Abstract There is obvious need in the mapping community for significant advance in the level of automated stereo compilation . Computer-based digital compilation provides the reflexibity and sophistication necessary for this advance. Domains of mapping application range from natural terrain to cultural sites, and these both present extreme difficulties for traditional area-based cross-correlation mapping techniques. Digital compilation comes in two basic flavors: area-based and feature-based. Recent results in feature-based matching have shown this to be, generally, a superior approach, and indicate that this will be an important source of advance for mapping capability [2, 10, 12]. We discuss here results and current of research in feature-based surface mapping at Stanford. Our discussion, while focussing on the analysis of stereo pairs, also describes research in 3-D surface inference from single images and techniques for using single image analysis to aid in stereo mapping.