into polar components (magnitude and direction) and residuals were calculated again. ABM and STDEV of the polar residuals quantify the obtained accuracy. This approach is analogous to the procedure described in Section 4.3.1. In this case the extent dependency of the mean error is realized by the size of the correlation window. To obtain accuracy for 'annual creep rates", the magnitude has to be divided through the number of years between the measurements.
Panadda Suebpimpawong
43.3. Automatic post processing of vector fields Inevitably there are some large errors in most of the creep vector fields obtained. They are mainly single mismatches and larger areas with faulty correlations due to of structure, shadow major changes effects or homogeneity of surface pattern. The different superficial extension of these two kinds of errors makes it difficult to eliminate them fully automatically without a considerable loss of data. We obtain- ed the best results with a relatively simple filter procedure. In the regu lar raster structure of the vector field we calculated STDEV of the angle for the 3 x 3 surroundings of each single vecto. This calculation is difficult due to the discontinuous function of angle units. STDEV be- tween neighbouring vectors of 360°, 1" and 2" cannot be calculated in the common way. We therefore calculated the mean vector vm of the 3 x 3 surrounding first