Color distortions are minimized by taking into account the effect of the reﬂectance properties

of the projection surface. (c) The expected projection. This cannot always be achieved due to the limitations of additive color

mixing and the hardware.(d) The original stereoscopic image taken from (ste, ).

Phong, B. T. (1975). Illumination for computer generated

pictures. Communications of the ACM, 18(6):311–

317.

Raskar, R., Welch, G., Cutts, M., Lake, A., Stesin, L., and

Fuchs, H. (1998a). The ofﬁce of the future: A uni-

ﬁed approach to image-based modeling and spatially

immersive displays. In Proceedings of the 25th an-

nual conference on Computer graphics and interac-

tive techniques, pages 179–188. ACM.

Raskar, R., Welch, G., and Fuchs, H. (1998b). Spatially

augmented reality. In First IEEE Workshop on Aug-

mented Reality (IWAR98), pages 11–20.

Ridel, B., Reuter, P., Laviole, J., Mellado, N., Couture,

N., and Granier, X. (2014). The revealing ﬂashlight:

Interactive spatial augmented reality for detail explo-

ration of cultural heritage artifacts. Journal on Com-

puting and Cultural Heritage (JOCCH), 7(2):6.

Seitz, S. M., Matsushita, Y., and Kutulakos, K. N. (2005). A

theory of inverse light transport. In Tenth IEEE Inter-

national Conference on Computer Vision (ICCV’05)

Volume 1, volume 2, pages 1440–1447. IEEE.

Svoboda, T., Martinec, D., and Pajdla, T. (2005). A conve-

nient multicamera self-calibration for virtual environ-

ments. Presence, 14(4):407–422.

Torrance, K. E. and Sparrow, E. M. (1967). Theory for off-

specular reﬂection from roughened surfaces. JOSA,

57(9):1105–1112.

Tsai, R. (1987). A versatile camera calibration technique

for high-accuracy 3d machine vision metrology using

off-the-shelf tv cameras and lenses. IEEE Journal on

Robotics and Automation, 3(4):323–344.

Zhang, Z. (2000). A ﬂexible new technique for camera cal-

ibration. IEEE Transactions on pattern analysis and

machine intelligence, 22(11):1330–1334.

A Long-range Vision System for Projection Mapping of Stereoscopic Content in Outdoor Areas

297