AN IMPROVED ALGORITHM FOR NOAA-AVHRR IMAGE REFERENCING

The algorithm for NOAA-AVHRR image referencing previously presented in this journal by Ho and Asem is discussed and improved. Applying their procedures, difficulties were found in obtaining proper inputs (such as exact time difference (dt) between nodal time and time of the first scan line) and hence calculating adjusted values for satellite altitude and inclination angle. The improved algorithm is based on the same equations but uses nominal inputs where possible (satellite altitude, inclination) and adjusts the others (dt and ascending nodal longitude) using one ground control point (GCP) in a simulation process. The algorithm was developed and evaluated to geocorrect NOAA-images (size 512 by 512) on a Personal Computer AT. To make it faster, we present a combination of the orbital model with the advantages of the GCP-method by an automatic determination of randomly distributed reference points (REF) and bilinear regression analysis. A set of six representative images was geocorrected and mean errors were calculated using independent GCPS. The results showed a pixel displacement of 0.3 to 1.5 in line and 0-5-2.3 in row respectively. One picture of an almost cloud-free scene is illustrated.