Proceedings 18 th International Symposium CIPA 2001 
Potsdam (Germany), September 18 - 21, 2001 
FACADE RECONSTRUCTION OF DESTROYED BUILDINGS 
USING HISTORICAL PHOTOGRAPHS 
Christian Bräuer-Burchardt and Klaus Voss 
Digital Image Processing Group, Friedrich-Schiller-University 
Emst-Abbe-Platz 1-4, D-07743 Jena, Germany 
email: cbb@pandora.inf.uni-jena.de 
KEY WORDS: Monocular Reconstruction, Camera Calibration, Image Rectification 
ABSTRACT 
A methodology for the metric reconstruction of the planar surface of facades of destroyed buildings using historical single-view 
photographs has been presented in this work. The methodology is based on using a-priori object properties like linearity, parallelism, 
perpendicularity, symmetry etc. in addition to information extracted from the photographs. The methodology involves the determina 
tion of line segments and vanishing points and, if necessary, intrinsic camera calibration. The methodology may be a useful tool for 
tasks like documentation of historical architecture, surveying, and restoration. Reconstruction examples and results of facades of 
destroyed buildings are presented and error estimation results are given. 
1. INTRODUCTION 
Historical buildings often have been destroyed or damaged by natural phenomena and human activities. But in recent times people 
are interested in reconstruction of destroyed buildings in order to restore or rebuild the buildings or just for documentation. Unfortu 
nately, the reconstruction of destroyed historical buildings poses a great problem owing to the lack of architectural records in most 
cases. Historical photographs are the only means available for this task. But usually such photographs do not have metric properties. 
In the case of historical photographs stereo views are not to be expected. So one may give up to determine the metric properties of 
the reconstructed model or, and this will be the purpose of this work, find a method for metric reconstruction using only image in 
formation. 
Assuming that neither the camera information nor any reference points are available, projective 3D-2D-mapping is not reversible and 
hence metric reconstruction is theoretically impossible. However, certain image information may be extracted from a photograph by 
a human observer using common reception patterns together with usual construction rules. These construction rules, if not obviously 
violated, should complete our a-priori information about the imaged architectural object. Mainly, these are feature oriented proper 
ties like linearity, parallelism, perpendicularity, and symmetry. 
Assuming that such properties hold, the observer has an imagination of the object shape, e.g. one supposes lines to be parallel or 
perpendicular and surfaces to be planar, orthogonal, etc. This information should be included in monocular reconstruction algorithms 
and denoted by a-priori knowledge or information. 
Fig. 1: Historical photograph of the university library in Jena (Ger- Fig. 2: View of destroyed buildings in Jena 
many), destroyed during world war two 
The idea of architectural object documentation using photographs is due to Meydenbauer (1834-1921). He was the founder of the 
“Kgl. Preußische Meßbildanstalt” in Berlin (Germany) in 1885. His aim was to collect architectural records together with photo 
graphs of national and international historic monuments in order to make possible a later reconstruction of the buildings in the case 
of destruction or damage by natural phenomena or wars. In order to do this he developed a methodology which allowed the metric