lk = Lee & Kuo wh = Worthington & Hancock hw1 = Haines & Wilson, version 1 (eccv) hw3 = Haines & Wilson, version 3 lk was done with 150000 iterations. wh was done with 25000 iterations. bard: lk=8h,54m wh=44m,18s hw3=1m,33s head: lk=20h,4m wh=1h,50m hw3=3m,14s sunev: lk=22h,42m wh=1h,53m hw3=4m,17s venus: lk=21h,22m wh=1h,46m hw3=4m,31s (Core 2 Duo 2Ghz, 1gig ram, single thread implimentation.) -lk failed for sunev. One gets the impression that running lk for longer would get better results, but only marginally as its convergance is unbearably slow.