Elastic matching versus rigid matching by use of dynamic programming

An elastic matching method is proposed, based on dynamic programming, to take account of distortions in images. It makes use of the fact that dynamic programming provides not only a global transformation, but also one-to-one correspondences between primitives, and therefore carries much more information about geometrical deformation. A deformation model that provides a much more accurate matching result is then estimated by using a minimum-mean-square criterion.<
>