When measured scanning electron microscope (SEM) image contours are used to generate or verify optical proximity correction (OPC) models, it is important that the contours are correctly aligned. If this is not the case, alignment errors will negatively impact the OPC model quality or its verification result. We present the approach we have developed to accurately align SEM contours for a variety of structure types. We discuss what one should align them to and how to do this, making the distinction between the case where the contours are to be used for verification of an already existing OPC model, or for generation of an OPC model. We show that our approach to contour alignment also offers the possibility to do contour averaging. We demonstrate the validity and limitations of our approach for a large number of SEM contours, taken from a 22-nm random-logic poly-type application.