A novel automatic and robust remote sensing image registration method is proposed.At first

the feature points are identified as local extrema both in 2D-space and scale space of the image

and the local invariant feature descriptors are extracted in the neighborhoods of the feature points

named as scale invariant feature transform（SIFT）.Then an initial matching set is obtained by matching the SIFT features based on the distance gather.At last

the matching point set is divided into inner and outer point using robust random sample consensus（RANSAC）algorithm

and the image transform model is accurately estimated by the inner sub-set.The repeatability and matchability of SIFT features are tested by the simulated experiment.The validity and the robustness is validated by the experiment with satellite image of the proposed method.