Longer Registration

Experiments 25112004-1 to 25112004-3

Here I tried to see the effect of longer registration, with a greater number of knot-points towards the end. I am curious to see if the results continue to improve at a finer resolution and, if so, how long it takes.

• 25112004-1: uses ssd
• 25112004-2: uses abs_diff
• 25112004-3: uses wtd_abs_diff

All settings files are in the appendix.

Results: See /Long_Optimisation-25112004.

Conclusions: better results are obtained by running the longer (~2 hours versus about 15 minutes in most of the previous experiments) optimisation. However, results are not optimal or stunningly good. The three objective functions used give unique yet similar results. It is not obvious to the eye which objective function works best.