Matching topographic structures in stereo vision

Stereo vision is important m determining three-dimensional positions of tile visible surfaces in images. Most existing stereo matching methods involve matching low-level image features such as edges. A stereo matching technique based on matching high-level image features is described in this paper. The high-level image features extracted are topographic structures, arc segments and region segments. Matching results in stereo images are also discussed. Key wordw Stereo matching, feature extraction, edges, topographic structures, computer vision.