Existing multimodal registration algorithms depend on optimal transformation estimation based on feature point matching or deformation prediction based on depth learning. However, the former fails to ...