A computational approach to fractal analysis of a cityscape's skyline

This study proposes a semi-automated approach for cityscape analysis which is based on calculating the fractal dimension of a cityscape's skyline. A software tool was developed which consists of an intensity-based skyline extraction module combined with a box-counting approach for calculation of the fractal dimension. Obstacles such as power-lines, vertical poles or cranes which interrupt the skyline can automatically be excluded from the analysis. The paper describes the methods involved and presents three pilot experiments using the new approach which indicate that: (1) If trees intersect the skyline they typically increase its fractal dimension; (2) Different types of cities can be distinguished by their characteristic skylines; (3) The process to determine the best fit skyline in an image can require user intervention. It can be semi-automated by using the local minima of the skyline's fractal dimension which is interpreted as a function of the image's intensity cut-off values.