Application of rotation- and translation-invariant overcomplete wavelets to the registration of remotely sensed imagery

摘要

Abstract: A wavelet-based image registration approach haspreviously been proposed by the authors. In this work,wavelet coefficient maxima obtained from an orthogonalwavelet decomposition using Daubechies filters wereutilized to register images in a multi-resolutionfashion. Tested on several remote sensing datasets,this method gave very encouraging results. Despite thelack of translation- invariance of these filters, weshowed that when using cross-correlation as a featurematching technique, features of size larger than twicethe size of the filters are correctly registered byusing the low-frequency subbands of the Daubechieswavelet decomposition. Nevertheless, high- frequencysubbands are still sensitive to translation effects. Inthis work, we are considering a rotation- andtranslation-invariant representation developed by E.Simoncelli and integrate it in our image registrationscheme. The two types of filters, Daubechies andSimoncelli filters, are then being compared from aregistration point of view, utilizing synthetic data aswell as data from the Landsat/Thematic Mapper and fromthe NOAA Advanced Very High Resolution Radiometer.!17