Parameter setup

Parameters to setup for the cross-correlation algorithm are in all cases:

• The size of the search area in x and y. Given a starting point, how far will be searched in pixels around the point to find for a matching point? Setting this value depends on the expected level of accuracy for the provided positions.
• The size of the measurement area in x and y. Given a candidate position for a match, how many pixels should be used to compute the cross-correlation criterion? Figure 5 shows the importance of choosing a right measurement area size.

  

Figure 5: Importance of a right definition for the measurement area. The measurement areas are shown as overlapping dark boxes over the image. Above: it is too small, as it would induce a periodical pattern effect (any other star is a good candidate). Below: it looks large enough to prevent false detections.

As specified, parameter settings depend on the quality of the input estimates for the offsets. The implementation gives the following possibilities for input estimates: header offsets, file offsets, or no offsets (automatic search). Each case is described hereafter.

One word about edge effects: computing a cross-correlation criterion for a point lying on an image edge would mean a loss in precision and probably artefacts. For this reason, it has been decided not to compute any cross-correlation if the involved point lies too close to an edge in one of the two input images. This sets a constraint on the choice of correlating point only.