This paper presents the first known image reconstruction of complex probe tip shapes and the removal of those shapes from reentrant topologies that are encountered in critical dimension (CD) measurements. Algorithm improvements are described that enable reentrant image reconstruction. In addition, a solution is presented which eliminates image artifacts that result from noise when using Legendre transform or '' slope-matching '' reconstruction techniques. The new methods are compared to existing technology and demonstrated on reentrant features. Although demonstrated with two-dimensional profiles, the methods are readily extendable to three-dimensional morphologies. Tip wear effects on CD measurement are investigated and compared with the previous state-of-the-art method ('' tip width subtraction '') and fully reconstructed images, clearly showing superior measurement stability for the image reconstruction method. Finally, CD measurements derived from reconstructed images are compared directly with Hitachi S4000 and S5000 cross-section SEM (X-SEM) data. The results show a close match between the SEM and CD AFM data. (c) 2005 American Vacuum Society.