138 
PRE-PROCESSING EFFECTS ON CLASSIFICATION OF SATELLITE IMAGERY 
Raad A. Saleh* and Eugene Derenyi 
Department of Surveying Engineering 
University of New Brunswick 
Fredericton N.B. Canada E3B 5A3 
ABSTRACT 
Quantitative analysis of satellite images require that these images be distortion free or fitted to a geographic 
reference system. Geometric correction and transformation are used for this purpose. The transformed images 
involve radiometric values derived from the original images, using a resampling technique. The accuracy of 
the resampling technique must be known to assess any changes in the original information content and the 
impact on the subsequent analysis. 
Landsat Thematic Mapper and Spot images were resampled by nearest neighbour, bilinear interpolation, and 
cubic convolution. The classification results before and after resampling, using different strategies of signature 
calculation, were evaluated with respect to each others, as well as to reference data. In general, the 
information content has been modified by resampling, yielding different classification results, these differences 
were influenced by the degree of distortion in the original images. 
* Currently at the Department of Civil and Environmental Engineering, 
University of Wisconsin-madison, Mailing address: 
Environmental Remote Sensing Center ERSC 
1225 West Dayton Street, 12th Floor 
Madison, Wisconsin 53705 
USA