Multi-Scale Change Detection Based on Fractal Analysis

The aim of this work is to propose the base-line of a new approach for change detection, based on local fractal dimension analysis. The box counting algorithm has been used for generating local fractal dimension maps of the images registered at different date to be compared. From these maps the percentage of the spectral and texturale changes between the images analyzed has been obtained. From results obtained is has been proved that this methodology is capable to detect changes in different scales depending the windows sizes using to determine the fractal maps. That implies that it is posible to define the scale of the change that will be detected.