Main /
## LorusoTransforms## Estimating 3-D Rigid Body Transformations : A Comparison Of Four Algorithms## A. Lorusso, D.W. Eggert, and R.B. Fisher, Machine Vision Applications, 9 (5/6), pp.272-290, 1997.The computer vision field is driven by the idea of novelty, and the underlying assumption that different algorithmic approaches will necessarily result in quantitatively different levels of performance (accuracy, repeatability etc.). This paper, as the title implies, focusses on the main issue of comparing a group of object alignment algorithms which for several years had been viewed as competitors in their field. The final results however demonstrate quite clearly that the performance of these methods were indistinguishable at the level of machine accuracy. The reason being that all of them are based upon the same set of underlying statistical assumptions. This is a very important lesson. Fundamental differences between well implemented algorithms are due to the assumptions made NOT the algorithmic approach. Two apparently different algorithms will give near identical performance if they are based upon the same statistical principles. Understanding what constitutes a genuinely new approach, with performance different to existing techniques, requires us to relate algorithms at the level of assumptions. NAT 8/3/2005 |

Page last modified on March 05, 2012, at 03:54 PM