The processing times of the algorithms are discussed in section 5.4.
The iterative algorithms SJ87 and SS90 are not suitable for working with many
trajectories because of the high computational cost that becomes prohibitive
very soon as
grows. These algorithms cannot efficiently constrain
the correspondences, which leads to combinatorial explosion.
(The unconstrained motion correspondence is an NP-complete problem.)
A critical value of
is around 40, when SJ87 becomes
practically impossible to systematically test.
As pointed out in [9], this
greedy exchange algorithm may sometimes not converge at all,
going into an infinite loop -- if the iterations are not forced to stop by
limiting their number, as we actually do.
Apparently, this occurs more frequently as the point density increases.
In the original paper [11],
no efficient stopping rule was formulated for such cases.