In order to develop an operational method for the U.S. Navy/NOAA Joint Ice Center to extract ice velocity vectors from sequential advanced very high resolution radiometer (AVHRR) imagery, we have combined the maximum cross correlation (MCC) method with a spatial filtering technique on the image inferred ice motion vectors. We compute the cross correlations between images directly from the image brightness values rather than computing FFTs. The direct method allows greater flexibility in computational parameter settings and allows one to compute motion vectors near coastlines where irregular windows are required. By using a combination of statistical and spatial filters we can then retrieve coherent ice motion vectors in the presence of cloud contaminated imagery. A series of six satellite images of the Fram Strait region, from April 1986, was used to compute sea ice motion from pairs of sequential images. The resulting ice motion vectors were taken as a representation of the surface flow field derived objectively from the satellite imagery. Resulting vector motion fields were found to match well with manually tracked vectors for the same images, thus verifying the validity of the objective MCC method of computing ice motion. These techniques were applied to both the visible and infrared AVHRR channels and to images with different spatial resolutions yielding an overall bias accuracy of about 0.5 cm/s and standard deviations of about 0.9 cm/s. The MCC ice motion results were also compared with wind-driven numerical model simulations of the region. Marked differences between the MCC image-derived velocities and those from the numerical model were thought to be primarily due to a stronger ocean current than was present in the model ¿1991 American Geophysical Union