A proposedmulti-scale approachwith automatic scale selection for image change detection

In this paper a system for multi-scale change detection with automatic scale selection is proposed. The generation of the multi-scale data set is based on fractal net evolution approach. The set of scales used are selected optimally from the scale domain to ensure good enough representation of the scale domain. The change detection is performed on each scale image independently then the information extracted from all the scales is fused to construct the final change map. A scale-driven method is proposed to fuse the multi-scale information. The method detects for each object the preferred scale to model it and hence can get the change information for this object to form the final change map. The proposed system is compared against different multi-scale systems and proved to perform better.