ON THE GEOMETRICAL QUALITY OF PANORAMIC IMAGES 
Petteri Pöntinen 
Institute of Photogrammetry and Remote Sensing 
Helsinki University of Technology 
P.O.Box 1200, FIN-02015 HUT, Finland 
petteri.pontinen@hut.fi 
Commission V 
KEY WORDS: Geometric, Quality, Measurement, Mosaic, Sequences, Distortion, Calibration 
ABSTRACT: 
Creating of panoramic images is a common technique nowadays. The needed components are a sequence of overlapping digital 
images and a suitable stitching software. There are several software available for combining single images together just by few 
mouse clicks. They can also take care of the image distortions and adjust the radiometric differences so that the final image looks 
very consistent. In addition to the commercial software there are also frec ones, which can be downloaded from Internet. Panoramic 
images are used in many different contexts. Real estate agents have on their web pages 360 degrees panoramic views of interiors of 
the houses for sale. Panoramic views can be found in many virtual reality and multimedia presentations, advertisements, work of arts, 
etc. Depending on the usage there are different requirements for the panoramic images. For a real estate agent it might be enough that 
the panorama is seamless and gives impressive presentation of the interior of the house. The most important thing is that the image 
looks impressive. But this is not always the case. If the images are used for measurement purposes the geometrical quality of the 
panoramic images is important. Even though panoramic images are not very often used in measurement tasks there are some cases 
where they might be advantageous. Sometimes, for example, the image must be taken quite far away from the object in order to see 
enough control points. This might reduce visible details. But using panoramic images taken closer to the object might give both 
enough control points and visible details. This paper concentrates on the geometrical quality of the panoramic images created from 
concentric image sequences. The main goal is to understand the suitability of panoramic images for photogrammetric measurements. 
1. INTRODUCTION 
Developments in imaging technology have increased the 
production and use of panoramic imagery. Panoramic images 
can be found in advertisements, artworks, virtual reality and 
multimedia presentations, etc. In the Internet numerous 
impressive panoramic images can be found. One interesting way 
to utilize panoramic images is to combine them with laser 
scanner data for modelling and visualization purposes (Haala et 
al., 2004; Reulke et al., 2003; Rónnholm et al., 2003; Scheibe et 
al., 2004) 
There are two main streams in digital panoramic image 
capturing (Luhmann et al, 2003). First, the panoramic image 
can be constructed from a concentric image sequence or second, 
captured using rotating line scanner. In the first case there is no 
need for a special camera. Instead, an image sequence taken 
with a digital frame camera or scanned from analog photographs 
can be stitched to a panoramic image with proper software. 
There are more than 30 commercial stitching software on the 
market (Remondino et al., 2004) and some software can be 
downloaded for free from Internet. Stitching can be based on 
corresponding points of adjacent images (Luhmann et al., 2004) 
or on the whole overlapping area (Szeliski, 1996; Póntinen, 
1999). The projection centre of the camera should be stable 
during the camera rotation, but if the object is far away from the 
camera or approximately planar, small deviation from the 
concentricity does not prevent the stitching. To make the image 
sequence better concentric, a special camera adapter can be 
used (see Figure 1). This concept is widely in use because of its 
low costs. 
The other concept, rotating line scanner, requires more 
investments but produces directly panoramic output with very 
high resolution. There are some commercial devices available, 
like EyeScan from KST Dresden GmbH and German Aerospace 
Centre (DLR), or SpheroCam from Spheron VR AG. 
Panoramic images can be used also for photogrammetric 
measurements (Antipov et al., 1984; Hartley, 1993; Luhmann et 
al., 2004). Some possible applications concerning construction 
machines and building sites are listed in (Hoske et al., 2004). In 
the case of exact measurements the geometrical consistency of 
the panoramic image becomes important. In order to maximize 
the consistency the used instrument, either a rotating line 
scanner or a frame camera with panorama adapter, must be 
calibrated. The calibration of the instrument includes the 
determination of the. camera parameters (camera constant and 
principal point), lens distortions and the eccentricity of the 
projection centre from the rotation centre. In the case of a 
rotating line scanner also the tilt and inclination of the imaging 
sensors with respect to the rotation axis, resolution of rotation 
and so called tumbling must be solved (Amiri Parian et al., 
2003; Amiri Parian et al., 2004). The geometrical modelling and 
calibration of an EYESCAN M3 panoramic camera can be 
found in (Schneider et al., 2003). 
This paper concentrates on the geometric consistency of 
stitched image sequences. The more time, money and effort is 
spend for creating the panoramic image mosaic, the better is the 
result. But if someone buys a digital camera, a panoramic 
  
  
  
   
  
  
    
  
  
    
   
  
  
    
  
  
  
  
  
  
  
  
  
  
  
  
   
    
  
  
  
   
   
   
   
  
   
   
  
   
   
  
  
  
   
     
   
   
  
  
   
  
  
   
  
  
    
Internati 
adapter 
sequenc 
close ra 
or she ¢ 
this que: 
In char 
discusse 
concent! 
stitching 
measure 
  
Figure 
Depend; 
distortio 
hundred 
cannot 
distortio 
Calibrat 
field m 
capturec 
station € 
station | 
1996). 
camera . 
the seq 
(Hartley 
projectiy 
taken fr 
formulat 
(Wester- 
neglecte 
generall: 
remove | 
While 
correlati 
be taken 
there is 
correctic 
listed t 
paramete