In this study we present a method to build 3D corneal atlases of different populations using a registration step based on inter-surface volume minimization. First the construction method is presented. It is based on a global factor computation in order to minimize the volume between several surfaces. Then the significance of the choice of the matching step is shown with the comparison of two atlases expected to be nearly identical: male vs. female corneas. Finally two atlases are presented and compared, for two different populations: right myopic and hyperopic eyes. Our study demonstrates that the matching step is crucial to correctly compare two surfaces, and shows two clinical applications of this method.