Quality-aware deformable models for change detection

In this paper we present a novel approach to image-based change detection by extending the model of deformable contour models (snakes) to function in a differential mode. We make use of prior information for an object (an older record of the shape of the object) and accompanying accuracy estimates. Our objective is to update this preexisting geospatial information (shape and corresponding accuracy). We introduce a new framework, comprising change detection and versioning, to differentiate change detection from the identification of numerous statistically identical versions of an object. Experimental results are presented to demonstrate the performance of our approach.