Abstract The correction of inter-channel image distortions in multispectral imaging is the key process for high-precision applications. In this work, image distortion in a special filter wheel camera with sensor actuation based focus adjustment is investigated. Based on the analysis of the optical path, error compensation models at different levels are derived and qualitatively compared with each other. The experimental evaluation of the models shows that the proposed extended radial-tangential model achieves the highest image registration accuracy, especially in the marginal image areas. Furthermore, the stability of these algorithms is evaluated with repeated measurements.