Enhanced change detection using nonlinear feature extraction

This paper presents an application of the kernel principal component analysis aiming at aligning optical images before the application of change detection techniques. The approach relies on the extraction of nonlinear features from a selected subset of pixels representing unchanged areas in the images. Both images are then projected into the aligned space defined by the eigenvectors associated to largest variance (eigenvalues). In the transformed space, unchanged pixels of both datasets are mapped next to each other, thus reducing within-class variance. The difference image that results from differencing the (kernel) principal components is likely to provide a more suitable representation for the detection of changes. A bi-temporal subset of Landsat TM images validates the proposed approach, which is used to provide a suitable representation before applying the change vector analysis and the support vector domain description.