When cinder and glance both are using rbd backend then there doesn't exist any optimization when we upload a volume to image service.
Similar optimization exists when we are creating a bootable volume from image where we use a COW clone across the glance and cinder ceph pools.