Adaptive technique for change detection in VHR multispectral images robust to registration noise

This paper presents an automatic context-sensitive technique robust to registration noise (RN) for change detection (CD) in multitemporal very high geometrical resolution (VHR) remote-sensing images. Exploiting the properties of RN in VHR images, the proposed technique analyzes the distribution of the spectral change vectors (SCVs) computed according to the change vector analysis in an adaptively quantized polar domain. The method studies the behavior of SCVs falling into each quantization cell at different resolution levels (scales) in the polar domain to automatically reduce the effects of RN on the final change detection map. This information is jointly exploited with the spatial-context information contained in the neighborhood of each pixel for generating the final CD map. The spatial-context information is modeled through the definition of adaptive regions homogeneous both in spatial and temporal domain (parcels). Experimental results obtained on real VHR remote-sensing multitemporal images confirm the effectiveness of the proposed approach.