An automatic approach of image registration is developed to register the multi

t

emporal images of different sensors. A point in the coastline derived from SeaWiFS and AVHRR images is taken as a ground control point （GCP）

and the correlation

r

elaxation （CR） technique is employed to search the counterpart point of GCP. The correction of GCP values is guaranteed by a decision rule

which will improve the accuracy of registration. The relationship between the n th order bivariate transforms and the accuracy of registration is studied. The averages of RMS show that the larger number of GCP will lead to the more accurate image registration

while the high order bivariate transforms can not reduce that RMS of two images. The large number of GCP obtained by this method give a chance to study new transforms for geometric rectification to improve the accuracy of image registration. The approach described in the paper can be used in dynamic monitoring and integrating of Satellite data.