Utilisation of Radial Basis Functions (RBF) for reconstruc tion of damaged images became common technique nowadays. This paper deals with computation and utilisatio n of auxiliary points in order to further increase the ability of RBF to restore damaged areas in image. Our goal was to achieve the best possible results in acceptable time of computation. We put stress mainly on cases, where the image is heavily damaged e.g. by extreme noise. In these cases our new proposed approach achieved very usable r esults that even surpassed our expectations.