Simultaneous least-squares registration of satellite time series

This paper presents a new technique able to register satellite time series made up of several images within a global Least Squares adjustment. The method improves traditional ‘master-to-slave’ registration approaches and extends the pairwise registration concept towards block adjustment. Its main advantage concerns the combined analysis of multiple images, avoiding standard matching between one ‘master’ and several ‘slave’ images. This results in a better control of errors and a more robust and reliable Least Squares solution with full error propagation. Application to Landsat TM images is reported to show the performance of the simultaneous Least Squares registration.