SAMPLING OF BIDIRECTIONAL REFLECTANCE FROM MULTIANGULAR HICH 
RESOLUTION AIRBORNE IMAGERY 
Antero Kukko 
Department of Remote Sensing and Photogrammetry 
Finnish Geodetic Institute 
PO Box 15, FIN-02431 MASALA, FINLAND 
Antero.Kukko@fgi.fi 
KEYWORDS: Acquisition; Sampling; DEM/DTM; Aerial; High resolution; Multispectral; 
ABSTRACT: 
The present investigation describes a method for sampling bidirectional reflectance information from multiangular airborne images. 
The method uses high resolution surface models in the determination of the location of the imaged point on the ground and in the 
image, and finally the orientation of the measured surface fragment. 
Two test sites were imaged with a wide range of azimuth angles at two different times. A high resolution HRSC-A stereo camera was 
used for image acquisition. Algorithms to reconstruct the image acquisition and retrieve the image samples from the HRSC-A image 
data, used with GPS and INS data and automatically derived high resolution digital terrain models, were implemented and used to 
determine the viewing and illumination geometry on the target surface. The image digital number of a sample point was recorded as 
an uncalibrated reflectance measure. A large number of directionally defined samples and a wide angular range of sample geometry 
were obtained by taking the surface orientation, extracted from the high resolution digital surface model, into account. 
The images were classified in order to aggregate the samples of a certain surface type acquired from the different images. The 
sampled reflectance data were tested and analysed by investigating the bidirectional reflectance of seven agricultural and forest 
targets. Angular uncertainty of the data, and other sources of error affecting the data quality were studied. 
Directionally defined reflectance data were acquired to assist in future the modelling and correction of bidirectional effects on 
airborne optical images used in mapping, urban modelling and for establishment of a bidirectional reflectance database. The 
multiangular image data, the developed sampling methods and the obtained bidirectional dataset proved to be feasible for 
investigations of the bidirectional effects of natural targets. Airborne imagery, including scanner and frame images, combined with 
digital surface models permits extensive investigation of the bidirectional reflectance of a wide range of natural objects and large 
habitats. 
1. INTRODUCTION 
Viewing and illumination conditions play a critical role in the 
interpretation of remote sensing and aerial images, e.g. in the 
processes of change detection, classification and image 
mosaicking. Most natural surfaces scatter incident radiation 
anisotropically (Beisl, 2001; Sandmeier et al., 1998). For such 
surfaces, sufficient description of the reflectance requires 
knowledge of the full angular distribution and reliable 
interpretation of remotely sensed data calls for knowledge 
regarding all possible surface types (Arnold et al. 2002). 
Surface anisotropy is usually described in terms of bidirectional 
reflectance distribution functions (BRDF). These functions 
formally describe the scattering anisotropy of a certain type of 
surface dependent on illumination and viewing directions (0, 
9;) and (0, p,), namely 
dL,(0;,9;.0,.9,;A) 
(0.0.0.0. A) 5 0 nr 2 
/,0.0,.0,.9,:4) eh 
where dE =LicosOdw and dw;=cos Ode; (Beisl, 2001; Jensen, 
2000; Sandmeier & Itten, 1999.) 
Bidirectional effects on an image are more prominent when a 
large field of view is used, e.g. in aerial photography (Beisl, 
2001). Bidirectional effects cause brightness variation in an 
image, depending on the sensed surface type and its condition, 
the geometric properties of sun-viewing conditions, and the 
wavelength used. The brightness variation reduces the quality of 
multi-temporal monitoring and classification if the effect is not 
corrected because the same surface may seem significantly 
different in brightness in the images acquired with different sun- 
viewing geometries. Bidirectional effects are the most severe 
when the images are taken from the direction of incident light 
and cause a ‘hot spot’ on the image (Hapke et al., 1996). 
Beisl (2001) describes methods for sampling and correcting 
bidirectional effects in imaging spectrometer data. 
Hyperspectral data were acquired using an imaging HyMap 
spectrometer at three different times and with two perpendicular 
flight directions. Sampling was carried out by aggregating the 
pixels representing similar surfaces. This approach improved 
the angular sampling, which was initially poor. 
Airborne Cloud Absorption Radiometer (CAR) has been used 
extensively in to retrieve the surface reflectance of various types 
of land cover such as forested wetland and desert (Soulen et al., 
2000), cerrado and dense forest (Tsay et al., 1998), and sea ice 
and tundra (Arnold et al., 2002). In these studies, clockwise 
circular flight tracks were used in data acquisition over 
homogenous targets. 
Pellikka (1998) uses video images for natural resource 
assessment. Correction methods were developed for 
bidirectional effects on video images acquired with a 
multiangular video system. Separate video frames were 
     
  
   
    
   
   
   
  
   
   
   
    
  
   
  
  
  
  
  
   
  
   
  
  
  
  
  
  
  
  
  
   
  
   
  
  
  
   
  
  
   
   
  
   
  
   
  
   
   
  
  
  
  
   
  
  
  
  
  
    
  
    
   
    
Inte 
cor 
nun 
dec 
me: 
fun 
illu 
als 
imé 
Th: 
fiel 
Go 
san 
pra 
of. 
nat 
ins 
fra 
are 
pix 
an 
me 
pre 
are 
in 
fl 
in 
re 
lu 
th