Wavechange: a Procedure for Change Detection Based on Wavelet Product Spaces

Our objective was to develop a procedure for change detection which is less sensitive to problems caused by misregistration, atmospheric effects and variations in vegetation phenology. The procedure is based on redundant wavelet transforms. Wavelet correlation was explored by taking point wise products of adjacent scales in order to enhance coefficients due to changes and smooth out noises. Local maxima were detected on the product spaces at varying spatial resolutions. The detected maxima were used as seeds for a multiscale region growing algorithm. The results showed that the method is not sensitive to geometric and radiometric misregistrations because of the multiresolution approach to feature extraction.