Content uploaded by Fabio Remondino
Author content
All content in this area was uploaded by Fabio Remondino on Oct 12, 2014
Content may be subject to copyright.
SURFACE RECONSTRUCTION ALGORITHMS FOR
DETAILED CLOSE-RANGE OBJECT MODELING
Fabio Remondino, Li Zhang
Institute of Geodesy and Photogrammetry, ETH Zurich, Switzerland
E-mail: fabio@geod.baug.ethz.ch
Web: http://www.photogrammetry.ethz.ch
KEY WORDS: Surface Reconstruction, Modeling, Precision, Matching
ABSTRACT
Nowadays 3D modeling is generally performed using image or range data. Range sensors are getting a quite common source of data
for modeling purposes due to their speed and ability to capture millions of points. In this paper we report about two surface
measurement algorithms for precise and detailed object reconstruction from terrestrial images. Photogrammetry has all the
potentialities to retrieve the same details of an object that range sensors can achieve. Using advanced measurement techniques,
which combine area-based and feature-based matching algorithms we are able to generate dense point clouds of complex and free-
form objects, imaged in closely or widely separated images. Different examples are reported to show the potentiality of the methods
and their applicability to different close-range data sets.
1. INTRODUCTION
Three-dimensional modeling from images is a great topic of
investigation in the research community, even if range sensors
are becoming more and more a common source and a good
alternative for generating 3D information quickly and precisely.
3D modeling of a scene should be meant as the complete
process that starts with the data acquisition and ends with a
virtual model in three dimensions visible interactively on a
computer. The interest in 3D modeling is motivated by a wide
spectrum of applications, such as animation, navigation of
autonomous vehicles, object recognition, surveillance,
visualization and documentation.
In the last years different solutions for image-based 3D
modeling have been developed. Most of the current reliable and
precise approaches are based on semi-automated procedures,
therefore the introduction of automated algorithms is a key goal
in the photogrammetric and vision communities. 3D modeling
methods can be classified according to the level of automation
or the required input data while their strength is reflected by the
variety of scene that can be processed and the level of detail
that can be reconstructed.
The common fully automated ‘shape from video’ framework
[e.g. Fitzibbon & Zisserman, 1998; Nister, 2001; Pollefeys et
al., 2004] requires good features in the images, very short
baseline and large overlap between consecutive frames,
requirements which are not always satisfied in practical
situations, due to occlusions, illumination changes and lack of
texture. So far, automated surface reconstruction methods, even
if able to recover complete 3D geometry of an object, reported
errors between 3% and 5% [Pollefeys et al., 1999], limiting
their use for applications requiring only nice-looking 3D
models. Furthermore, post-processing operations are generally
required, which means that user interaction is still needed.
Indeed the most impressive results are achieved with interactive
methods and taking advantage of the environment constraints,
in particular for architectural objects. For different applications,
such as cultural heritage documentation, semi-automated
methods are still preferred as smoothed results, missing details
or lack of accuracy are not accepted.
In this article we report about two surface matching algorithms
developed for the precise and detailed measurement and 3D
modeling of complex and free-form terrestrial objects, like pots,
reliefs, statues, façades, etc. Commercial photogrammetric
stations generally fail with tilted close-range images, therefore
the topic still need some developments. We will concentrate
only on the measurement of the object surface, assuming the
calibration and orientation of the images already performed. As
the network configurations that allow full and precise camera
calibration are usually very different from those used for scene
reconstruction, we generally first calibrate the camera using the
most appropriate set of images and afterwards recover the
orientation parameters of the scene’s images using the
calibration results. The orientation is generally performed by
means of a photogrammetric bundle adjustment, extracting the
required tie points with automated approaches [Remondino &
Ressl, 2006] or manual measurements.
The first surface measurement algorithm presented afterwards
matches the points in image-pairs, the second one works
simultaneously with many images. Both methods require some
seed points between the images at the beginning of the process,
to initialize it and improve the performances near surface
discontinuities. The seed points can be provided manually
(stereo or monocular measurements) or extracted automatically,
leading to a fully automated surface reconstruction method. The
number of seed points depends on the set of images, their
disparity and texture content. Starting from these seeds points, a
dense and robust set of correspondences covering the area of
interest is generated.
Our research aims to combine area-based and feature-based
matching techniques to recover complete and detailed 3D
surfaces. The methods can cope with depth discontinuity, wide
baselines, repeated pattern, occlusions and illumination
changes.
In the next sections, after an overview of image-based modeling
works and matching strategies, the two matching strategies are
described in details. Then some examples demonstrating the
potentialities of the algorithms and their applicability to
different close-range data sets are reported and discussed.
Results in form of 3D point clouds, shaded and textured 3D
models are shown.
2. 3D MODELING FROM IMAGES
Recovering a complete, detailed, accurate and realistic 3D
model from images is still a difficult task, in particular if
uncalibrated or widely separated images are used. Firstly
because the wrong camera parameters lead to inaccurate or
deformed results. Secondly because a wide baseline between
the images generally requires the user interaction in the
measurement phase.
The research activities in terrestrial image-based modeling can
be generally divided in area-based [e.g. Pollefeys et al., 2004]
and feature-based [e.g. Schmid & Zisserman, 2000] methods. A
more detailed classification of point-based methods is:
1. Approaches that try to get automatically a 3D model of the
scene from uncalibrated images (also called ‘shape from
video’ or ‘VHS to VRML’ or ‘Video-to-3D’). The fully
automated procedure widely reported in the vision
community [Fitzibbon & Zisserman, 1998; Pollefeys et al.,
1999; Nister 2001; Mayer, 2003] starts with a sequence of
images taken with an uncalibrated camera. The system then
extract interest points, sequentially match them across the
view-pairs and compute the camera parameters as well as
the 3D coordinates of the matched points using robust
techniques. This is done in a projective geometry
framework and is usually followed by a bundle adjustment.
A self-calibration, to compute the interior camera
parameters, is afterwards performed in order to obtain a
metric reconstruction, up to a scale, from the projective one.
The 3D surface model is then automatically generated by
means of dense depth maps on image pairs. See [Scharstein
& Szeliski, 2002] for a recent overview of dense stereo
correspondence algorithms. The key to the success of these
automated approaches is the very short interval between
consecutive images. Some approaches have been also
presented for the registration of widely separated views
[Pritchett & Zisserman, 1998; Matas et al., 2002; Xiao &
Shah, 2003; Lowe 2004] but their reliability and
applicability for automated image-based modeling of
complex objects is still not satisfactory, as they yield
mainly a sparse set of matched feature points. Dense
matching results under wide baseline conditions were
instead reported in [Megyesi & Chetverikov, 2004; Strecha
et al., 2004].
2. Approaches that perform an automated 3D reconstruction of
the scene from oriented images. The automated 3D
reconstruction is generally based on object constraints, like
verticality and perpendicularity [Werner & Zisserman,
2002; Van den Heuvel, 2003; Wilczkowiak et al., 2003] or
using the geometric epipolar constraint [Gruen et al., 2001].
3. Approaches that perform a semi-automated 3D
reconstruction of the scene from oriented images. The semi-
automated modeling rely on the human operator and
produced so far the most impressive results, in particular for
architectural objects [Debevec et al., 1996; El-Hakim, 2000,
2002; Gibson et al., 2002]. The interactive work consists of
the topology definition, segmentation, editing and 3D data
post-processing. The degree of automation increases when
certain assumptions about the object, such as
perpendicularity or parallel surfaces, can be introduced.
Manual measurements are also performed in some projects,
generally for complex architectural objects or in cultural
heritage documentations where highly precise and detailed
results are required [Gruen et al., 2004]. Manual measurements
are time consuming and provide for less dense 3D point clouds,
but have higher reliability compared to automated procedures.
3. MATCHING FOR SURFACE MEASUREMENTS
Image matching represents the establishment of
correspondences between primitives extracted from two or
more images. In its oldest form, image matching involved 4
transformation parameters (cross-correlation) and could already
provide for successful results [Foerstner, 1982]. Further
extensions considered a 6- and 8-parameters transformation,
leading to the well known non-linear Least Squares Matching
(LSM) estimation procedure [Gruen, 1985; Foerstner, 1986].
Gruen [1985] and Gruen & Baltsavias [1986] introduced the
Multi-Photo Geometrical Constraints into the image matching
procedure (MPGC) and integrated also the surface
reconstruction into the process. Then from image space, the
matching procedure was generalized to object space,
introducing the concept of ‘groundel’ or ‘surfel’ [Wrobel, 1987;
Helava, 1988].
Even if more than three decades have been devoted to the
image matching problem, nowadays some important limiting
factors still remain. A fully automated, precise and reliable
image matching method, adaptable to different image sets and
scene contents is not available, in particular for close-range
images. The limits stay in the insufficient understanding and
modeling of the undergoing processes (human stereo vision)
and the lack of appropriate theoretical measures for self-tuning
and quality control. The design of an image matcher should
take into account the topology of the object, the primitives used
in the process, the constraint used to restrict the search space, a
strategy to control the matching results and finally optimization
procedures to combine the image processing with the used
constraints. The correspondences between images are matched
starting from primitives (features and image intensity patterns)
and using similarity measures. Ideally we would like to find the
correspondences of every image pixel. But, in practice,
coherent collection of pixels and features are generally
matched.
A part from simple points, the extraction of feature lines (see
[Dhond & Aggarwal, 1989; Ziou & Tabbone, 1998] for a
review) is also a crucial step in the surface generation
procedure. Lines (edgel) provide more geometric information
than single points and are also useful in the surface
reconstruction (e.g. as breaklines) to avoid smoothing effects on
the object edges. Edge matching [Vosselman, 1992; Gruen &
Li, 1996; Schmid & Zisserman, 2000] establishes edge
correspondences over images acquired at different standpoints.
Similarity measures from the edges attributes (like length,
orientation and absolute gradient magnitude) are a key point for
the matching procedure. Unfortunately in close-range
photogrammetry, the viewpoints might change consistently;
therefore similarity measures are not always useful for edge
matching.
4. STEREO-PAIR SURFACE MEASUREMENT
The first developed algorithm is a stereo matcher with the
additional epipolar geometric constraint. The method was
firstly developed for the measurement of human body parts
[D’Apuzzo, 2003] and afterwards also applied to full human
body reconstruction [Remondino, 2004] and rock slopes
retrieval [Roncella et al., 2005]. It has been now extended to
include also edge matching. The main steps of the process are:
1. Image pre-processing: the images are processed with the
Wallis filter [Wallis, 1976] for radiometric equalization and
especially contrast enhancement. The filter enables a strong
enhancement of the local contrast by retaining edge details
and removing low-frequency information in the image. The
filter parameters are automatically selected analyzing the
image histogram.
2. Point matching: the goal is to produce a dense and robust set
of corresponding points between image-pairs. Starting from
few seed points well distributed in the images, the
automated process establishes correspondences by means of
LSM. The images are divided in polygonal regions
according to which of the seed point is closest. Starting
from the seed points, the automated process produce a
dense set of image correspondences in each polygonal
region by sequential horizontal and vertical shifts. One
image is used as template and the other as search image.
The algorithm matches correspondences in the
neighborhood of a seed point in the search image
(approximation point) by minimizing the sum of the squares
differences of the gray value between the two image
patches. If the orientation parameters of the cameras are
available, the epipolar geometric constraints between the
images can also be used in the matching process. Generally
two stereo-pairs (i.e. a triplet) are used: the matcher
searches the corresponding points in the two search images
independently and at the end of the process, the data sets
are merged to become triplets of matched 2D points.
3. Edge matching: the approach extracts line features based on
the edge detection and linking proposed in [Canny, 1986]
and [Henricsson & Heitger, 1994]. For each image, only the
edges longer than a certain threshold are kept. Afterwards
an edge matching is performed for each image pair of the
set. Firstly the middle points of the edges are matched,
providing a preliminary list of edge correspondences.
Afterwards, starting from the matched middle point, the
other points lying on the edge are matched in a propagative
way.
4. 3D Point cloud generation: the 2D matched points and edges
are transformed in 3D data by forward ray intersection,
using the camera exterior orientation parameters.
The developed matching process works on image pairs and
integrates the epipolar constraint in the least squares estimation,
limiting the patch in the search image to move along the
epipolar line. To evaluate the quality of the matching results,
different indicators are used: a posteriori standard deviation of
the least squares adjustment, standard deviation of the shift in x
and y directions and displacement from the start position in x
and y direction. Thresholds for these values are defined
manually for different cases, according to the level of texture in
image and to the type of template. The definition of the seed
points is generally crucial, in particular if discontinuities are
present on the surface. The matcher, working only with stereo-
pairs, is less robust than a multi-image strategy which takes into
account all the available and overlapping images at the same
time, but it is still able to provide for accurate and detailed 3D
surfaces.
5. MULTI-IMAGE SURFACE MEASUREMENT
The multi-image matching approach was originally developed
for the processing of the very high-resolution TLS Linear Array
images [Gruen & Zhang, 2003] and afterwards modified to
accommodate any linear array sensor [Zhang & Gruen, 2004;
Zhang, 2005]. Now it has been extended to process other image
data such as the traditional aerial photos or close-range images.
The multi-image approach uses a coarse-to-fine hierarchical
solution with an effective combination of several image
matching algorithms and automatic quality control. Starting
from the known calibration and orientation parameters, the
approach (Figure 1) essentially performs three mutually
connected steps:
1. Image pre-processing: the set of available images is proc-
essed combining an adaptive smoothing filter and the Wal-
lis filter [Wallis, 1976], in order to reduce the effects of the
radiometric problems such as strong bright and dark regions
and optimizes the images for subsequent feature extraction
and image matching. Furthermore image pyramids are gen-
erated.
2. Multiple Primitive Multi-Image (MPM) matching: this part is
the core of the all strategy for accurate and robust surface
reconstruction, utilizing a coarse-to-fine hierarchical match-
ing strategy. Starting from the low-density features in the
lowest resolution level of the image pyramid, the MPM
matching is performed with the aid of multiple images (two
or more), incorporating multiple matching primitives (fea-
ture points, grid points and edges) and integrating local and
global image information. The MPM approach consists of 3
integrated subsystems (Figure 1): the feature point extrac-
tion and matching, the edge extraction and matching (based
on edge geometric and photometric attributes) and the re-
laxation based relational matching procedure. Within the
pyramid levels, the matching is performed with an exten-
sion of the standard cross-correlation technique (Geometri-
cally Constrained Cross-Correlation -GC3-). The MPM
matching part exploits the concept of multi-image matching
guided from object space and allows reconstruction of 3D
objects by matching all available images simultaneously,
without having to match all individual stereo-pairs and
merge the results. Moreover, at each pyramid level, a TIN
is reconstructed from the matched features using the con-
strained Delauney triangulation method. The TIN is used in
the subsequent pyramid level for derivation of approxima-
tions and adaptive computation of some matching parame-
ters.
3. Refined matching: a modified Multi-Photo Geometrically
Constrained Matching (MPGC) and the Least Squares B-
Spline Snakes (LSB-Snakes) methods are used to achieve
potentially sub-pixel accuracy matches and identify some
inaccurate and possibly false matches. This is applied only
at the original image resolution level. The surface derived
from the previous MPM step provides well enough ap-
proximations for the two matching methods and increases
the convergence rate.
The main characteristics of the multi-image-based matching
procedure are:
• Truly multiple image matching: the approach does not aim at
pure image-to-image matching but it directly seeks for im-
age-to-object correspondences. A point is matched simulta-
neously in all the images where it is visible and exploiting
the collinearity constraint, the 3D coordinates are directly
computed, together with their accuracy values.
• Matching with multiple primitives: the method is a robust
hybrid image matching algorithms which takes advantage
of both area-based matching and feature-based matching
techniques and uses both local and global image
information. In particular, it combines an edge matching
method with a point matching method through a probability
relaxation based relational matching process.
• Self-tuning matching parameters: they are automatically de-
termined by analyzing the results of the higher-level image
pyramid matching and using them at the current pyramid
level. These parameters include the size of the correlation
window, the search distance and the threshold values. The
adaptive determination of the matching parameters results
in higher success rate and less mismatches.
• High matching redundancy: exploiting the multi-image con-
cept, highly redundant matching results are obtained. The
high redundancy also allows automatic blunder detection.
Mismatches can be detected and deleted through the analy-
sis and consistency checking within a small neighbourhood.
More details of the matching approach are reported in Zhang
[2005].
Figure 1: Workflow of the automated DSM generation approach. The
approach consists of 3 mutually connected components: the image pre-
processing, the multiple primitive multi-image (MPM) matching and
the refined matching procedure.
6. EXPERIMENTS
We have performed many tests on different close-range data
sets with the two surface reconstruction approaches. So far the
results are checked just visually, as no reference is available. In
the future an accuracy test should be performed. In the next
sections we report results from widely separated images,
untextured surfaces and detailed heritage objects. More
examples are reported in our homepage.
Test1. Three images of the main door of the S. Marco church in
Venice (Italy) are used. The image size is 2560x1920 pixels.
The triplet is acquired under a wide baseline (base-to-distance
ratio ~ 1:1.4) and very fine details are present on the object.
Both methods could correctly retrieve the surface details, as
shown in Figure 3. The stereo-pair strategy matched
approximately 590 000 points between the two pairs while the
multi-image matching recovered ca 700 000 points.
Figure 2: The three images of the church’s façade acquired under a large
baseline.
Test2. A very small pot (ca 3x4 cm) is modeled with the two
presented matching strategies. Six images, with a size of 1856 x
1392 pixels are used. The detailed results are shown in Figure 4
as textured and shaded surface model.
Test 3. The data set consists of 6 images of a Maya relief in
Edzna, Mexico. The object is approximately 4 meters long and
2 meters high. The images have different light conditions and
scales. Due to the frontal acquisition, the upper horizontal part
of the relief is not visible in the images, leading to some gaps in
the matching point results and some stretching effects in the
meshed model. Both methods could reconstruct all the details
of the heritage. The stereo-pairs approach (performed on 4
pairs) generated ca 860 000 points and 7 900 edges while with
the multi-image strategy a cloud of 1 940 000 points and 23 000
edges was produced. The results are shown in Figure 5.
Figure 3: Results of the stereo-pair matching method (above): recovered
3D point cloud, displayed with pixel intensity values and a particular of
the generated shaded model. Views of the shaded and textured surface
generated with the multi-image method (below).
Figure 4: Three (out of 6) image of the small pot (above). The surface
reconstructed with the stereo-pair matching and with the multi-photo
approach.
7. CONCLUSIONS AND OUTLOOK
We have presented two matching strategies for the precise
surface measurement and 3D reconstruction of complex and
detailed terrestrial objects. The stereo-pair approach constraints
the search of correspondences along the epipolar line while the
3D coordinates of points and matched edges are computed in a
second phase, using rejection criteria for the forward ray
intersection. The multi-image approach is more reliable and
precise but requires very accurate image orientation parameters
to exploit the collinearity constraint within the least squares
matching estimation. The maximum orientation errors in image
space should be less than 2-3 pixels.
The two approaches use points and edges to retrieve all the
surface details and they have both advantages and
disadvantages. They can be applied to short or wide baseline
images and can cope with scale changes, different illumination
conditions or repeated pattern. Employing the precise LSM
algorithm, they can recover sub-pixel accuracy matches. They
both need some seed points to initialize the matching procedure
and the number of seed points is strictly related to the image
texture and surface discontinuities.
The results so far achieved are promising but more tests have to
be performed as well as an accuracy assessment of the two
strategies. Photogrammetry has all the potentiality to retrieve
the same results (details) than range sensors. But to asses the
accuracy of the systems is not an easy task. Assessment on the
whole measured surface would require the two models to be in
the same reference systems or to set one model as reference and
transform the second one into the first reference system.
REFERENCES
Baltsavias, E.P., 1991: Multi-Photo geometrically constrained
matching. PhD Thesis, Institute of Geodesy and
Photogrammetry, ETH Zurich, Switzerland, 221 pages
D’Apuzzo, N., 2003: Surface Measurement and Tracking of
Human Body Parts from Multi Station Video Sequences. PhD
Thesis Nr. 15271, Institute of Geodesy and Photogrammetry,
ETH Zurich, Switzerland, 147 pages
Debevec, P., Taylor, C. and Malik, J., 1996: Modelling and
rendering architecture from photographs: a hybrid geometry
and image-based approach. ACM Proceedings of SIGGRAPH
’96, pp. 11-20
Dhond, U.R. and Aggarwal, J.K., 1989: Structure from Stereo.
IEEE Transaction on System, Man and Cybernetics, Vol. 19(6),
pp. 1489-1510
El-Hakim, S., 2000: A practical approach to creating precise
and detailed 3D models from single and multiple views.
IAPRS, 33(B5A), pp 122-129
El-Hakim, S., 2002: Semi-automated 3D reconstruction of
occluded and unmarked surfaces from widely separated views.
IAPRS, 34(5), pp. 143-148, Corfu, Greece
Fitzgibbon, A and Zisserman, A., 1998: Automatic 3D model
acquisition and generation of new images from video sequence.
Proceedings of European Signal Processing Conference, pp.
1261-1269
Foerstner, W., 1982: On the geometric precision of digital
correlation. International Archives of Photogrammetry, Vol.
24(3), pp.176-189
Foerstner, W., 1986: A feature based correspondence algorithm
for image matching. International Archives of Photogrammetry,
Vol. 26(3), Rovaniemi
Gibson, S., Cook, J. and Hubbold, R., 2002: ICARUS:
Interactive reconstruction from uncalibrated image sequence.
ACM Proceedings of SIGGRAPH’02, Sketches & Applications
Gruen, A., 1985: Adaptive least square correlation: a powerful
image matching technique. South African Journal of PRS and
Cartography, Vol. 14(3), pp. 175-187
Gruen, A. and Baltsavias, E., 1986: Adaptive least squares
correlations with geometrical constraints. Proc. of SPIE, Vol.
595, pp. 72-82
Gruen, A., and Li, H., 1996: Linear feature extraction with
LSB-Snakes from multiple images. IAPRS, Vol. 31(B3), pp.
266-272
Gruen, A., Zhang, L. and Visnovcova, J., 2001: Automatic
reconstruction and visualization of a complex Buddha Tower of
Bayon, Angkor, Cambodia. Proceedings 21.Wissenschaftlich-
Technische Jahrestagung der DGPF, pp. 289-301
Gruen, A., Zhang L., 2003. Automatic DTM Generation from
TLS data. In: Gruen/Kahman (Eds.), Optical 3-D Measurement
Techniques VI, Vol. I, ISBN: 3-906467-43-0, pp. 93-105.
Gruen, A., Remondino,F., Zhang, L., 2004: Photogrammetric
Reconstruction of the Great Buddha of Bamiyan, Afghanistan.
The Photogrammetric Record, Vol. 19(107)
Helava, U.V., 1988: Object-space least-squares correlation.
PE&RS, Vol. 54(6), pp. 711-714
Henricsson O. and Heitger F., 1994: The role of key-points in
finding contours. ECCV'94, Vol. 2, pp. 371-383
Lowe, D., 2004: Distinctive image features from scale-invariant
keypoints. International Journal of Computer Vision, Vol.
60(2), pp. 91-110
Matas, J., Chum, O., Urban, M. and Pajdla, T., 2002: Robust
wide baseline stereo from maximally stable extremal regions.
Proceedings of BMVC, pp. 384-393
Mayer, H., 2003: Robust orientation, calibration, and disparity
estimation of image triplets. 25th DAGM Pattern Recognition
Symposium (DAGM03), Number 2781, series LNCS,
Michaelis/Krell (Eds.), Magdeburg, Germany
Megyesi, Z. and Chetverikov, D., 2004: Affine propagation for
surface reconstruction in wide baseline stereo. Proc. ICPR
2004, Cambridge, UK
Nister, D., 2001: Automatic dense reconstruction from
uncalibrated video sequences. PhD Thesis, Computational
Vision and Active Perception Lab, NADA-KHT, Stockholm,
226 pages
Pollefeys, M., Koch, R. and Van Gool, L., 1999: Self-
calibration and metric reconstruction in spite of varying and
unknown internal camera parameters. IJCV, Vol. 32(1), pp. 7-
25
Pollefeys, M., Van Gool, L., Vergauwen, M., Verbiest, F.,
Cornelis, K., Tops, J. and Koch, R., 2004: Visual modeling
with a hand-held camera. IJCV, Vol. 59(3), pp. 207-232
Figure 5: Three (out of six) images of a Maya relief (above). Results of the stereo-pair matching (4 pairs) displayed as shaded model (with a close
view of the mesh) and as textured model (middle). Shaded model obtained with the multi-image matcher simultaneously run on 6 images (bottom).
Pritchett, P. and Zisserman, A. 1998: Matching and
reconstruction from widely separated views. 3D Structure
from Multiple Images of Large-Scale Environments, LNCS
1506
Remondino, F., 2004: 3D Reconstruction of Static Human
Body Shape from Image Sequence. Computer Vision and
Image Understanding, Vol. 93(1), pp. 65-85
Remondino, F. and Ressl, C., 2006: Overview and
experiences in automated markerless image orientation.
IAPRS, Comm. III Symposium, Bonn, Germany. In press.
Roncella, R., Forlani, G. and Remondino, F., 2005:
Photogrammetry for geological applications: automatic
retrieval of discontinuity in rock slopes. Videometrics VIII -
Beraldin, El-Hakim, Gruen, Walton (Eds), SPIE-IS&T
Electronic Imaging Vol. 5665, pp. 17-27
Scharstein, D. and Szeliski, R., 2002:. A taxonomy and
evaluation of dense two-frame stereo correspondence
algorithms. IJCV, 47(1/2/3): 7-42
Schmid, C. and Zisserman, A., 2000: The geometry and
matching of lines and curves over multiple views. IJCV, Vol.
40(3), pp. 199-233
Strecha, C., Tuytelaars, T. and Van Gool, L., 2003: Dense
Matching of Multiple Wide-baseline Views. IEEE
Proceedings of ICCV’03, Vol.2, pp. 1194-1201
Van den Heuvel, F., 2003: Automation in architectural
photogrammetry. PhD Thesis, Publication on Geodesy 54,
Netherlands Geodetic Commission
Vosselman, G., 1992: Relational matching. Lecture Notes in
Computer Science, No. 628, Springer Berlin, 190 pages
Wallis, R., 1976: An approach to the space variant
restoration and enhancement of images. Proc. of Symposium
on Current Mathematical Problems in Image Science, Naval
Postgraduate School, Monterey, CA
Werner, T. and Zisserman, A., 2002: New technique for
automated architectural reconstruction from photographs.
Proceedings 7th ECCV, Vol.2, pp. 541-555
Wilczkowiak, M., Trombettoni, G., Jermann, C., Sturm, P.
and Boyer, E., 2003: Scene modeling based on constraint
system decomposition techniques. IEEE Proceedings 9th
ICCV, pp. 1004-1010
Xiao, J. and Shah, M. 2003: Two-frame wide baseline
matching. IEEE Proceeding of 9th ICCV, Vol.1, pp. 603-610
Wrobel, B., 1987: Facet Stereo Vison (FAST Vision) – A
new approach to computer stereo vision and to digital
photogrammetry. Proc. of ISPRS Intercommission
Conference on ‘Fast Processing of Photogrammetric Data’,
Interlaken, Switzerland, pp. 231-258
Zhang, L., Gruen, A., 2004. Automatic DSM Generation
from Linear Array Imagery Data. IAPRS, Vol. 35(B3), pp.
128-133
Zhang, L., 2005: Automatic Digital Surface Model (DSM)
generation from linear array images. PhD Thesis Nr. 16078,
Institute of Geodesy and Photogrammetry, ETH Zurich,
Switzerland, 199 pages
Ziou, D. and Tabbone, S., 1998: Edge Detection Techniques
- An Overview. Journal of Pattern Recognition and Image
Analysis. Vol. 8, pp. 537-559
