CITY STRUCTURE ANALYSIS ON QUICKBIRD IMAGERY BY MULTISCALE RADON TRANSFORMATION

Abstract. The radon transformation is a proven image analysis technique in medical image processing, although there are some applications with satellite images. Among these applications maybe the most interesting one is based on the feature that this mathematical transformation is capable of detecting roads and road junctions. This option makes the transformation similar to the well-known Hough transformation, but radon has better performance parameters.
Roads are ground objects having linear element borderlines, which can be extracted from radon analysis. Since roads and streets are crucial objects in urban environment analysis, they are excellent sources to get information about the city structure.
The current paper introduces the short mathematical background behind the technique, and then a multiscale approach is described. A Quickbird image captured about Budapest was used to create a test data set with different scales, where the radon transform is applied. The different scale image layers show different details about the road structure, therefore about the city architectural structure. The result of the analysis can be presented in GIS database, thus the extracted information is able to be reused in further sophisticated applications.