The study investigates the use of remote sensing techniques in extracting accurate 3D highway geometric parameters using a case study in Hong Kong. A stereo-pair of Very High Resolution (VHR) IKONOS satellite images (1-m spatial resolution) covering Hong Kong Island and the Kowloon Peninsula are first geo-referenced using a polynomial-based generic pushbroom model and a sufficient number of Ground Control Points (GCPs) acquired by static Global Positioning System (GPS) measurements. The highway geometric parameters are extracted semi- automatically using stereoscopic measurements and the geometric accuracy of the alignments were assessed by comparing the same alignments extracted from stereo aerial images delivered in 10-cm spatial resolution. The Root Mean Squared (RMS) errors in both vertical and horizontal directions of the stereo IKONOS images were found to be within 0.8 m. With this accuracy, the geometric parameters of an interchange (radius and center of the horizontal curve, and length and curvature of the vertical curve) in the study area are extracted from both the IKONOS and aerial datasets. It is found that there is only less than 0.5 m difference between the radii extracted from the two datasets. For the vertical alignments, the measurements on the same loop ramp derived from the stereo aerial and IKONOS images are fitted to a symmetrical parabolic curve. The results show that the geometric parameters extracted from the stereo IKONOS images are very close to those extracted by the stereo aerial images. The findings of this paper can generally contribute to the knowledge-based components of geographic information system (GIS) database and to intelligent transportation systems (ITS).