A New Hybridized Rigid-Elastic Multiresolution Algorithm for Medical Image Registration
Xingang Liu, Wufan Chen, Guangjie Chen · 2005
We present a new hybridized rigid-elastic multiresolution algorithm of registration for medical image. The rigid deformation is achieved by a global affine transformation function composed of B-spline, in which the knots of B-spline are the four vertexes of the medical image. If we increase the quantity of the knots of the B-spline along the X and Y coordinates axes, the deformation function will be more complex and "elastic", which can complete elastic matching to align the local deformation in the medical image. In 16 groups of contrast tests, the hybridized algorithm achieves a total accuracy rate of 81.25%, while Kybic's elastic algorithm only achieves a total accuracy rate of 31.25%. It is proved that the hybridized algorithm is more robust than Kybic's elastic algorithm of registration.