| Algorithm 1. ME detection post processing. |
| Params: minMicroSz: the minimum size in frames of a ME (τ/4 in our experiments) maxDist: the maximum distance between two clusters to be merged (2∙τ in our experiments) |
| 1: Find the predicted and disjunctive ME intervals: I = {(s0, e0), (s1, e1), …, (sn, en)} 2: doMerge ← True 3: while doMerge do 4: doMerge ← False 5: for i = 1 to len(I) do 6: m1 <- (ei-1 – si-1) 7: m2 <- (ei – si) 8: if (m2 – m1 ) < maxDist then 9: merge(Ii, Ii-1) 10: doMerge ← True 11: break 12: end if 13: end for 14: end while 15: for i = 1 to len(I) do 16: if (ei – si) < minMicroSz then 17: remove(Ii) 18: end if 19: end for |