Abstract: In this paper,Gaussian-geometric moment(GGM) and their translation and rotation invariants have been introduced and a GGM-based image registration method has been proposed.Firstly,on the basis of GM,gauss kernel is employed to define the GGM;secondly,the GGM invariants are derived by means of the Gaussian-complex moment and an independent and complete set of GGM rotational invariants is given;finally,feature description capability of GGM by the influence of the scale parameter is fully studied.The capability of feature matching and image registration of both moments has been comprehensively tested by experiments,the results of which show that the proposed moments are much better than geometric ones in noise resistance and feature description.