A supervised learning approach to landmark-based elastic biomedical image registration and interpolation
Mark P. Wachowiak, Renata Smolikova, Jacek M. Żurada, Adel Said Elmaghraby · 2003
Biomedical image registration often requires local elastic matching after initial global alignment. Due to their universal approximation property, neural networks may be used for landmark-based elastic registration. A supervised learning approach using backpropagation, Bayesian regularization, Gauss-sigmoid networks, and radial basis function networks is presented for 2D elastic registration.