Automatic offsets

In this case, the algorithm should try to determine automatically the offsets between frames. It will look around the same position in all consecutive planes. The default search area is then increased to $\pm 80$ pixels, relying on the assumption that the offsets lie within 15% of the detector size. For SOFI/ISAAC, this means $\pm 75$ pixels. The measurement area is kept at $\pm 25$ pixels.

The total number of pixel operations in this case amounts to:

$(2\times80+1)^2 \times (2\times25+1)^2 = 67,420,521$

which is roughly a factor 60 in computation time compared to header offsets. It is easy to understand why it is preferrable to have reliable offsets in input (10 seconds computation time) rather than use an automatic search (10 minutes computation time). The automatic mode is also more prone to false detections, because the search area is bigger, thus the probability of mistaking positions increases.