This paper presents a satellite image co-registration procedure aiming at simultaneously estimating multiple affine transformations between a set of multi-temporal or multi-source satellite images, reducing error accumulation and improving metric precision. The approach is based on synchronization, a method that seeks to infer the unknown states of a network of nodes, where only the ratio (or difference) between node pairs can be measured. In our case states represent affine transformations. The proposed method globally combines via synchronization pairwise transformations computed for all the image combinations of the multi-temporal sequence, beyond the traditional image-to-base approach available in remote sensing and GIS packages. Results obtained with Landsat and Sentinel-2 images reveal that the algorithm can be used not only to perform the actual co-registration, but also as a diagnostic tool to evaluate the quality of transformation parameters through a comparison with basic co-registration methods, as well as with global least squares adjustment.
